The Nature of Conscious Experience


Reality, Meaning and Knowledge

    An account of the world and our place in it

                             Jonathan CW Edwards 

Hyacinth Macaws


1 Introduction

1.1 The need to rethink

1.2 A note on context

1.3 A note on method

Part I: An account of the world

2 Reality

2.4 ‘Direct’ versus ‘indirect’ perception

2..5 What is it to be ‘evenly blue’?

2.6 Different blue dispositions

2.7 Difficulties with dispositions

2.8 What are ‘objects’ and are there any?

2.9 Signs will depend on the way of receiving

2.10 Space, time and disposition

2.11 Must all aspects of the outside world be dispositional?

2.12 Qualities as purely proximal and asymmetrically relational

2.13 A coherent view of physics

2.14 Summary

3 Complementarity: illustrations and explorations

3.15 What do gravity and knight’s moves look like?

3.16 Cause: Hume, Mackie, modern physics and Leibniz

3.17 Causal chains and unknown manifestations

3.18 Reality check: do we need manifestations to …?

3.19 Are manifestations causes?

3.20 Dynamic and manifest time

3.21 Manifest movement and motion

3.22 No empty space inside an atom

3.23 Rainbows and flashing cat’s eyes

3.24 Pragmatism and Positivism

3.25 Parts and wholes: mereology

3.26 Emergence

3.27 Fermions, bosons and Descartes’s space.

3.28 Man as part of his world

3.29 God and the God’s eye view

3.30 Summary: a sequential universe

4 Meaning

4.31 Language as an internal process

4.32 Meaning by, meaning to and other meanings

4.33 Why meaning is never constant

4.34 Meanings as dynamics and as manifestations

4.35 Dynamic, manifest and intermediate meanings

4.36 Two complementarities

4.37 Do words and their meanings exist?

4.38 Manifest meanings must be immediate

4.39 Pseudoconcepts and pseudodynamics

4.40 Aboutness

4.41 How meanings might be set up

4.42 Mordant loops

4.43 Variable specificity: from 50% to 99%.

4.44 Nouns and verbs

4.45 How can words be markers for both dynamics and manifestations?

4.46 Truth

4.47 Summary

5 Knowledge

5.48 What is knowledge?

5.49 What knowledge has been said to be

5.50 The need for multiple pathways

5.51 Roles for path complexity

5.52 Differences between us and animals

5.53 Knowing machines and knowers

5.54 The problem of the knower knowing

5.55 A return to the problems of systems and functional role

5.56 The form of manifest knowings

5.57 Knowing and physics

5.58 Self-knowledge

5.59 Self-knowledge of a knower

5.60 Summary

Part II: The Human Subject

6 The nature of the human subject

6.61 Subjects and creatures

6.62 Options for an inner subject

6.63 Manifestations to: the role of input.

6.64 Homuncular problems: real and false

6.65 Everything coming together and real-time experience

6.66 ‘Orthodox’ views of the human subject

6.67 Global workspace theory

6.68 Integrated information theory

6.69 Neuroanatomical workspace theory

6.70 What would the output role of customary subjects be?

6.71 The arguments for multiple subjects

6.72 The subject as integrating unit

6.73 The individual cell as subject: historical ideas

6.74 Which cells might be customary subjects?

6.75 What sort of entity would provide a cellular subject?

6.76 General requirements for an acoustic cellular subject

6.77 Specific models for an acoustic subject

6.78 Marrying experience with fundamental physics

6.79 Unpacking the binding problems

6.80 The syntax of binding

6.81 Possible dynamics for binding

6.82 Dimensions and degrees of freedom

6.83 How long is an experience?

6.84 How do we talk about experience?

6.85 The difference between knowing and self-knowing

6.86 Some zombic considerations

6.87 How do we talk about richness?

6.88 The problem of manifest evaluation

6.89 A workable panexperientialism

6.90 Summary: a complementarity-based view

Part III

7 Descartes: a reappraisal (to be added)

8 Leibniz: a reappraisal (to be added)

1. Introduction

1.1 The need to rethink

As children, we grow up happy to think of the world as consisting of stuff, things, animals, people and the like. We variously think of these in terms of being, changing, doing, desiring or believing. We think we can see and hear things and know about them, and that, although we may be mistaken, there are truths about how things are.

On the other hand, with time, and in particular if involved in science or philosophy, many of us come to admit that our talk of stuff and things, or knowledge and truth, raises many problems. When examined minutely, stuff seems to lose the features that made it stuff-like in the first place – hardness, opacity, colour, or taste. It becomes less clear what deserves to be called a thing. Do a mountain, a rainbow, a tower block apartment or a flower fully qualify? What does? Do we have a clear idea of what we mean by something being coloured (what if, like a rainbow, it neither reflects nor emits light)? Do we have a clear idea of what a person is, rather than a body or a soul? What is it for such an entity to have beliefs or to know? Knowing can be challenged both in terms of the possibility of certainty and of what knowledge could actually be. Similarly, we may doubt the existence of truths about the world if such truths can only be defined in relative terms.

There are many ways of responding to these difficulties. Ignoring them can work not just in daily life but in many of the branches of science that made great progress in the twentieth century. However, on fundamental issues, and on issues to do with perception, thinking and language, which will form the biggest challenges for the twenty-first century, ignoring the problems of the childhood view of knowing the truth about stuff and things is not an option. Nor, I think, can we ‘dissolve’ the problems, as suggested by Wittgenstein, simply by avoiding muddled language and restoring ordinary usage. There is certainly a lot of muddled language around, but the central problems relate to non-verbal ideas, like time, space, cause and effect. Just as when children, shown that blue and yellow paint make green but blue and yellow light make white, have to reset their non-verbal ideas associated with the word colour, we need to develop new ideas about the structure of the world in a wider sense.

Another approach, perhaps most associated with Immanuel Kant, is to accept that there are difficulties with the childhood view and conclude that our view of the world can only hope to be an approximation to reality since we are just not equipped to ‘know what the world is really like’. Variations on this position seem to be common, with the implication of some sort of glass ceiling to what we can know, with some unattainable position above the glass from which a God might know everything.

I am going to argue for a more optimistic and forward-moving position: that if we work through the metaphysical jungle of difficulties with the childhood view persistently enough we can come out the other side with a clear view that leaves nothing ‘unknowable’ in principle. The exercise is not pain free. It requires a ruthless reassessment of how we should view reality, meaning in language and knowledge. But it has pay-offs.

The common thread in all three cases is the recognition that each presents us with pairs of complementary aspects, that are incommensurable, in the sense that one cannot be reduced to the other. This complementarity is indirectly related to what Neils Bohr rather obscurely proposed for modern physics but I am interested in a much older and more general idea, surfacing repeatedly in history in Parmenides, Newton, Schopenhauer and many others. So I am not suggesting that it is new, but rather that if fully explored it pays greater dividends than may have been apparent.

This complementarity appears in many forms and ramifications but as far as I can see it stems ultimately from a distinction between what may be called dynamics and manifestation. In plain usage it is the distinction between:

What is ‘really going on’.


What is, was or otherwise might be ‘really here, now’.

The former is what we might call underlying reality: a reality of patterns of change in the world that we infer from appearances, with varying degrees of precision. The latter might be called apparent reality: a manifestation of underlying reality that is undeniably evident, but may not signify what it seems to signify. These are not, however, two different types of anything, but merely two conceptions of the same everything. Moreover, I believe they are as relevant to microphysics and cosmology as they are to human experience.

In even simpler terms, complementarity might be seen as between cause and effect. This will not do, however, because, as Aristotle pointed out, cause covers several different ideas. The complementary divide I shall propose is subtler, but perhaps in the end simpler, than cause and effect. I shall not attempt a concise definition at this point because the words needed have to be put into new contexts that will give them new meanings. Philosophers please note straight away that my use of manifestation is not quite that of Charles Martin and John Heil and their commentators; it cuts at different joints. Context–sensitivity of meaning in language is a key component of my proposed approach, not least because words behave in complementary ways on different occasions.

As I see it, in our natural languages, many words have at least two complementary meanings, one of which is implicit in any one ‘utterance’. Thus the statement that ‘The tomatoes are red.’ meaning that the goings on we call the tomatoes includes a tendency to reflect the sort of light that makes us have a sense of red, is the same as the statement that ‘The tomatoes are red.’ meaning that I have a sense of red tomatoes here. It is only in the context of a mixture of trichromat and colourblind people in a neuropsychology lab equipped for different condition of illuminating tomatoes that the difference between the two sentences starts to be obvious.

Complementarity within language has received relatively little attention but was highlighted by Wilfred Sellars and also by Rudolph Carnap. Sellars distinguished statements about either a ‘manifest image’ or a ‘scientific image’, which he saw as complementary. While this is related to the schema I am proposing I am not convinced that Sellars saw how deeply complementarity is buried in the structure of individual words within sentences.

This invisible embedding of complementarity in language may be almost as big a problem for sorting out a consistent world-view as the counterintuitive nature of complementarity itself. However, once recognised, it may also reveal deeper levels of sophistication in the ordinary man’s, if not the child’s, view of the world. The distinction between a dynamic meaning of reality as what is really going on from an apparent meaning of what is indubitably real in the sense of manifest here and now, will come as no great surprise to most people and crops up in the literature time and again. Schopenhauer’s Die Welt als Wille und Vorstellung (doubtfully translated as The World as Will and Representation) may not have captured the imagination of many, but that is the commonplace idea that he wanted to justify.

When it comes to knowing, the complementarity story gets still more complex and can only be hinted at here. The essential problem is that knowing does not just require receiving signals from the world, it requires a computational apparatus, or ‘knowing machine’, that can cross-correlate patterns of signals under different conditions (as in turning an object over in one’s hand) to make it possible to infer what dynamic structure in the world could give rise to such patterns-in-contexts. Worse still, to be justified, whatever ‘knower’ is informed of the conclusions of the knowing machine (whether or not they are the same thing) must have evidence for its machine’s reliability. There are reasons to think this is impossible. Processes that deserve the name ‘knowing’ do seem to exist, and indeed may have no theoretical limits, but I suspect they are not just based on the relation between known and knower we assume. There is a more complex set of relations between known and knowing machine, and knowing machine and knower that must be broken down into at least these two complementary components. Once separated, the historic paradoxes of ‘knowing’ may become tameable.

The idea that we have to consider the world in two complementary ways may seem untidy. I will try to argue not only that it amply justifies the extra complexity but also that once accepted it fits better with our intuition than a single aspect view (and rids modern physics of any special spookiness). We do think in terms of cause and effect being asymmetrical and incommensurable even if we overlap the two aspects in our thoughts and words at a level that is frequently implicit and unconscious.

Moreover, I do not see the goal as just metaphysical tidiness. I think a rigorous complementary approach may be crucial to the next phase of human learning. For four hundred years we have unraveled the dynamics of the outside world. The challenge now is to understand the relation of dynamics to manifestation within us: what it is to know such things, which in turn means understanding what it is to be a knower. As of 2012 I do not think there is a useful consensus on how to approach this. Neuroscience has revealed some of the operations of our knowing machines, like the line-recognising cells described by Hubel and Weisel. What it has failed to do is give any account of the knower that the knowing machine supplies. I suspect that is because the full implications of the complementarity inherent in our way of knowing have not been assimilated.

To summarise, and inevitably oversimplify, the view I want to explore, is that the so-called puzzle of the division between physical and mental is just the asymmetry between change and what you get from change, the existence of which even a child can understand. Without it, the world collapses to something like a ‘Matrix’ or the whim of Descartes’s evil demon. With it, you have a real world that hides nothing if we can read the signs right.

1.2 A note on context

Reading historical attempts to describe a broad view of the world it is hard not to be struck by the fact that those worth reading show no consistency whatever in format or adherence to the conventions of any particular discipline. They vary from Presocratic fragments of a few Greek words about walking into a river to William James’s hefty Principles of Psychology, via Newton’s Principia Mathematica and Whitehead’s Process and Reality. It seems there are no rules of form or content to follow, which is liberating but daunting. I will reach into a range of disciplines in what seems a helpful order. My only specific intent in terms of format is to present ideas in chunks of roughly 1000 words, for reasons that may become apparent.

Trying to understand the way the world really is is equally the aim of ‘science’ and of ‘philosophy’. In recent times there has been a divide between the disciplines and even a suggestion that philosophy can legitimately look for an account of the world alternative to that of science. For me that is just small-mindedness. A valid scientific account of the world must stand up to philosophical critique and vice versa. The only science we need an alternative to is bad science, and likewise for philosophy. Part of the problem is that at the interface, notably in the study of our perception of the world, there is an lot of bad science and bad philosophy about, a common feature of which is that scientists and philosophers both misrepresent what the others propose. So I can agree that both many neuroscientists and many contemporary philosophers have missed the point but not that either science or philosophy misses the point.

In the Analysis of Matter, Bertrand Russell notes that ‘It may be admitted that most of what has passed for philosophy would not have been very useful to the men of science; but that was chiefly because philosophy was no longer being created by men like Descartes and Leibniz, who were of supreme eminence in science as well. It may be hoped that this state of affairs is coming to an end.’ Seventy five years later the hope has not yet been fulfilled but maybe things will change. For me, the two people who were most successful in crossing the jungle of difficulties with childhood realism to a clear view were the same scientist-philosophers noted by Russell: René Descartes and Gottfried Leibniz.

Descartes’s account is often quoted, but to my mind usually misrepresented, in part owing to the influence of philosophers such as Gilbert Ryle, who had no experience of the business of building scientific ideas that Descartes was addressing in his Meditations. There are undoubtedly loose ends in Descartes’s approach, but I suspect that the Meditations still give valuable advice on how to form a robust view of the relation of man to the world. In terms of the basic art of thinking productively I think he is still way ahead of a large proportion of contemporary scientists and philosophers. Some specific aspects of his insights are, I think, essential to revisit.

Leibniz’s account is less often discussed and mostly with the implication that it is some sort of arbitrary fantasy. As David Wiggins not unfairly put it in response to my attempt at defence: ‘but, in the Monadology, Leibniz does not give any reasons for anything he says’. Yet, as someone having been guided to Leibniz by someone who noted that I had drawn much the same conclusions, he seems to say all that is necessary. The difficulty seems to be how to introduce, using the standard language of stuff, things, truth and knowledge, a different set of ideas that might work better (and, in Leibniz’s case, seem to presage developments in science 300 years later). It is intriguing that several people, including Russell and James, hint that they think Leibniz might prove to have been more right than wrong if only one could see what he really meant. That people have often not seen what he meant is suggested by the frequent running together of the mature ideas of his Monadology with earlier, superseded ideas, and a range of other howlers about ‘psychophysical parallelism’ and ‘blind monads’, which are not Leibniz.

Although I cannot claim to know what either Descartes or Leibniz thought at the time, I have some ideas about the most productive interpretations of their work, and will explore these further later.

There is no doubt that a number of more recent scientists and philosophers, including Eddington, Pauli, Feynmann, James, Russell and Chomsky, have written interestingly on the issues of how we should view reality and our knowledge of it. However, I have a distinct sense that whereas Leibniz had come out of the other side of the jungle of difficulties with a simple and comprehensive schema, more recent commentators show various signs of still being entangled in the process of reaching a clear view. The cynical view would be that everyone attempting to improve on Leibniz, including Kant, has merely demonstrated their failure to fully reach and understand the strength of his position. A more measured and charitable view might be that there is indeed obscurity in Leibniz, and a vagueness that leaves room for genuine clarification as much as it does for overgenerous interpretation. Either way, I do see a hesitancy in recent accounts.

Bertrand Russell perhaps did more than any other recent philosopher to produce an account that was informed in depth by scientific theory. Sadly, his work seems to have been eclipsed by a sterile debate between poorly defined positions usually described as ‘materialism’ or ‘physicalism’ on the one hand and various forms of ‘dualism’ on the other. Russell came to a position he called neutral monism that appeals to many people’s common sense. The idea is that the mental and the physical are just two different aspects of a single sort of ‘world structure’. Everything, including human ‘minds’ follows the same set of mathematical causal laws, but at the same time the events that ensue have an intrinsic quality that, for us at least, is what we call our experience, consciousness or mentality. I like Russell’s general framework but I think he makes some serious errors in his final analysis that belie a failure to appreciate just how incommensurable the dynamic and manifest aspects of reality are. As a result his monism had no real pay-off. But he is a good scene-setter.

I will return to these historic figures and a number of others in due course but at this point I would like just to make a short comment on method and then move on to the basic reasons for taking complementarity seriously.

1.3 A note on method

The primary obstacle to a world view based on complementarity is that it is counterintuitive. Even for the narrow sense of complementarity found in quantum theory there is a feeling that things are a bit desperate if we have to accept something so peculiar as having two mutually exclusive descriptions of everything. To suggest that a more general form of complementarity pervades our daily lives, our language and our thoughts so that not just ‘electron’ but words as diverse as space, time, red and belief each have at least two meanings that are not just quite different but totally disanalogous can presumably only be worse. Much of what I want to say is about the extreme and perhaps insuperable difficulty most people have in taking such an idea fully on board. If Russell did not do so, despite being highly motivated, intelligent and informed in detail about physics, it cannot be a trivial task.

One can of course defend the counterintuitive on the grounds that mere intuition has no place in any serious enquiry. Arguably, the whole point of both science and philosophy is to show that intuitions often run into trouble when pushed to the limit and are better replaced by more subtle but consistently workable ideas. Adherence to certain intuitions come-what-may, which seems to be an approach at least as prevalent amongst contemporary professional philosophers as in any previous era, seems unlikely to get very far. Nonetheless, new ideas, like the earth being round or orbiting the sun, are only more workable if they make predictions, and making predictions from ideas usually cannot be done without the help of intuitions. There is a story that everyone thought that a spaceship travelling almost at the speed of light would look squished in one direction until Roger Penrose pointed out that it would look squished the other way. If true, Penrose had a better ‘grasp’ of the relation of the maths to the reality and such a grasp comes pretty close to having the right intuitions about it. Some intuitions you cannot do without. These may be different from the first sort in some identifiable way. They tend to involve an element of rational inference. However, the literature is not clear on this and I suspect for good reason.

The way to handle the problem of intuition seems to me to be pretty much as recommended by Descartes. At some point the accumulation of paradoxes raised by empirical observations makes it necessary to challenge everything with a sceptical stance. You then identify those intuitions that seem to generate consistent workability: not just survival in the face of one or two counterarguments but survival in the face of being turned around to work in every possible way day in and day out in all sorts of practical contexts. Finally, you totally rebuild a framework of ideas, replacing intuitions that lead to inconsistency with ideas based on intuitions found to generate consistent workability, even if they may appear more abstract than those they replace.

To this I would add a very desirable bonus for any ideas used as replacements: that they end up satisfying certain basic intuitions better than their predecessors. I will use as an analogy the way people conceive what it is they see in a mirror. Most people have the intuition that their faces are reversed right to left in a mirror but would deny that they are reversed top to bottom. This is odd because whatever a mirror does to right and left it ought to do to up and down, shouldn’t it? If we make the unfamiliar (yet rather sensible-sounding) suggestion that the mirror only reverses front to back, we in fact have a consistent account. The right of the face is seen on the right, the top at the top and so on. It fits with the intuitions that right to left and up and down relations should be on a par and that reversing front to back makes sense as what mirrors do do. The reason for the false intuition can also be seen. Your face in the mirror could equally be ‘you’ reversed right to left and rotated around a vertical axis or ‘you’ reversed up and down and rotated around a right to left axis. We just assume the former because we do not often go upside down!

Descartes claimed that at the end of his sceptical journey he could establish ‘clear and distinct’ ideas that he could rely on. Here I would be more cautious and suggest that however far we may have got in replacing inconsistent intuitions with apparently consistent ones there are likely to be several more levels of replacement to go. Like at the opticians, what seems ‘clear’ may, surprisingly, be made even more clear with another tweak of a lens. In fact I think that if we are prepared to apply Descartes’s approach repeatedly and indefinitely we are likely to get through a few more interesting levels quite soon.

Thus, what I would call ‘common sense’ and give a stamp of ultimate approval is the belief that we should base our world view on the use of those intuitions that stand up to test under the most stringent conditions time after time while rejecting those intuitions that fail. Unlike Russell, I do not see common sense as something superseded by physics. For me, common sense is the peculiarly human skill of doing or thinking something a bit more complicated in order to get a more reliable result. If the eggs are on top of the potatoes in the supermarket trolley it is common sense not to put the eggs into the shopping bag first and then the potatoes. You put the eggs to one side and put the potatoes in first. That sort of common sense goes on being paramount right through to interpreting the most obscure physics.

The double-edged nature of intuition can also be seen in language use. The intuition that languages have words with a single definable shared ‘ordinary’ usage is, as we shall see, unworkable. However, the intuition that in the two sentences ‘What time shall we meet?’ and ‘How much time will we have?’ the word time must have a different meaning, may take us a long way. The intuition that giving each word a very precise definition might help has paid few dividends. As Charles Travis has pointed out, the effect of context on word meaning is almost limitless. The intuition that the way to make the meanings of words clear is to be very careful about how we build them into statements, paragraphs and documents may be more fruitful.

What I hope to show is that if we are prepared to question our deepest intuitions about our thoughts and our words to start with, as Descartes advised, and then sift carefully through all the evidence in front of us, there may be a way to return to something that can satisfy our deepest intuitions after all. What would be comforting would be to find oneself at the top of a hill looking down on a familiar land, from a new perspective, but still a familiar land and maybe even more familiar through understanding how closely we know it.

Part I: An account of the world

2 Reality

2.4 ‘Direct’ versus ‘indirect’ perception

I am starting with perception partly for its own sake but also because I think it is the most direct route to an understanding of what the job of physics really is. Discussion of perception in philosophy focuses on whether we perceive the world directly or indirectly, with many going for direct. In science, the debate is considered settled in favour of indirect, but in a different sense from that used by philosophers. I think this debate is full of language muddles and misses the real issues entirely. Nevertheless, it is useful to follow through some of the basic positions taken on it because they show the sorts of arguments that people often feel need to be raised.

The intuitive or ‘naïve’ view is said to be that we think we perceive things directly and transparently. Nothing seems to intervene between ‘me’ and the table I am perceiving now. I have no sense of any mechanism. The brown colour I see seems to be the colour of the table itself, not some internal sign of a brown colour.

The problem with the naïve view is that if we make any attempt to analyse how the human sensory apparatus gathers information about the world we have to come to the conclusion that there are very complex mechanisms between the table and ‘me’ involving all sorts of chemical and electrical media and collation and cross-referencing of signals. Moreover, we have to accept that the brown colour that is manifest is indeed our nervous system’s sign for a complex reflectance disposition of the table’s surface. I will enlarge on these issues shortly, but for the moment I will simply point out that unless one accepts that perception is indirect in this sense trying to understand it in biological terms hits a brick wall. Not only do you hit empirical contradiction but after a few examples it becomes clear that the naïve view never could make sense.

Philosophers may accept this but claim that perception is ‘direct’ in another sense: that I do not see what the table is like via seeing what some interposed representation of it is like.

The first objection made to such mediation by ‘a representation’ is that we have no reason to believe in some extra metaphysical category of ‘sense data’, making up representations floating in some sort of internal experiential space for us to peer at, available for being sensed in a way that ‘ordinary stuff’ is not. I will try and explain why I think the objection to sense data is groundless, although I admit that in order to find a convincing response one has to bite the bullet of what sense data might be in a way that current neurobiology totally fails to do.

I think we do need a metaphysical framework with a place for sense data. However, that does not mean that there need be ‘another, different, sort of stuff’. What I will suggest is that the relations between sense data and human subjects are based on the same rules that cover the relations of tables, light, air, lenses and suchlike. All such relations involve the determinate outcomes or results of packets of change as manifest to a further packet of change. Nowhere is there any ‘stuff’ other than packets of change and their manifest outcomes. Changes and their outcomes are not different things, since neither can exist without the other. The subtlety is that manifest outcomes are aspects of both the occasion of change that determines them and the next occasion of change that they help to determine – hence creating the chain we call causation. Sense data are only unusual in that they are the outcomes directly manifest to the packet of change that is a human subject. These claims may appear obscure at this point, not least because they are oversimplified, but they will be explored in detail in due course.

The direct realists’ second main objection to the idea of indirect perception is that when someone says ‘I see X’ or ‘I perceive X’ X must be an object in the world such as a table because that is what the words ‘see’ and ‘perceive’ ordinarily mean. The objection is that to say that I ‘see’ a representation in my head is to misuse the word. What this shows, I think, is not that indirect theories are wrong, but rather that the ordinary English usage of words like ‘see’, ‘perceive’ and ‘representation’ may not be very helpful, at least not without careful contextual qualification, for understanding our true relation to the world. We are not seeing what the sense data are like, but it is very doubtful that we are ‘seeing what X is like’ either. Although philosophers like to pride themselves on the rigour of their argument my impression is that they are often prepared to hang on to simplistic usages of words long after the ordinary person with common sense would have accepted that such usage is not really up to the needs of rigorous argument. The indirect perception theorist does not want to imply that we perceive sense data in the sense that we think we perceive things in the world. Rather, the claim is simply that perceiving things in the world is mediated by some closer relation involving sense data.

A third objection to indirect theories of perception is that they rely on what is called the argument from illusion and that it is wrong to extrapolate from unusual illusory situations to an ordinary case like seeing a table. I think this objection is little more than a philosophical ostrich strategy since the arguments for indirect perception go far wider than illusion. The argument from illusion is merely a simple way of introducing the idea that the naïve view might be problematic; if we can see things moving that do not move, or the size of things apparently changing with our viewpoint, or ‘objects’ in dreams that are not there, then maybe our percepts are not necessarily directly related to what is going on in the world outside us. If we need something like sense data in these situations maybe they mediate percepts that are not illusory too.

There is a much wider and more powerful argument against the idea that we see directly, without internal data, what things are like. Paradoxically, it is also an argument against common versions of indirect perception that postulate internal representations that are somehow ‘copies’ ‘maps’ or ‘analogues’ of the outside world in the way that diagrams or models are.

It is not that illusions are exceptions in which we have percepts or images that are not what the world is really like. The point is that it is completely absurd to think that the appearances of percepts could be ‘like’, or resemble, the bits of the world they signify, even in terms of things like shape. Locke seems to get this wrong. The situation bears no relation to that of a map, diagram or model. The first point is that we have no reason to think that an effect should ‘resemble’ goings on, or patterns of change that, directly or indirectly, give rise to it. The second point, which I am gradually arguing towards, is that real goings on do not, in themselves, on their own, have any appearances. They are not ‘like anything’ in that sense. (Note that this is a sense that cannot be constructed from the individual words taken out of context. One pattern of dynamics can be similar to another but that is not what we mean by ‘like something’ in the manifest sense.) We confuse the idea of ‘what it would be like if we were there’ with what is going on when we are not there. The dynamic patterns determine (indirectly) the patterns of appearances or manifestations to us but these are not the appearances of the goings on (how could a sweet taste be what the goings on in sugar and saccharine molecules are like?). At this point the distinction may still seem obscure but it may be clearer following an illustration.

Key point: Neither the idea of ‘direct’ perception nor that of ‘indirect’ perception via internal map-like representations are helpful accounts of our relation to the world. Both wrongly assume that the manifest patterns of experience should be ‘like’ the dynamic patterns that give rise to them.

2.5 What is it to be ‘evenly blue’?

If I turn to an ordinary perceptual situation it may become clear why I think the shortcomings of both direct and representational accounts of perception are based on much more pervasive considerations. I apologise to physiologists who take the following as obvious, but they may find a few novel twists at the end.

Having turned my gaze away from the table, I am currently looking at our blue enameled AGA cooker. As I pass my gaze over the cooker front I sense that the entire front is the same intense rich dark cobalt blue. There is not a hint of an inconsistency in the colour of the enamel, no fading, chipping, discolouration with use or whatever. Yet, if I introspect a little I can see that my opinion about this constant colour is derived from a totally inconstant spectrum of light reaching my eye from each part of the cooker front. Some areas are in deep shadow, reflecting almost no light. In other places the shiny enamel reflects bright highlights. Yet other areas reflect yellowish tinges from the beech flooring or white grid patterns of light from a French window frame.

What in fact I mean when I say that I see that the cooker front is a constant blue is that I have unconsciously inferred that every part of the cooker front instantiates the same disposition to reflect certain wavelengths of light preferentially. I must immediately add that ‘inferred’ will require as much unpicking as ‘see’. However, all I require the word to imply here is a certain sort of indirectness requiring outcomes of comparisons. I will return to its full meaning shortly, and in more detail in the section under knowledge. I use the awkward word instantiates to avoid a potential linguistic pitfall. I am implying that with every bit of cooker front you get the same disposition operating – no more and no less.

I am not being directly presented with an intrinsic quality of an object, but with a sign (or signs) of a relational disposition. Much of what my eyes take in relates to the cooker’s relations to other things. The even blue I am ‘seeing’ is a sign of an inferred disposition of a surface. Moreover, this disposition is complex. It includes both a tendency to preferential scattering reflection of short wavelength light with highly efficient absorption of other wavelengths and a different tendency to reflect, without scatter, all wavelengths equally, in the way that a mirror does (the shinyness). As such I see it as an evenly shiny deep blue cooker.

Of note here is that I do not see paler blue where there is a superimposed faint reflection of a white door frame. I see a deep blue cooker reflecting a white door frame. I am unable to see the two components of colour as ‘paler blue’.

Most paradoxically, if I try to find and focus on what one might assume to be ‘patches uncomplicated by reflections that are the plain blue I assume the rest is’ I find I cannot get any sense of what colour they are. Concentrating on what in the corner of the eye seemed a nice blue bit makes the local blueness vanish: not the general blueness of the cooker (which still hovers in the background) but the local ‘blue feel’ itself. The raw sense of blue often assumed to be a direct and ‘uninterpreted’ sensation turns out only to exist as an unconscious (or ‘preconscious’) interpretation of a general tendency. It is easier to appreciate the need for this with a dark shiny surface than a pale matt surface. Nevertheless, it has nothing to do with ‘illusions’. It tells us about the way normal vision works. If you look around most of what you see is likely to be more like a shiny cooker than an evenly illuminated matt surface. The ‘transparency’ of perception that is the apparent absence of anything between us and the object is in fact an ‘opacity’ that hides from the human subject the brain processes that generate the manifestations it beholds.

The crux is that this is in no way seeing directly how something is. I see an evenly shiny blue cooker. However, to do so I must rely on a highly sophisticated computational apparatus in my sensory pathways that can ‘infer’ from present and past inputs that a very complex pattern of retinal stimulations equates to an evenly blue shiny cooker. That must mean that somewhere in my head is a subdomain of my nervous system S that is receiving an output from such a computational system that signals ‘evenly blue shiny cooker’ to S; otherwise why would the computational system bother to make this inference?

As I hope to make clear, ‘blue’ is not quite the property of things that our ordinary language suggests it is. The concept of blue we grow up with, which we take to be valid and a sound basis for statements like ‘the cooker is blue’ turns out not to have the referent (something it refers to) we thought it had in the real world. We have to recast it in terms of new concepts that match with what is really going on. Even in terms of disposition, as I shall come to next, blue is not one but a collection of quite different dispositions, which happen to share mental signs. Childhood blue is what I shall call a pseudoconcept. In exploring what is really going on in the world pseudoconcepts crop up with great regularity, which is my reason for being sceptical about the philosophical idea that we should respect ordinary language use. In this field, much of it turns out to have no real referent.

I have already slipped in a number of terms without too much definition, including goings on, change, dynamics, disposition, cause and manifestation. This is deliberate because the meanings of these words that I think most useful can only emerge once I have brought together a range of background issues. Moreover, the meanings are going to be multiple and context dependent. I could introduce new words but I think that may obfuscate. What I can say now is that dispositions are the same as causal powers or tendencies. Dynamics are patterns of change that are instances of operation of causal powers. However, I will subsequently suggest a subtle but important relation between aggregate dynamics, or everyday goings on, and individual dynamic packets. Manifestations are how things appear, or what they are like to something but, again, I shall introduce a relation between two possible meanings with different scopes. Manifestations may also be instances of operation of causal powers, which is initially confusing. I will try to make clear how I am using these terms but I will use them differently in different places, as I believe is inevitable given the structure of natural language. Hopefully the non-verbal ideas will emerge.

Key point: The ‘rawest’ of our sensations are based on inferences about dynamic patterns in the world. They signify patterns of operation of causal dispositions, not intrinsic qualities of objects.

2.6 Different blue dispositions

When I say that for the cooker to be dark blue is for the cooker’s surface to instantiate a disposition to reflect light a certain way some people may feel uneasy. There is an instinctive sense that to be blue is not just a tendency or disposition to reflect light but actually to do some blue reflecting. However, if asked if my cooker is blue when I go out of the room or on a moonless night when it is in complete darkness most people will say yes, it is still a blue cooker. If lit by a sodium street light so that it seemed black with yellow highlights the answer to the question ‘Is your cooker black?’ would be ‘No, actually it’s blue.” To be blue only requires a disposition. To be seen to be blue requires that disposition to operate in such a way that we can infer the disposition but has only a very indirect relationship to the amount of blue light arriving at any particular point on the retina.

Part of the problem here is the problem of the multiple meanings of ‘The tomatoes are red.’ in the introduction. If this is a statement about a tendency of the tomatoes, whether the tendency is being cashed out by red light rays bouncing off is not relevant. We have an ordinary language phrase to illustrate this. We can say ‘No, the tomatoes only look red, they are yellow, but cunningly lit with red lights.’ The various possible situations of light bouncing off make no difference to being red. Moreover, if they are red, we will experience them as red in all but the most seriously confusing or inadequate viewing conditions even if 90% of the light they reflect is blue.

When we say tomatoes only look red we seem to be saying that we realise that the conditions are so atypical that we tend to infer the wrong reflectance disposition for the surface of the tomatoes. We are still talking in terms of reflectance. However, there is also the situation where I might be blindfolded with a mask and escorted to an unknown place and laid on my back, at which point someone opens a tiny hole in the mask over one eye and asks ‘what do you see’. I might just answer ‘blue’, not being able to relate the colour to any particular object or even material. It might be sky or ceiling or a coloured card. This does suggest that we can see blue directly. Thus, one might argue that there is a difference between knowing something is blue, even in the dark, and seeing it to be blue and that the latter is crucially dependent on getting a raw blue signal right at the beginning of the sensory process in the retina. But, as indicated above with the blue of the cooker disappearing when looked at hard, it is not that simple. A vast literature is now available in neuropsychology that shows that raw input signals to the retina can be completely ignored (if you stare long enough at any static object it disappears) or converted into a whole range of ‘raw feels’ depending on how later neural processes collate these signals with others.

The explanation for this counterintuitive situation must be that ‘seeing something as blue’ and ‘seeing blue’ involve different collation processes prior to something being manifest to the subject. In the second case blue does not seem to be a property of anything. We explain this by saying that the light coming in to the eye is blue, but if this is a property of the light it is completely disanalogous to the property of the cooker (and again disanalogous to that of a blue light emitting diode). What this emphasises is that perception involves inferences that have to draw on relational aspects of patterns of input signals. If we can infer the presence of a cooker we can then infer a disposition for the cooker. The sense of blueness is ‘about’ the cooker. If we cannot infer anything to attribute blueness to we have to infer a disposition that operates at some other point in the input causal chain. Not being able to trace back the causal story to an object we have to infer a disposition about the interaction between signals reaching the eye and our sensory receptors – the ‘blueness’ of light.

This raises two points. Some of the signs sent out from our collating apparatus may turn out to be generated in response to several quite different dispositions because these are often associated in causal chains. Moreover, these signs in themselves may give no indication what disposition they actually signify. My earlier suggestion that the collating apparatus infers dispositions was intended to entail a very limited interpretation of the word infer. It does not imply ‘action’ by an ‘agent’, both of which terms I consider to be misleading and best avoided. Nor does it imply any ‘theory’ or ‘understanding’ in the sense that the disposition need be describable in other terms which allow some sort of independent verification of the collation process. As for so-called ‘theory of mind’ the inferences children make about dispositions of coloured objects involve no theory about light, reflection or whatever. All that is required is that they have a collating system that generates outputs that correlate to dispositional patterns in the world. Some would argue that the generation of signs in response to patterns of disposition by cross-collation of inputs is not true inference. That there is more to some sorts of inference I would not deny but exactly what may be poorly understood (I will return to this issue in the section on knowledge).

While being sceptical we might perhaps argue that the shininess of the cooker is only a property of the most superficial layer of the enamel and the blue is a property of the pigment embedded in the transparent vitreous matrix of the paint. Perhaps the cooker is in fact the deep blue colour and the highlights created by the shiny surface just interfere with our appreciation of the blueness in some places. However, the depth of the blue and the highlights are interdependent aspects of a single complex reflectance disposition. If there were no shine the blue would not be deep; blood red marble looks white before it is polished.

The question of different layers having different dispositions does, nevertheless, raise a different point. The deep blue is a dispositional property of the enamel, not of the rest of the cooker, which is mostly cast iron and stainless steel. A ‘white car’ is a collection of bits and pieces of all sorts of colours including a very thin layer of white cellulose paint on some of the outside. The front of my AGA cooker that I took to be an even deep blue is in fact made up of a frame plate on to which are loosely hinged three door plates. They form an ‘object’ with the rest of the cooker only in the sense that they were all delivered from the same AGA warehouse. Other parts of the ‘cooker’ are much more firmly bonded to the floor, wall or fitted cupboards than the door plates are to the frame plate.

So, when I say I see a blue cooker it seems that what I mean is that I am inferring various dispositional features manifest to me in one way or another as part of the goings on within certain spatial and temporal domains. I file these away under ‘qualities of objects’ but neither ‘quality’ nor ‘object’ may mean what it seems to mean. Language obscures the real situation. Phrases like ‘looks blue’ reveal that if we stop and think we can recognise that things are a bit more complicated. Ideas of objects may allow our collating systems to learn to distinguish, and generate signs for, the distinction between the blueness disposition of cookers and the blueness disposition of light but this may come at the price.

Key point: our sensory systems often use the same manifest signs, or appearances, for unrelated dispositions that are commonly associated in causal chains, again indicating that the manifestations cannot be ‘like’ the dispositions they are signs for.

2.7 Difficulties with dispositions

The discussion of the blue cooker raises two important questions for our understanding of physics. Are all aspects of the underlying ‘out there’ reality of the world in fact dispositions (i.e. tendencies or causal powers)? Do these dispositions belong to objects or are they, for instance, just local features of a universe? These may seem to be issues for armchair philosophical debate. However, I would suggest that they are not, for two reasons. Firstly, sorting these questions out may show us why modern physics is in fact no stranger than the old physics. The mysteries people often talk of simply disappear. Secondly, sorting them out may provide a way of working out the nature of the human subject to which the world seems to be manifest in experience.

Before addressing these major questions I need to consider two difficulties with the idea of dispositions, discussed in the philosophical literature. I might have avoided these problems, at least in part, if I chose other terms, like causal power, but my impression is that all the terms in this area suffer from conflation of meanings to much the same extent, if in slightly different ways.

The first difficulty is with the idea of dispositions existing in the absence of any other underpinning or ‘qualitative’ property that is the way that some entity with the disposition ‘really is’. For instance, the enamel’s disposition to reflect blue light, is underpinned by the enamel being of a quality that makes it tend to do that.

The second difficulty is with the idea of dispositions existing, or being instantiated, without that requiring that they are ever manifest. In what sense would the disposition to reflect blue light be actually present for a cooker enameled in the dark and assembled in the dark and smashed to powder before ever exposed to light? The need to take this issue seriously was emphasized by CB Martin. Although Martin makes some important points in favour of thinking in terms of dispositions I shall argue that they always exist as instances of operation, and in that sense are always manifest. Moreover, we have no reason to ask for anything underlying them. The cooker in the dark may seem to contradict this but I will argue not. The real difficulty is in defining dispositions in the right terms.

In philosophy a popular example of a troublesome dispositional property is fragility. A fragile vase can be said to be disposed to break if hit by a hammer. We tend to think of this disposition as being grounded in some intrinsic aspect of the vase. It is also easy to feel uncomfortable about the idea that vases that can never be hit by hammers are still endowed with this disposition. (Think of the disposition of Spode vases to break when trodden on by dinosaurs.)

These two problems are interrelated. To avoid having genuinely present dispositions that are defined by realizations in situations that can never occur we appeal to qualities. Perhaps the fragility disposition does not exist all the time, only some underlying features of the china relating to chemical bonds between atoms.

There are two ways one can think of this. A traditional view would be to say that the vase is in fact an aggregate structure, consisting of a collection of molecules that are bound to each other in the way typical of china. The idea that dispositional properties are based on underlying’ properties probably derives from the general view, which must have preceded any formal science, that the properties of big things tend to be ‘explicable’ by the combined properties of the parts. Such an analysis seems to work in a wide range of situations.

However, if we then look at the binding properties of the molecules these turn out to be a tendency to attract other molecules and maintain them in proximity unless overcome by high kinetic/ thermal energy. They are dispositions or powers to retain proximity under certain circumstances and not otherwise. So there is no real difference between the type of property at the two levels. We seem simply to have shifted disposition to a smaller level.

Note, however, that at small scale the sort of definition we need for dispositions tends to become evident. We tend to think of molecular dispositions as manifest, or at least ‘cashed out’ in terms of the vase sticking together. However, note that the ‘not otherwise’ component is not cashed out. Similarly, fragility is equally a tendency to hold together if not hit (or dropped) and break if hit or dropped. Something that disintegrates without impact is not fragile. Remaining intact is just as much a manifestation of fragility as coming apart, if it occurs under the appropriate circumstances.

A possible weakness of the aggregate view of the jar is that it is tantamount to suggesting that there is no real jar entity to be fragile, just an aggregate of molecules with microproperties. This may be a salutory reminder that in a whole range of contexts, like AGA cookers built in to a kitchen, a lot of properties we ascribe to large functional complexes are really better seen as properties of components. Yet we might still want to consider a vase a legitimate entity with regard to fragility because the fragility may not just be a function of molecular binding. It may reflect an elongated narrow neck, or protruding handles. Little chips of china are not fragile in the same way; they do not crack when dropped.

I only want to mention this briefly at this point, but there is another way of looking at the jar, informed by a combination of classical thermodynamics and modern field theory and maybe also some ‘common sense’. The jar may be seen as a true entity because its existence as a single coherent structure brings with it certain energy bearing modes, such as modes of rotation and the elastic or acoustic modes that allow it to ‘ring’ if tapped. These may seem rather trivial aspects but there are reasons for thinking that they are crucially relevant to any concept of ‘vaseness’ or being a vase. A vase is something that you can pick up and turn over (rotate) in one piece, unlike a pile of books. It is firm or ‘sound’, in a way directly related to its ring. Even the reflection of transmission of light by a vase is partly dependent on its macroscopic modes.

I will argue later that these modes are as real entities as others. The fragility of the vase can be rephrased in terms of the stability of elastic (alias acoustic) modes. A glass that shatters when someone sings a high note is showing that it has an acoustic mode that, when inhabited by sufficient energy of vibration, self-destructs. Fragility then becomes a dispositional property of a single entity. That entity is not the traditional concept of jar-as-whole-made-of-parts since these modes have no parts, but they do occupy the entire ‘jar domain’.

Either way, it looks, so far, as if whichever level we look at we have dispositional properties that are always associated with some sort of manifestation.

The next step in the argument is to ask whether we think that either the molecular binding in the first case or the mode in the second are in themselves underlain by some non-dispositional qualities. Perhaps we could say that there are some ‘chemical bonds’ there. However, ever since quantum theory provided us with an explanatory, rather than just an empirical, theory of chemistry, the chemical bond has turned out to be something that we can only understand as some form of dynamic disposition of the patterns of electron orbitals around nuclei – a disposition to maintain proximity. If we then ask about the electrons we find descriptions in terms of mass, charge and spin. These might seem to be ‘qualitative’ at last, but in fact they are all defined in terms of dispositions. Negative charge is a disposition to repel negative charge and attractive positive charge. We know nothing more about it than that. Mass is a disposition to resist acceleration and also to attract through ‘gravity’ (or maybe warp spacetime). It is dispositions all the way down.

The philosophers may still object that although we can only know these properties as dispositions they must be underpinned in some way by qualities of things. The argument is that you cannot just have little patches of powers, powers must reflect qualities of what has them. I want to come back to this later because the arguments tend to get rarified. However, at this point I would simply ask what could possibly be meant by such underpinning qualities if they are not in themselves dispositional. Would not underpinning imply a disposition to bring the disposition along too?

The objection to the idea of dispositions all the way down, without underpinning by ‘qualities’ is closely related to a concern that dispositions must be the dispositions of some ‘physical’ thing, rather than just floating free in the universe. If there is no underpinning quality belonging to some thing then how can the disposition belong to any particular thing? This concern fits with our instinctive childhood approach but has never been much to do with physics. Whatever these underpinning ‘physical’ qualities might be they are not something that physics says anything about, which makes the term ‘physical’ in its lay sense rather ironic. Dispositions do seem to need to be tied into ‘bundles’ or patterns in some way but that is not helped by being tagged to some ill-defined sense of what the bundle might really be like if we were an all seeing God. I shall return to this issue shortly.

What does seem reasonable is the idea that dispositions cannot just exist as abstract tendencies (an example of the more general problem of the existence of abstract universals). For this reason I think it makes sense to suggest that a disposition only ever exists as an instance of operation – as a pattern of goings on. That is what makes it real rather than just a mathematical structure. If the disposition is fragility and in a given case breakage does not occur, the disposition, as fully formulated, has still operated, because it is a disposition to break under some conditions and not under others. If it is being blue in the sense of reflecting twice as much of the incident blue light as of the red, this disposition can still be said to have operated when the amount of light for both colours is zero. (Whether we say the disposition is manifest is a more complicated question that I will address later.) Things staying the same is as much a part of what is going on as not staying the same. It is in a sense still a pattern of dynamics but one with some zero values. It is still quite a different matter from what is manifest as a result.

I have so far not distinguished between the need for all real dispositions to be instances of operation and for them to lead to manifestations. To equate these functions will do for the purposes of this section but will need unpicking later, as I think there are at least three conceptions of manifestation, two legitimate and one illegitimate, which need to be explored. Moreover, manifestations are themselves dispositional in the sense that they contribute to causation. My aim at this stage is simply to indicate that we need to replace our concept of stuff with concepts of disposition or causal power.

So there are reasons for thinking that dispositions may do much of the work we intuitively think as done by ‘qualities’. At this point I need to deal with various issues about the context of dispositions, before returning to the question of whether all features of underlying reality are dispositional.

Key point: although philosophers have raised various problems about dispositions there are reasons to think that all we can know of the world are the instances of operation of patterns of dynamic disposition.

2.8 What are ‘objects’ and are there any?

Common language usage and also most day to day thinking takes it as read that the world contains ‘objects’. Objects are entities that are a certain way, in a certain position, of a certain colour, weight or size. They thus have qualities and are ‘in a certain state’. As indicated above, these definitive features of objects may not be quite what they seem. Inasmuch as we know them we know instances of operation of dispositions. This calls into question whether or not we really have a clear understanding of what we mean by objects. That we may not has been explored at length by James Ladyman in his book Every Thing Must Go.

The word object seems to imply a dynamic role; -ject comes from to throw. However, the dictionary suggests that this role can be as varied as something presented to the mind or something to which something is done. I suspect that any specific dynamic role has been lost from most people’s understanding of the word. In fact, most of us probably think of objects as non-dynamic: as things that just are a certain way, and only as a byproduct get involved in relations to other things.

If we insist, as some philosophers would, that we see or perceive objects rather than signs of objects (or goings on) because that is what seeing and perceiving mean then we need to be careful we do not have a circular definition of meanings. If objects are defined as things we are related to by seeing and seeing is defined as our relation to objects we have no indication what either ‘seeing’ or ‘object’ means. This should make us go back and check whether the word ‘see’ can have the sort of meaning that the sentences you find it in would suggest.

In similar verbal constructions, like ‘I hit the ball’ the object is a ball and the verb tells us something about my influence on it. However, for ‘seeing a table’, we are not considering our influence on the object. That does not mean that seeing is just passively receiving input. It requires the brain to actively compare different inputs before recognition of an object is achieved. However, these ‘activities’ do not influence the table (except in an irrelevant sense that applies equally if the person is asleep).

The problem that we have with language is that we assume that its constructions relate unproblematically to what is really going on in the world. A construction of subject, verb and object is seen as a straightforward indication of some sort of dynamic relationship in the world. This seems to work for ‘I hit the ball’. However, linguists point out that it gets murkier when we compare ‘I persuaded John to buy a ticket.’ And ‘I expected John to buy a ticket.’. When it comes to ‘I see a shiny cooker.’ It seems likely that there is no dynamic relation that logically fits into this verbal structure. Seeing is a pseudodynamic idea that seems to refer to some dynamic relation but in fact does not. If we want to give a truly usable dynamic description we have to recruit different sorts of verbal construction.

Wittgenstein claimed that ‘what can be said at all can be said clearly, and what we cannot talk about we must pass over in silence’. This seems to imply that the language we have at present, and the way we use it, are all that are possible. It seems to imply that we cannot have useful and meaningful thoughts for which we have no words readily available. As a practicing scientist this strikes me as patently untrue. Science is all about having new ideas about the world that we then have to create new language to describe. Ordinary life would seem to work the same way. All our words have at some time been coined in order to label some idea already in somebody’s mind. Moreover, in mathematics new languages are constantly being devised to handle new thoughts. When a child first sees an algebraic equation x = 4y he or she has no idea what it could possibly mean until it is explained what sorts of thoughts can be handled more easily by this sort of language than by ordinary English. If we want to clarify our thoughts about the world we should be looking for more subtle language forms to suit new ideas, not to find ideas to suit the language forms we have.

So it is not going to be any good defining objects as things that we can see or perceive.

What might be more helpful is the historical concept of objects as ‘things presented to the mind’, which might also be construed as things from which we receive signals that indicate their nature. That seems a reasonable start. However, there are two caveats relating to the two questions raised previously. Firstly, we have to consider what aspects of ‘nature’ signals might indicate. Secondly, we have to consider how the mind can know which signals relate to which objects. What are presented to the mind are signals; objects have to be inferred from the way these signals behave under various different conditions. It is not clear that our ways of dividing up what is presented to us in experience are good ways, or even that dividing up is legitimate.

Nevertheless, following the approach suggested, perceiving an object will be receiving such signals in a way that allows us to infer something about the signals’ indirect causal origin. It is hard to see how we could infer anything about any non-causal properties of objects since they could not be responsible for us receiving any signals. We come back to the idea that whatever entities there may be in the world we can only know of them in terms of their dispositional causal powers, not some non-causal features that might be described as ‘what they are made of’. Talk of what something is made of is really just talk of how complex things may be broken down into components, it never gets us nearer to ‘the way things are’ other than in terms of their causal dispositions. This, I believe, relates to a common misconception about science: that its job is to ‘reduce’ the complex to the simple. It certainly does a lot of that, but that is not ultimately what science is for, its real job is to relate our experiences to what is really going on – to relate cause and effect.

The signals we receive from the world may be said to represent the world in a loose sense, as loose as the sense in which members of parliament represent people or a flag represents a nation. What we would not expect is for these signals to resemble the causes that determine their patterns because they are not themselves causal dispositions, but rather some of the things that the dispositions are to be cashed out as, and very indirectly, at that. We have images that signify certain patterns of operation of disposition but are not images of those operating dispositions.

This point goes beyond the point that the neural processes that make someone sense red will not look red to a neuroscientist peering at the persons brain with a microscope. Neither are the processes red in that sense, nor are they even presenting to the owner a resemblance of some property such as redness or shape that an external object has because the dynamic properties that are being represented simply do not have aspects that can be ‘resembled’ in this way.

This is not to deny that dynamic processes may not be ‘modeled’ by other dynamic processes with the same or similar dynamic parameters. But the dynamics inside our brains that generate the manifestations we behold are totally invisible to us. Moreover, they would not be expected to share parameters with what they signify. If we see three oranges there is no reason to think that what is manifest to the internal subject includes something in triplicate, any more than any of the words in this sentence come in triplicate. To think that representations must resemble their referents in this way is to fall into the true homunculus fallacy – the idea that there has to be some sort of internal ‘seeing’ that resembles the optical apparatus that first gathers data from the world. It is easy to assume they should, and there is no doubt that there will be triplication of certain events at early stages of the sensory process, as in the retina, but we would expect the final manifestation to the internal subject to be based on a sign of triplication, not three similar signs. If not, it is hard to see what all the pre-collation mechanisms are for – why not plug the retina straight into the soul! The signs manifest to the subject will be organised in a way suited to decision making about the world, not to modeling the external objects to be decided on.

Objects, as separate parts of the world, that are a certain way, and are presented to us as separate things that are that way, are fictions of our childhood view of the world. Like Ladyman, I think they should go. What we have reason to think exist are patterns of operating disposition, or goings on. Unlike Ladyman, however, I do think that some of these patterns we can reasonably consider as genuine ‘packets’ or ‘things’ in the sense of individual entities. My view may be close to that of Shoemaker’s ‘bundle theory’ here, but I agree with Heil that ‘bundle’ has an unfortunate implication of parts that I would want to avoid. The neutral term ‘packet’ borrowed from quantum physics, seems less troublesome. Some of these pattern packets have much the same domains as childhood objects, but the two categories should not be confused. Moreover, my motivation for considering there to be genuine individual entities is not to provide objects. Ladyman is right in this respect – we do not need them. What we do need are subjects, of which more in due course.

Although I may continue to use the term property for simplicity I think it is probably true to say that I am sceptical of the philosophers’ concept of a property of something. I think it may be more useful to think of packets of disposition as aspects, or perhaps properties, of the world as a whole. (Modern physics strongly suggests that we should think of them simply as asymmetries of the universe.) If individual entities can only be conceived in terms of dispositions it seems redundant to regard these dispositions as properties of the entity rather than just the entity itself.

Key point: the everyday concept of an object is not a good basis for building a view of the world. Instances of operation of dispositions are best thought of as patterns in the world rather than properties of objects. Sometimes these patterns fit quite well with the domains of everyday objects, but that does not mean that they are equivalent to the everyday concept of the object.

2.9 Signs will depend on the way of receiving

Another aspect of the idea that our perception of the world is via an internal system of signs is that the sensations human subjects have when we see, hear, touch or smell things must be entirely dependent on the sort of collating apparatus we have. Thus my blue cooker can only be blue, in the sense we think we share, for observers that have not only the same set of retinal receptors, but also the same general methods for sorting and collating initial inputs.

We are taught at school in art lessons that there are three primary colours – red, blue and yellow. We are then taught in science that the there are three primary colours – red, blue and green. The second set may be considered more accurate, since it makes sense that the existence of ‘primary’ colours is an artifact of our having these three sorts of colour receptor in the retina.

Yet there is this odd fact that we think yellow is at least as much a ‘pure’ colour as green and analysis of use of words for colours throughout different communities suggests that we have a universal sense of four ‘pure’ colours. The precise reasons for this may not be clear but there is at least one plausible explanation.

We also have colour-indifferent photoreceptors (rods). This means that the simplest distinction the visual sensory system has to make is a light-dark one. If we then add in whether or not blue-sensitive receptors are being stimulated we have two possibilities for any perceived brightness – brightness accounted for by blue light or brightness not accounted for by blue light. If the brain receives a strong brightness signal but no blue signal it should be expected to allocate a colour to that brightness that we might call yellow.

If we then add in a red receptor we can split the yellow part of the spectrum into red and green, according to whether the red receptor gets a high signal (signifying red) or a low signal (signifying, by default, green) relative to the level of ‘non-blue brightness’ inferred from collating the first two receptor inputs. This raises the puzzling question as to why we also have a green receptor.

There are at least two factors that might explain the apparent redundancy of colour receptors. One is that the central field of vision has few rods so has to work largely on the cones alone. The other is that we also need to make use of the total brightness signal as a way of detecting form through shadow. However, the sophistication of the collating system is illustrated by other oddities, like the fact that despite there being many fewer cones in the peripheral visual field, we do not see peripheral objects as ‘less colourful’.

It seems that the existence of ‘four pure colours’ arises through a complex collating programme in our visual systems that may reflect both the selective pressures of what is useful for our survival and random quirks of structure due to the incremental nature of the evolutionary process. Maybe if red receptors happened to evolve before blue we would have a different sense of the pure colours. There is nothing ‘transparent’ in the relation of mind to object. It is full of murky magical tricks, even if hidden from view.

There is also the more general argument, disregarding any suggestion of phenomenal experience, that any sign received from some part of the world will be dependent not just on the causal dispositions of that bit of the world but equally on the disposition of the bit receiving the influence. Thus an electron is disposed to receive a repulsion signal from another electron but a proton will receive an attractive signal from the same electron. Colour perception is entirely dependent on the dispositions of the receiving apparatus to be influenced by light of different wavelengths. Smell is dependent on the disposition of receptors to be occupied by molecules sniffed in from the world. Two molecules quite different in many respects may smell identical to one animal but quite distinct to another. If the same receptors were expressed on tongue rather than nose cells then the dispositions of molecules that determine ‘smell’ to us, would give ‘tastes’, which the brain collates quite differently for evolutionary reasons. Thus it is not possible for us to think of a smell as an inherent property of a substance, independent of who is smelling it.

All properties that we sense are properties determined by the ‘fit’ of incoming influence and what is being influenced. One of the jobs of science is to find a language that can separate these. For simple things like electrons, dispositions can be described simply, in terms of other negatively or positively charged entities. The disposition takes two forms, repulsion and attraction, manifest as push and pull. However, for a perfume molecule fitting into an olfactory receptor there are millions of possible receptors to interact with and countless possible collating systems that link these receptors to a site where a smell is manifest. Smell is not a concept like negative charge, if anything it is more like push.

I have said very little of the historic literature on this topic, which is extensive. This is largely because these are insights that can and have been reached at many points in history and the historic accounts tend to overburden the topic with misleading jargon. Nevertheless, it may be fair to say that much of what has been said so far follows the account John Locke gives for what he calls ‘secondary properties’. These are properties that objects appear to have in themselves but which Locke realised are more properly seen as ideas raised in the mind by signals coming from the objects.

Locke was of the view that certain other properties, such as shape and size, which I will come to next, could be considered ‘primary’ because they really are intrinsic features of the objects. Although there is a sense in which Locke makes a valid distinction between two categories, I think it is important to note that in a crucial sense his view is misleading because all properties relate to dynamic dispositions to generate manifest ideas. Locke’s distinction turns out to relate to a subtle difference between our ways of knowing relatively simple interactions in the outside world and our ways of knowing the highly complex interactions of our own internal information gathering apparatus. This will be very relevant later but for the time being the key point is that even our sense of space is incommensurable with the space of dynamics.

Key point: the way the world appears to us is as much to do with us as the world.

2.10 Space, time and disposition

On an occasion when I remarked to some fellow philosophy students that I thought all features of the outside world are dispositional, I was asked ‘What about the property of being at time t.’ One could also ask about the property of being two metres long. Surely, if we can see at time t by the clock that X matches a two metre ruler, these are established qualities of X, rather than dispositions?

However, a careful consideration of the properties that do seem to be dispositional casts doubt on this. In detail, these dispositional properties might be ‘a preferential tendency to reflect blue light over a length of two metres and the period t-5 years to t+5 years; a tendency to prevent the transit of other solid objects over a length of two metres and the period t-5 years to t+5 years, etc. etc.. Thus, the properties of being at time t or two metres long can be seen not as additional properties but merely as the spacetime domain within which the dispositional properties we have signs of are operating. Time and space (in this sense) are the metric of operation of dispositions. (CBM MIN p1.) To say something is two metres long is to say that a full description of any of the dispositions involved will include that the domain of operation is of two metres length. If we take dispositions always to be token instances of operation then they do not even exist in the absence of their spacetime domain.

An important implication of this is that neither time nor space, as here defined as the metric of causal dispositions, have any appearance. If dispositions have no appearances then their extent cannot have an appearance. (cf Berkeley, 1710) Locke was wrong to suggest that our ideas of shape are more veridical or more like the shapes in the world they are about than, for instances, our sense of colour. This at first seems a truly weird conclusion, but it has to be drawn. And, after all, we could all probably agree that time and space alone do not have appearances.

We need to accept that the space of the outside world is not ‘really like anything’, nor is the time. The felt ‘spatial’ and ‘temporal’ aspects of the manifestations that elements of the outside world determine in our brains cannot be what the spatiotemporal domains of those dispositional elements are ‘in themselves like’ for the same reasons that the mental sign we call green is not ‘like’ the disposition of a surface to reflect mid-wavelength light. Not only are the signs of space and time and of colour not like the dispositional events that lead to them, they do not even fall into the same type of general category. Changes and what the changes result in, in the sense of dynamics and manifestations, cannot stand in any sort of analogy to each other. (However odd it may seem, this approach does have the great advantage that it relieves us of the burden of trying to work out how the sensed space of the Grand Canyon can exist inside a skull.)

It may be worth commenting at this point that the argument just given seems to be extremely difficult for many people to accept and it is uncertain that even some of the greatest minds in the field, like Newton and Russell, really appreciated its full implications. Newton does seem not just to produce the argument, but to stress it, in his first ‘Scholium’, or explanatory aside, in the Principia. He tells us not to confuse the absolute space and time of physics with the space and time of everyday life. He mentions a number of differences. He contrasts them as absolute and relative, true and apparent, mathematical and common. He seems to be emphasising that they really are different in kind. Nevertheless, when it comes to his detailed comments it does not seem that he took this difference to be quite as stark as I think one must. He appears to have regarded apparent space and time as merely imprecise, relative versions of the metric of his dynamics. Pace Newton, I think we need to accept that the two senses of space and time are as different as the pattern of moves the pieces in a game of chess make are different from the resulting pattern of pieces on the board - the difference between change and what that change determines.

Russell mentions the difference between the spacetime of physics and the space and time of experience but as for Newton, hints that he sees the difference as one of partial resemblance rather than difference in kind. On the other hand, going back two thousand years, Parmenides and Heraclitus both do seem to have seen just how different change is from manifestation. My impression is that they may have seen the problem more clearly than Newton.

Why is it so difficult for most people to see that these two types of concept are so unlike and related by indirect causation rather than resemblance? The first likely reason is the fact that the dichotomy is usually made invisible by our language. The second likely reason, I believe, has to do with the fact that most of our thinking is designed to ‘cancel out’ the nature of the internal manifestation so that we can concentrate on causal chains (not just cause-effect but cause-cause-cause-effect) in the outside world, similarly tending to make the cause/manifestation distinction invisible.

I have mentioned the language problem in relation to ‘The tomatoes are red’. Even there, where we appreciate two meanings, we tend to think they are not so different: as in ‘The tomatoes are actually red.’ and ‘The tomatoes look red but they are yellow because they are under a beam of red light. The implication is that being red and looking red relate to our mistaking of ‘one colour for another’. I hope I have convinced the reader that nothing could be further from the truth. The (first) word red relates to a completely different kind of concept in these two sentences.

Exactly the same applies to space, as in ‘My great-aunt’s house really is small.’ And ‘My great aunt’s house looks so small (60 years on).’ It is not that the way it looks is ‘unreliable’ or unveridical’. It is not that the space of experience does not quite match the space of physics. (In the example one sense of space does not match a remembered sense of space, that is all.) It does not ever ‘match’ because it is a different kind of concept. The space of physics is part of what is now thought of as a single metric or framework for dynamic causal processes. The dispositional properties of entities in the world, the unfolding of which constitute ‘what is really going on’, operate in a metric of spacetime. It is unclear what spacetime could mean other than in the context of something going on. Change is measured in spacetime. In contrast, no dispositions are operating within an experience. No causal relation is going from right to left in my experience of the world. It would not take 0.5 seconds for my view of the table to get to my view of the AGA cooker. Even in spatial terms there is no reason to think that a line in my experience is composed of dots. The sign from the sensory collating apparatus will mean ‘line’ not ‘dot+dot+dot’. The space of experience is not a metric of anything. It is not something by which experience is measured because you cannot measure in experiences. Moreover, we experience space and time differently, not as a unified spacetime.

There are all sorts of ways of trying to describe the argument in the above paragraph and I know from experience that many people do not follow the argument no matter how one puts it. I am also aware of the extreme fragility of language here. A subtle shift in the interpretation of any of the words used may indeed render the arguments incoherent and I am very aware that I have taken some liberties so far that will need to be corrected later. Nevertheless, I think that common sense tells us that we have to accept the dynamic spacetime and experienced space and time are completely different kinds of concept for the same reasons as for red and green. Locke’s primary properties follow the same rules as the secondary ones.

Key point: Locke’s primary properties of shape and duration are the domains of operation of dispositional patterns and the way they appear to us is no more independent of us than the secondary properties such as colour.

2.11 Must all aspects of the outside world be dispositional?

Having argued that many properties we take to be ‘qualitative’ in the sense of ‘what things are like’ are in fact operations of dispositions, or unfolding tendencies, the question arises as to whether absolutely all features of underlying reality must be dispositional. Colours and tastes look to be dispositional. Size, shape and duration might be thought of as non-dispositional in themselves, but, as argued above, it looks as if they can be seen as the extent of dispositions and as such can no more constitute ‘what things are really like’ than the dispositions they delineate.

The fundamental properties of physics are also only known to us as dispositions, and are defined as such. Mass is both a tendency to resist acceleration in the face of force and a tendency to attract other masses, (or warp spacetime). More esoteric properties of quarks like charm and colour follow the same principle.

There are some aspects of the world that I want to save for discussion in later sections because they relate to language and knowledge: like meaning, aboutness (as a dream is about a train journey), truth, veridicality and certainty. These may or may not be dispositional, but they are not the sort of ‘qualitative’ properties that some would suggest underlie the dispositions I am concerned about at this point.

Running through all the properties of concrete (in the sense of token or instantiated) entities that I can think of I cannot find an instance that does not follow the general rule. However, this might of course be because I am missing important exceptions. The more convincing strategy is to see if it even makes sense for features of the world to be non-dispositional.

People want some non-dispositional properties presumably because they feel there should be ‘ways things are’. Ways things are in the naïve sense turn out to be dispositional properties, signs of which are manifest to us in experience, as I have indicated. We need dispositions and we need the experiences they are disposed to determine. If that is not enough then quite what more is wanted? Do we really want a ‘triple aspect’ theory? What justification is there for proposing a third category, which so far we have never come across? For this intuition, of there being an additional category of how things are, to be justified, we would have to assume that we are somehow genetically programmed to know that there is a certain way the world is, without ever having had manifest evidence. Such knowledge would have to be gained without any causal mechanism. The proposal of such a category seems not only unparsimonious but devoid of any motivation.

As I shall return to later under a discussion of the logical positivist approach, there is nothing wrong with positing aspects of the universe that at present we have no empirical evidence for. It would be perfectly reasonable to suggest that there are dispositional aspects to elements of the universe like charge, spin, and strangeness but which so far we have no inkling of. Most physicists believe that indescribable dark matter makes up most of the mass of the universe. It is quite a different thing to propose an aspect of the world quite different in category from everything we know: neither a disposition nor a manifestation in experience but a ‘quality’ for which we are unable to give any cogent account at all. It is for this reason that I think we can reasonably suggest that not only all the features of the world that we know, but also all the ones that we might reasonably surmise might exist, are either instances of operation of dispositions or the manifestations those dispositions help to determine.

My viewpoint is close to that argued by Shoemaker and is widespread amongst physicists. Resistance from philosophers seems to be partly a matter of immovable intuition, but some have raised specific objections. CB Martin argues that the suggestion that the only properties things can have are dispositional ones is absurd, based on the argument that you cannot just have a world made up of an endless chain of tendencies or dispositions operating. At some point these must be tendencies or dispositions to something – some non-dispositional event – a real happening or occasion perhaps. I very much agree. Note that this is not a problem of circularity or infinite regress, as some have suggested, but simply that whether you have a single disposition or a chain of dispositions, at some point some disposition has to be a disposition to something.

Martin is keen on the idea of casting the causal world in terms of dispositions and manifestations and my choice of the word manifestation may well have been influenced by him. However, Martin’s concept of manifestation is rather different from mine and the distinction raises an issue that will be important throughout this text.

For Martin, a manifestation is what we might ordinarily call an event. If an egg is a packet of disposition to hatch, then hatching is the manifestation of that disposition. However, all that we can know about hatching is that a certain instance of operation of a disposition has run its course and that subsequently other packets of disposition come into play, like a chick. (Unlike Leibniz, as usually construed, I think we have to say that the universe is not just the continuing harmonious progression of a fixed set of immortal monadic dispositions but that packets can cease to exist and others come into being.) We have evidence of some sort of watershed in the history of dispositions we can infer but it is still a history of dispositions leading to dispositions. This history can be considered real in that the dispositions are concrete instances rather than just abstract mathematical structures but it is not clear to me that Martin has got what he wanted which was some sort of end result other than a fabric of dispositions. In a sense he needs his manifestation to be to God, so that He can say ‘Aha! It hatched’.

To my way of thinking we have not cashed out our dispositions for sure until we have an experience. At least then the buck stops somewhere. Dispositions are not only real in being instantiated but also in that they determine something, and ultimately the only thing we have to be sure about is a manifest experience. Thus for me the one indisputably legitimate meaning of ‘manifestation’ is a human experience. What is not legitimate is the idea that we can say there is a manifestation just on the basis that ‘it happened and God saw it’. It is not legitimate if only because modern physics has shown that deep down the universe does not work like that.

However, I do think that there is another potentially legitimate meaning for manifestation: one well worth pursuing. This is the influence of one or more packets of disposition A,B,C on another packet of disposition D as ‘experienced’, defined in purely operational terms, by D. This makes a manifestation mind-independent but not interaction-independent. It would be Leibniz’s perception by D of A,B,C except that he might call influence the implications for harmonious progression. It is the manifestation of a hurricane that is reported as ‘Miami experienced 100mph winds tonight’. Put another way, a manifestation cannot be epiphenomenal, it must have causal power. That seems to make it a disposition, and it does, but it may not make it an instance of change (dynamic): more of which later. And I agree with McKitrick that the burden is on those who want the world to be more than a chain of dynamic and manifest dispositions to say what more they want. (McKitrick has argued that if manifestations are dispositions they would be unobservable but I think this can be sorted out when we come to the mechanics of knowing.)

Although physics says nothing much about manifestation in this broad sense I will argue later that it cannot really do without it. It is particularly relevant to quantum theory in that without the idea that inanimate disposition packets do actually ‘manifest’ some determinate signal to each other, we have Schrödinger’s deliberate absurdity of a cat that is both dead and alive until someone looks in its box. We want a mind independent reality. However, we cannot ask for an interaction–independent manifest reality, as I shall come to next.

Key point: we have no reason to think there are non-dispositional aspects of the outside world. If there were, they would be aspects the like of which nobody has ever encountered.

2.12 Qualities as purely proximal and asymmetrically relational

Martin has a second criticism of Shoemaker, linked to the first. He claims that we know that the world is not just made of dispositions because we are aware of qualitative aspects. But he also goes on to suggest what he himself calls the ‘surprising identity’ of dispositions and qualities. This identity thesis has been extended by Heil.

There might seem to be a sleight of hand here, and I think there is, but I also think there is mileage in the approach. I see no objection to either of Martin’s claims: that we are aware of qualities and that qualities are dispositions. The problem, I think, lies in ensuring that the right disposition is identified with the right quality. Words like disposition, manifestation and quality cut the world at certain joints. We need to be sure we are matching the joints up correctly.

The problem I see here is to avoid confusion of two quite different concepts of ‘qualities’. Martin, as perhaps Russell, equates these with the term ‘qualia’ used to describe elements of phenomenal experience. The suggestion seems to be that qualia are just a restricted ‘mental’ version of qualities that all things have.

This seems very odd because, as I understand it, the term ‘qualia’ is a deliberate neologism designed to distinguish qualities-as-sensed in experience from qualities as mind-independent properties of objects, which I agree with Martin are dispositional properties of the objects (if we allow these). Yes, we need a qualitative side to our story, but we cannot run together the two meanings. Locke’s qualities were properties of objects that gave rise to ideas in the mind that signify these properties. Qualia are the elements of those manifest ideas, not the qualities that give rise to them. The unification of the two concepts may be desirable, but we need to be careful how we go about it.

Martin himself seems to give a clear description of what qualia must be. They are aspects of manifestations, which are always determined by the interaction of at least two ‘dispositional partners’. A smashing of a vase is the interaction of the fragile vase with a rigid hammer. A tasting of pepperiness is the interaction of the pepper with a taste receptor. A playdough hammer and a sugar taste receptor will not give the same manifestations. But there is a conflation here. Manifestation is being used in two different senses. In both senses there is an ‘occasion’ or happening but in the first the quale of a crashing sound, belongs to a manifestation in the head of someone watching the destructive act. Out there in the world there are either no qualia or, if one takes a panexperientialist view, which Martin, (and certainly Heil) seems to resist, there must be two sets, one for the vase experiencing the impact of the hammer and the other for the hammer experiencing the vase in its path. If we take the alternative manifestation of fragility of remaining intact in the absence of impact we seem to have a vase accused of having qualities but not in relation to any occasion out there in the world – we are back to the occasion of experience within the observer and the qualia pattern signifying colours of the ceramic glaze. Or worse, we have a ‘public’ manifestation that is no more than the illusory manifestation to God.

One of the things that muddies the water here is, I think, Martin’s claim that dispositions and qualities can exist without being manifest. Clearly, qualia are manifest. Surely qualities must be the same, unless we are thinking in terms of abstract universals. The wider underlying problem is that if we are to do rigorous metaphysics we cannot expect to make use of everyday objects as examples because these are almost certainly not metaphysical primitives. Martin himself acknowledges this when he points out that everyday properties of an object are not properties of the object ‘holus bolus’. The implication is that some reallocation of entity status may be needed (not necessarily ‘reductive’).

In fact, as will be analysed further in section II, we have good reasons for thinking that in the second case the manifestation we recognise as pepperiness does not occur at the taste receptor. The taste receptor receives some stimulus and this may be in a way that we can call a manifestation in the sense of having some ‘feel’ to it. However, that will not be the feel we are familiar with, which can only arise after a further chain of interactions in the brain have collated the input from the receptor with many other signals. What the interaction with the receptor does highlight is that at every step in a causal chain what is passed on is dependent on the fit of the interacting elements, or what one could call the way one is manifest to another.

Martin implies that the qualia of experience must be in some way similar to qualities of objects, but there is a crucial difference. Qualia are those aspects, to try and find an appropriate word, of relations between causal dynamic elements in a brain that are experienced by an element S. They are not properties of S but of ‘everything else for S’. Contemporary philosophical jargon makes no reference to such terminology but to my mind that is just too bad for contemporary philosophy. The ideas are in no way new to philosophy because they are entirely consistent with Leibniz’s conceptions of the apperceptions of entities that accompany the progression of those entities in harmony with others. The qualia belonging to S are its autobiography in terms of how it apperceives the world. They are intrinsic to its history but only by dint of relation to other things. They are what the world is like to S, not what S is like.

Another way to unpick the problem of Martin’s model might be to say that a quality is the outcome of a dynamic disposition as manifest in the next event or occasion (to the one with this outcome). Qualities seem to be manifest and we think of manifestation as following disposition. However, I think this would be to misread Martin’s message. The blue quality of the cooker is the blue disposition of the cooker in his terms. It is remarkably hard to sort out the linguistic tangle here, which, again, I think is in large part due to trying to do metaphysics on everyday objects that are not suited to rigorous metaphysical talk.

To my mind, the use of the new term ‘qualia’ should not be seen as introducing some new concept to work alongside the concept of qualitative property of an object but rather as a new concept that replaces the old concept, which has become redundant. However, we might also want an equivalent for qualia for mind-independent causal relations.

Given that we are talking about confusions of terms I admit that I may have misunderstood the intended meanings in both Martin’s and Russell’s use of ‘qualitative’. Yet my best attempt to divine what they intend suggests to me an inconsistency in usage of the term. It is for this reason that I see Russell’s attempt at neutral monism as failing. It is quite difficult for me to understand how Russell could have made the mistake I am accusing him of. He argues clearly on so many other issues. Nevertheless, I end up concluding that this is one of those cases where words carry an argument along without actually carrying any valid concepts with them. This situation is undoubtedly widespread in all our thinking. The reason why I am writing this text at the age of 62 is that it has taken me all that time to disentangle my own confusions over such issues. Russell seemed so well placed to avoid the mistake, but he was as human as any of us and moreover he had immersed himself in a programme of logic that may have emphasised the logical rather than dynamic import of words.

I see this issue as potentially the most crucial tipping point in the development of a comprehensive and coherent complementarity-based world view. The idea that ‘qualities’, if we want to keep this word as well as ‘qualia’, are entirely proximal and asymmetrically relational (between X and everything else for that X) seems very hard for many people to absorb. Yet there is no coherent causal dynamic account of how it could be otherwise. If we want mind-independent qualities these still have to be interaction-dependent: qualities for partners, just like dispositions for partners. This might seem to require an unacceptable panexperientialism, but in due course I hope to show that this may not be quite such a bogey as it seems.

Key point: everything we know about causal chains indicates that appearances are not carried over from one causal step to another. Thus appearances must arise from the final immediate interaction in a causal chain.

2.13 A coherent view of physics

There are several further details and illustrations that I would like to explore in relation to the arguments so far. However, I think it may be helpful to come to what I see as the first main pay-off of these arguments straight away, especially for those readers who may see the arguments as straining the limits of what makes sense and maybe just causing unnecessary mental discomfort.

The first important implication of the complementary view of space, time and everything else is that modern physics need not be seen as posing any special problems about reality. If the dynamic processes, or packets of change, that link manifestations have no interaction-independent qualities then there need be no answer to the question ‘what do the goings on in the world really look like?’ That is to say that the processes that cause appearances do not in themselves have any appearance at all. It is not just, as Kant seems to imply, that we cannot know what the real appearance of the world is, in itself and without us watching; there is no reason to think there is one. Thus, there is no reason to ask the question ‘where exactly is an electron passing through a diffraction grating on its way to a fluorescent detector screen?’. An ‘electron’ is the unfolding of a pattern of change that need not be restricted to one point at any one time despite the fact that all or part of that change is only ever manifest through an interaction at a specific point. (If we allow inanimate operational manifestation it will be on the screen, if not the interaction at that point sets up a chain that leads to manifestations in a viewers brain.) The ‘mystery‘ of quantum superposition is a byproduct of mistakenly trying to think of a dynamic process in terms of a series of manifestations laid in a row.

This issue is not new. When Newton, against his own beliefs, decided that the earth must orbit the sun because of an invisible ‘force’ of gravity operating throughout the space in which the two bodies move he was admitting that dynamic processes do not have to be in a particular place at a particular time. I suspect that most children when first taught about ‘forces’ find the idea rather peculiar, but at the age of about six children take things on trust and if they seem useful they mostly stop worrying. Perhaps if children were taught quantum theory at the same age they would be just as happy with it.

Another way of putting the point is that change cannot be expected to be just a series of the sort of manifest states that are changing. This is the important lesson of Zeno’s paradoxes. The flight of an arrow is not just an infinite set of infinitesimally different states of an arrow at a place. If we run a very fast video camera we might capture ten thousand manifestations but none of these are the change that links them. The motion of the arrow is something else entirely.

The impossibility of motion in the outside world being what is manifest to us through our senses may seem strange because we all think we see movement. However, as I will explore more later, we know enough neurobiology to know that this is not the case. We experience signs caused by changes in the world that are manifest to us as what we call a sense of movement, but they cannot be ‘what physical motion is like in itself’. (The multiple meanings of words make it very difficult to be sure that one can convey these ideas effectively in a brief summary like this, but hopefully in the context of the whole text it will be clear. There is also the potentially confusing fact that in physics momentum can be manifest, but in a context not relevant to our sense of movement.)

Thus, there is no reason to think that we should want to know where ‘particles’ are in quantum physics during the change from a manifest interaction at one place and one at another. A wave function does not tell us that there is a particle here, or there, and it certainly does not tell us that the particle is both here and there. It says that the dynamic disposition that leads to the manifestation, that is particle-like in the sense of being at one location, is spread out.

As I understand it, this way of looking at quantum theory also makes sense of the meaning of the variables given in the wave function. It seemed to come as a surprise to physicists that these variables did not refer to features of a ‘state’ of an entity, like its position or momentum. They seemed to refer to the probabilities that the entity would be observed to have such features. But physicists should have expected this since the wave function is an account of the dynamics and dynamics is a matter of tendency or disposition – of what result is likely to be obtained. In previous theories of physics the dispositions were generally described in terms that assumed infinite divisibility. This meant that there could always be a definite answer to a question about what was expected. A disposition might predict a value of 0.7 for a quantity, so we could be sure it would come out 0.7. However, once a physical theory has built in to it the idea that all manifestations have a finite grain then there can only be probabilities. If our quantity is 1 unit then that theory has to describe dispositions that had a 70% chance of a positive result and a 30% chance of a negative one. The probabilistic nature of quantum theory should have been quite unsurprising. Descriptions of dynamics always gave probabilities in physics. It is just that they mostly gave certainties for single values instead of statistical ranges for multiple values.

So the vast literature about the merits of von Neumann’s, Bohm’s, Everett’s or Zureck’s interpretations of quantum theory misses the whole point of not just quantum physics but all physics. There never was a reason to think that dynamic packets need a ‘physical interpretation’ in terms of the language of manifestations because there is no reason to think of the two in any way similar. Rovelli has recently devised a ‘Relational’ interpretation of quantum theory. I have some sympathy with this because it almost gets rid of the false concept of ‘physicality’. But even so, in the account of his interpretation in the Stanford Encyclopedia of Philosophy Rovelli still talks of a ‘physical content’ of a theory as if the dynamics themselves are not enough.

As mentioned previously, the fundamental distinction between change (dynamics) and the manifest outcome of that change is something that was well known to natural philosophers as far back as we have records. If anything, the further back one goes the more sharply the contrast seems to have been appreciated. Heraclitus and Parmenides are the prime examples. Although they are often seen as giving opposite views, I think there is a sense in which they were expressing the same insight. As I indicated for Descartes and Leibniz, I cannot be sure what these historical figures really thought but I think there is an important potential commonality.

Heraclitus appeared to suggest that all there is is change. Parmenides appears to suggest that nothing changes. Taking into account my previous comments about the need to reassess what we might mean by ‘things’, I think these two positions are compatible in an interesting sense. If we take it that the elements of the universe are themselves ‘goings on’ or, in simple terms, patterns of change, then there is an important sense in which these are just what they are, and do not themselves change. If we see the fabric of the universe as a pattern of dynamics, rather than static structure, then if we call that change then Heraclitus makes sense and if we want to say that this dynamic fabric is just the way the universe is, and will never be ‘otherwise’, then Parmenides makes sense.

Confusion arises because we tend to think of ‘things changing’ as in a tree growing. It may be more helpful for us to think of a tree as a local pattern of dynamics or change, which tends to rather quiet in winter but more overt in spring. It is not so much that the tree changes but that the tree is a pattern of change. I think this is another example of an idea that although initially counterintuitive turns out to fit with intuition rather better than the alternative. As has often been pointed out in philosophy there is a paradox in the idea of a thing, as something with a certain set of features, changing, thereby presumably becoming a different thing since it must now have different features.

I will explore further several of these issues in the following sections. Time needs some more detailed thought, as does the issue of parts and wholes. The key message so far is that consideration of the way we know we ‘see’ things, if unpicked in a way that requires little or no technical knowledge not available to the ancient Greeks, tells us that what we call physics must always have been not an exercise in finding out ‘what the world is like’ but rather ‘what is going on’ that leads to the manifest outcomes, some of which we call our experiences. Releasing physics from any suggestion that it must say what ‘things are like’ allows us to see that quantum theory is no stranger than the laws of Newton and Maxwell. The next stage is to see how we can marry change and the outcomes of change into a single unified view of causality.

Key point: a view of physics based on complementary aspects of dynamics and manifestations, broadly equivalent to Leibniz’s appetitions and apperceptions, provides a coherent framework within which modern physics is no longer problematic.

2.14 Summary

The complementary aspects of dynamics and manifestation, of change and the outcome of change, have featured in accounts of the world back to Pre-Socratic times. They are built into physics at the very beginning. Modern physics forces us to face up to complementarity, but it is nothing new.

The problem about including manifestation in our account of the outside world is that the only manifestations we can be sure about existing are those of our own experiences, which appear to be generated in some unknown internal events. Thus, in Human Knowledge Russell notes:

“There is here a peculiarity: physics never mentions percepts [experiences] except when it speaks of empirical verification of laws; but if its laws are not concerned with percepts, how can percepts verify them?” (p.219)

The need for the manifestations of human experiences to form the ultimate reference point for any scientific law is hidden by various tricks of language and the use of standard devices like rulers and clocks but it is inescapable. All physics ultimately describes dispositional features of the world in terms of the manifestations of experience they determine. Thus it is absurd to suggest that the ‘mental’ world of experience is somehow outside the scope of physics. Physical laws are there to explain experiences. What may be true, nevertheless, is that most of the time science deals with relations in the world described in a way that deliberately bypasses any need to invoke the rules that relate local dynamics to manifestations in the final processes inside the human brain. There have always been exceptions, as when discussing rainbows, dreams, afterimages, ringing in the ears and suchlike. However, since the seventeenth century, when a natural philosopher was equally interested in astronomy and the ‘phantasms of colours’ in the mind, physics and neuropsychology have diverged to such an extent that it has become possible for people to claim that physics does not cover the manifestations of the mental realm.

The language of physics has been built in such a way that the patterns of change it describes can be concatenated to give more extended patterns of change without any reference to any intermediate ‘outcomes of change’. The mathematical formulations just seem to connect up to each other without requiring any account of ‘connections’. To an extent this is rigged. As soon as any random elements are involved, as they are in modern physics, or the description tries to deal with anything more than very simple interactions under straightjacket conditions, as in many body problems, the mathematics rapidly becomes useless. Solutions become non-computable. Yet this is not because the mathematics have anything missing. The connections are invisibly and fully subsumed into them. Nevertheless, the fact remains that to test the validity of the mathematics one always has to place oneself at one such connection point to experience an outcome. For reasons relating to the mechanisms of knowledge this connection point is seen as outside the realm of causation in spacetime -as mental. What we need is an account in which experience is not just Russell’s awkwardly necessary gold standard for the physical but a natural part of a general theory of causation.

Russell was close to AN Whitehead, although Russell clearly did not agree entirely with Whitehead’s position. Whitehead’s idea that what exist in the world are ‘actual occasions’ with both a dynamic and an manifest aspect seems to marry very well with my arguments so far. The pity about Whitehead’s metaphysical writings is that he becomes very difficult to follow and despite being a mathematician, tends to steer away from any application of his fundamental ideas to specific dynamic processes as described by the physics of his time. Russell gets closer but seems to be unclear about the exact relation of fundamental entities to events.

What is particularly attractive about the spirit of Whitehead’s approach is that it tries to marry the third person dynamic account of the world to the first person account of the manifestations of experience. Whitehead wants physics to make sense from the inside. He wants the observer to be a physical entity and, in a limited sense, for physical entities to be observers. The main weakness I see in Whitehead’s model is a lack of emphasis on the inherently asymmetrical nature of a causal occasion.

My discussion so far has focused on shifting from ‘stuff’ to the complementary pairs of change and outcome, dynamic and manifestation. What I would like to do in the next few sections is to go back to some issues of detail that arise in relation to what has been said so far and move a little closer to a general theory of causation. I will then move on to issues of meaning and of knowledge. Finally, I shall focus on the nature of the human subject itself with a view to returning to the issue of how the manifestations of experience can be seen as an integral part of a universal complementarity-based account of the universe.

3 Complementarity: illustrations and explorations

3.15 What do gravity and knight’s moves look like?

A number of issues have been touched on fairly briefly so far, with the aim of coming to a general position reasonably quickly. What I would like to do now is to go back and explore some of the issues further, to illustrate and strengthen the case. The first issue is the extent to which the aspects of physics that are often considered idiosyncrasies of quantum theory and Bohr’s complementarity are in fact inherent in all attempts at an account of how the world works and are likely to have been familiar to natural philosophers for many centuries. My suspicion is that the problem is very general. Nevertheless, I also think there may be specific reasons why quantum theory seems to be just that bit more counterintuitive than what has gone before.

We may kid ourselves that we can ‘see how gravity works’ but we cannot. Consider what Newton must have had to consider to find a way to ‘see how it works’. Perhaps the sun has attached to it a mass of tiny invisible hooks, one for every atom of the earth to link on to by its own hook (we need that otherwise there would be tensions between earth atoms that we do not find). In fact, worse still, every free electron and nucleus in the plasma of the sun would have to have its own set of hooks for each earth atom. It gets a bit crowded. The next problem is that since the earth is moving fast all the atoms would have to change constantly from one hook to the next one belonging to the next bit of space unless the sun hooks moved, but that creates other problems. What happens when the atoms are between hooks? Then we have to consider how the hooks pull. Maybe the sun hook moves a little bit towards the sun first and the atom hook follows. But that cannot be so because the hooks have to move together. But why should the atom hook move until the sun hook has moved? Then there is the problem of how a hook works. Is it made up a ‘hook atoms’ that hold on to each other with hooks. If so where does this stop?

A response to this might be that gravity is not really stuff so does not need to be joined up the way stuff is. At least stuff must be really like something. But modern physics has shown just how empty the concept of stuff is – it can only be described in terms of dispositions like gravitational attraction or resistance to acceleration. Moreover, light is as bad as gravity and even Newton realised that there was no simple way of describing ‘what light is [like]’.

The more you think about it Newton’s theory of gravity is just as much black magic as quantum theory, and by all accounts Newton was at least as worried about this as anybody has been about quantum theory. Leibniz, in his later years, had the edge on Newton, realising that trying to save the theory of gravity by somehow making it mechanical so that you could ‘see how the goings on work’ was futile because you cannot even see how they work for mechanical things like hooks. There is an infinite regress. You can never find out what change is like by thinking in terms of what the change determines.

Why do we have this feeling that we grasp how gravity works but not how quantum mechanics works? I presume that we have a sense that we ‘grasp’ an idea like gravity when our brain, probably after trial and error, finds a set of dynamic routines within its neural networks, that when run through will reliably generate thoughts that match up with what the theory of gravity predicts. These dynamic routines within the brain are not in themselves manifest to us either (although they help to generate experienced manifestations) and they will not be similar to the dynamic processes of gravitational interactions themselves. Nevertheless, there will be certain mathematical rules to the neural dynamics that in some indirect way complement the mathematics of gravity precisely enough for us to find that we can not just guess what manifestations gravity should cause but reliably predict them.

We can expect to have an unconscious pre-set system that complements external world dynamics in this way because we must use such a system all the time for interpreting all our inputs – for instance offsetting the effects of our eye movement when interpreting retinal stimuli. We will also use it for planning movements like throwing a ball for someone to catch. We can also expect this system to rescale any predictions, larger or smaller, to cope with things like perspective, without introducing any issue of ‘minimum grain’, since virtually all our biological functions operate well above the grain of fundamental particles. Newton’s laws assume that there is no grain. The differential calculus they are based on assumes no grain. The problem with quantum theory might be that it introduces an issue of grain so may need to be ‘grasped’ in a way less easily buried in unconscious routines. As soon as you have a minimum grain an indefinitely re-scalable dynamic routine ceases to be appropriate.

Unconscious brain routines do have a grain in the sense of only being able to give certain definite results. In a hearing test we either hear a sound or do not hear a sound. However, if as in quantum theory, grain introduces a point at which results are subject to random statistical scatter, then this is not going to help ‘grasping’. In quantum theory the useful results come from repeating an experiment enough times for the random variation to be factored out.

Maybe the problem is that quantum theory introduces a grain for manifestations but not for dynamics. Underlying quantum dynamics, judged by averaging over many measurements show no grain to extraordinary degrees of precision. Yet, because they dispose to a range of possible manifestation, individual manifestations must be different. (I suspect that an important factor in this is the spatial symmetry of dynamic laws. In simple terms, every direction has to have its fair share but individual manifestations have to take one direction or another.) Quantum theory forces us to model dispositions and manifestations according to different rules. This seems to require more conscious effort at least to begin with, but is not insurmountable; after a period of familiarisation (some remember it taking several months) quantum physicists come to work with their theories with much the same sense of ‘grasping’ as for classical theories. They may have difficulty conveying this to others, but they seem happy that they have the feel of the dynamics. Moreover, there is another dynamic system with a clear grain that a small proportion of the population have been getting the feel of for several hundred years: chess.

Consider first the bishop’s move, which seems to be along a diagonal line. But if a player shifts his pieces diagonally within a square without leaving the square that is not part of a move of chess. It signifies nothing in the game. And a bishop being exactly over the line between squares has no meaning in chess either as a position or indeed as a move. The move can consist of the player picking up the piece, tossing it about in his hand and putting it down in a new place after a sweep of the arm round the back of the board. Even if we consider the squares along the diagonal that the bishop’s move ‘goes past’, being in these places is not in any sense part of the move; it would be the result of a shorter move.

The knight’s move is perhaps even more like quantum theory. At least with very thin chess pieces it would be possible to pretend that a bishop’s move was mimicked by sliding the piece along a diagonal. However, for the knight’s move there is no fact of the matter whether it moves forward then sideways or sideways then forward. Moreover, it does not matter if there are pieces in the squares that either of these sequences would visit. There is just no fact of the matter ‘where the knight is on its way from one square to another’. Similarly, in a theory like quantum theory that states that only certain particular manifestations are possible the impossibility of ‘seeing’ what transitions between these manifestations are like, as if they were manifestations, is inevitable.

One might go back to argue that in real life a knight’s move must have some trajectory. However, this assumes the existence of a continuum of infinitesimal gradations, which needs an infinite structure to encompass. This is unparsimonious at best. We have to postulate infinite layers of complexity within everything – not a small price to pay for a good theory. As soon as one admits that it is more parsimonious to suggest that at some minimum scale there are no ‘in betweens’ you acknowledge that at some scale there is a knight’s move in which there is no fact of the matter what the trajectory is.

What remains unclear to me is to what extent the many physicists who happily ‘ride the bicycle’ of quantum theory appreciate just how unproblematic the metaphysical basis is. (Chess players need not worry.) Many are happy that it gives a ‘complete dynamic description’ but maybe many (even Richard Feynman?) have had a lingering sense that something is a bit weird. That seems to me to be a mistake.

Thus, as I see it, quantum theory only appears more counterintuitive than classical theory because it introduces the need to take grain into account for manifestations but not for dispositions and in doing so it forces us to face up to the difference between dynamics and manifestation, which we should have known about anyway. If we do not face up to that it all seems ‘spooky’, if we do, Zeno of Elea can at last sleep peacefully in the earth in the knowledge that modern physics has finally shown the value of his paradoxes.

Key point: there is nothing new about the fact that the dynamics of modern physics cannot be envisaged, gravity was the same. The existence of a quantum grain in modern physics means we should expect there to be no answer to ‘where objects are in between measurements’.

3.16 Cause: Hume, Mackie, modern physics and Leibniz

I have so far assumed that cause is a valid concept and that the real goings on in the world are causal goings on. Yet there are historical reasons for being sceptical about our concept of cause, both from philosophers like Hume and Mackie and from the form of twentieth century physical theories. These need to be taken seriously but I think that a refined version of the concept of cause can survive, maybe in the form that Leibniz preferred. All dynamic elements appear to be interrelated in a directional sequence, or progress in harmony.

Hume maintained that we have no evidence for causation in the world. All that we have evidence for is the conjunction of observed features of the world. We never observe ‘causing’. I think there is some value in this claim, particularly in the claim that we can never have proof that the dynamics of the world tomorrow will obey those of yesterday. However, we may need to reconsider Hume’s argument if we think of observations as the receipt of manifest signals determined by causal dispositional properties. Thus if I observe the blueness of my cooker I am not simply being informed of its state. The sensory collating system in my brain is informing me of an inference about a causal disposition. It seems that we do not have observations of states outside this context. What Hume may have thought of as observations of the way things are are in fact signs our brains use as tags for inferred causes.

Hume may still be right to say that we have to infer cause and cannot have any direct evidence of it. Nevertheless, it is not quite a question of observing repeated conjunctions and then using them to make inferences. Even our most basic sensations are themselves unconscious causal inferences, based on a system of comparing signals arriving through different routes thereby allowing us to ‘triangulate’ on to an object, and not just in terms of its ‘state’ but in terms of its dispositions.

Thus, although Hume claimed that all we have available as a basis for knowledge are our sensations, in a certain sense we have something else available, which is the unconscious computational machinery we use to extract inferences from patterns of input and present those as sensations. The only problem here is with what I mean by ‘we’. Whatever the sensations are available to may not be what the machinery is available to, or at least not in the same way. As I shall discuss under the issue of knowledge, I think we have to break down the concept of the human subject into experiencing knower and the computational knowing machine that supplies the knower. Perhaps the real lesson here is that Hume was pointing out that our naïve conception of knowledge is unworkable. That still leaves room for some subtler concept of knowledge to hold good.

The validity of a concept of causation has also been brought into question by attempts by philosophers such as Mackie to try to define the everyday meaning of ‘a cause of an event’. It turns out to be somewhat convoluted and hard to pin down. Mackie came up with the INUS definition for a cause: insufficient but necessary or unnecessary but sufficient. However, I think this complexity relates to two peripheral issues. Firstly, it is an illustration of just how much the meaning of a word is context dependent. Secondly, it reflects our desire to focus on those aspects of causality that are most worth remembering in everyday life. Thus, if the toast is burnt we think the fact that the setting on the toaster was too high is more the ‘cause’ than the frequency of the mains AC current, because the former is something that often changes and the latter does not. In any rigorous account of causation we need to give equal value to every dynamic pattern that contributes to ongoing events. As Leibniz put it, every element of the universe progresses in harmony with the whole of the rest of the universe. In modern physics terms, that harmony is importantly restricted by the concept of all causal processes falling within light cones, but the basic idea that every event has an immeasurably complex ‘causal cone’ fanning out behind it remains; as Martin put it, it is better to treat causality like the weather. (MIN 182)

There is also the problem that anything we want to know the cause of must ultimately be something manifest and that raises the question of manifest to what. Leibnizian harmony is a reflection of the ‘perception’ of each individual packet, which is a relation of this packet to everything else. Richard Feynman’s quantum electrodynamics posits a very similar many–to-one relation. I think this asymmetry has been underemphasized and I will argue that it should form the centre of a theory of causation that unifies human experience with the rest of the world.

It has been claimed by some that modern science does not have a place for anything like a traditional conception of cause. Ideas of cause based on familiar analogies like billiard balls colliding have gone. Units of stuff have been replaced by units of change. However, we still have the basic idea that causation is the existence of a sequence to dynamic patterns such that their interactions are always realised with a directionality. It may be, and generally is, perfectly possible for individual dynamic elements to occur in the opposite direction but for any real instance of dynamics one direction is that of cause and effect.

As an aside, this latter point has an interesting implication for any theories that propose ‘reverse causation’ in any situation in which human action is involved (as has been suggested recently by Daryl Bem). This has been called the Bilking Paradox. It turns out that any such theories will predict all possible outcomes, rather in the way that in a syllogism with incompatible premises all possible conclusions are valid. To propose retro-causation is effectively to propose that causes and effects are in a sequence that they are not in, since it is the direction of sequence that defines cause and effect.

The situation might appear to be complicated by certain interpretations of quantum theory in which it is proposed that wave functions can propagate ‘backwards in time’ as well as forwards. What this suggests to me is not that the sequence of cause and effect is reversible, but rather that within the dynamic pattern that links one manifest event to another there is no direction to the metric of spacetime. I will give more detail in the next section on time. Sequence is different from metric – it is an issue of cause and effect rather than just causal dynamic.

Key point: the meaning of causation has changed in modern physics but it still involves rules of association of dynamic patterns in sequence.

3.17 Causal chains and unknown manifestations

It is often said that physics only deals with relational aspects of the world. It is also often said that physics only deals with publicly observable facts. In reality I think physics and related branches of science can and do make use of a wider range of concepts and vocabulary. Nevertheless, in a sense the goal of science is to find ways of showing that understanding of the world can be translated into terms that are purely causal and relational and publicly available. As Russell indicates, as far as possible the ‘gold standard’ of the experience of a human subject is to be kept locked away in a vault, to be brought out for use as infrequently as possible.

This approach is almost certainly not very special to science. It is one we use in all sorts of practical everyday thoughts and discussions. It has something to do with the use of common language but I suspect more fundamentally it has to do with the fact that higher level intelligence, of the sort that humans make use of to survive, has a lot to do with understanding causal chains. Using tools, building shelters, hunting in groups, sowing crops and the like all require not just an ability to predict what will be seen if such and such dynamic disposition operates in such and such a situation, but what further events will be entailed further along a chain. Perhaps man is unique not so much in stringing many words in a chain as in stringing many ideas in a chain.

The scientific approach is to describe the effects of the operation of dynamic dispositions in terms of what dynamic dispositions can be expected to follow. Sodium hydroxide and ferric sulphate mixed in solution have a disposition to interact with each other to precipitate ferric hydroxide. If we look up ‘ferric hydroxide’ in the textbook we find it is a name for a substance with a disposition to re-dissolve in the presence of ammonia or further addition of acids. It can also be manifest to the senses as a red-brown mush, but one can skip looking at that point and still follow the causal chain. All that is needed to validate the scientific account is that the result of the final step of an experimental chain be manifest to the scientist.


Figure 1 shows in diagrammatic form how one might view the traditional scientific account. Note that this is not the picture I intend to end up with! Each arrow is the operation of a dynamic package. The red stars indicate what might seem to be the minimum requirement for manifestations of the sort we call human experiences. To infer the operation of a dynamic disposition we need to have access to manifestation involving evidence of how things started and also of how the ended up. In fact there are all sorts of variants on how this might be achieved and it might be argued that all that is required is one last manifestation that includes data both from immediate observation and from memory of a previous observation, or indeed a comparison of two memories. In practice, however, there are probably always at least two observations even if not quite in the simple relations shown.

The white stars indicate intersections between dynamic packages. The upper arrow and associated white stars are intended to indicate that the chain may diverge or converge. For Martin the white stars are public manifestations. However, as discussed previously, we have reason to think that these have no legitimate metaphysical status. Modern physics makes this explicit because there is no fact of the matter about space, time or anything in the absence of a frame of reference for at least a measuring, if not an observing, relation. Modern science does not in fact allow public manifestations. Yet life would be very much easier if there were some white stars a bit like the red stars out there in the world.

What the white and red stars would seem to have in common is that some sort of ‘causal baton’ is passed from one dynamic packet to the next. We might call it information, although that is a loaded word for other reasons. In some sense it seems that we have to allow that a preceding dynamic packet makes something manifest to the next one in order to inform it of how to proceed further. Eventually, I shall come to argue that we have no reason to think that there is any categorical different between the baton passing of a white star and a red star. However, in order to feel comfortable about that I think it is necessary to see how modern physics suggests that the diagram needs to be redrawn.

The dynamic packets of modern physics do not, in general, pass on information in a one-to-one relay race fashion. The physics of everyday life is almost entirely electrodynamics (the behaviour of electrons and photons), which underlies chemistry, the mechanics of matter and acoustics. As Richard Feynman points out in ‘Q.E.D.’ the progression of a dynamic packet such as a photon is informed via one indivisible relation with everything in the universe that is close enough to inform in the time available. In fact there are situations like reflection from two surfaces that suggest that the photon can ‘know’ about patterns too far away to be within its reach. With a little care one could argue that Leibniz was right to say that the whole universe is indivisibly manifest, at least in operational causal terms, to each of its indivisible (monadic) dynamic units.

This has two implications. One is that each white star is an asymmetrical relation between one dynamic pattern and the rest of the universe. The tiniest interaction is at least as complex as Martin’s weather. The second is that this white star entails countless other white stars that are the reciprocal relations of manifestation or passage of information. Thus if the universe, including packets B, C, D, to Googol, is manifest to A then the universe, including packets A, C, D, to Googol is manifest to B and so on. Because this creates a totally unknowable and non-computable causal mesh we have evolved to live by ceteris paribus (everything else constant) accounts of linear chains of events. We forget that the hammer smashes the vase in part because the vase is standing on a hard table, rather than on an eiderdown, and a meteorite has not just flattened the house anyway. What is manifest to any dynamic unit in this operational sense will be both rich and unified.

For this reason I think we have to replace martin’s ‘dispositional partner’ with whatever pattern the universe is manifesting at a given time. We cannot define dispositions for every possible pattern so for a rigorous metaphysics we have to define them in terms of every pattern they might pair with, which is how science sets about it.

Another key point to raise in relation to modern physics is that the manifestation of the universe to one of its dynamic elements is not only distributed in space, but also in time. This is one of the hardest things to get an intuitive grasp of, but we have to consider dynamics and manifestations (in this sense) both going on all the time. The passage of causal information that might be seen as the ‘end’ of one dynamic package is not the beginning of the next package because this might have ‘started’ a hundred light years earlier. If one can depict this at all it is going to look more like the diagram that I reach in section 3.30, having dealt with a few more counterintuitive issues!

I have hinted how the white stars may begin to resemble red stars, in terms of rich unified manifestations. It is too early to try to equate them but I will introduce briefly here the historical debate on whether or not we should do so. Do we want to have two sorts of intersection between dynamic packets, one with manifestation to something and one without, or one sort? If we take the view that experience is something very special to humans, as Descartes did, then we do want two. If, on the other hand we take Locke’s view that human experience ought to be an example of some more general property of the world that involves everything having a ‘mental’ aspect then it would be better if white and red stars were in the same category, even if they might differ in important details.

Thus we may want to suggest that these white stars are true manifestations in the sense that interactions have an experiential aspect for interacting inanimate entities in the outside world. This is the pan-experientialist approach. It has the advantage that it avoids the issue of how experience might have emerged from nowhere in the world with the evolution of man or perhaps some class of higher animals. If all entities experience their interactions then no sudden change is needed. Neither the historical anthropocentric view associated with religious thinking, nor the simple argument that we only know of experience in humans bear much weight here. In truth ‘I’ only know of ‘my’ experience; we infer its existence in other humans because they show outward signs of experience in the way we note in ourselves. This seems no reason to exclude the possibility of experience for entities that neither speak nor engage in eye contact, (including parts of our brains other than ‘I’). There are good reasons for thinking that the experiences of non-human, and in particular inanimate, subjects would be very different from ours but that is no reason to think there are none.

What is more troublesome about the pan-experientialist hypothesis is that it is not amenable to any direct testing. I do not think that this renders it meaningless but it puts all the work on to sound inference rather than direct observation. A similar situation arises in cosmology and study of the Big Bang, but pan-experientialism is admittedly even worse than this. As a result some would argue that it is simply not worth postulating. I think this is wrong. If we get to a stage where we can address the relationship between human experiences and the local processes that immediately underlie them in a range of situations where the experience has significantly different aspects, like dreaming, sleep, light anaesthesia or use of psychotropic drugs then we may find we can build a theory that links type of experience to local dynamic disposition. Testing may remain limited to the red stars of human experience but it is common enough for testing of general hypotheses to be limited to very narrow contexts in science.

It this point I am going to leave the relation between white and red stars unsettled. Nevertheless, I will return to them in the sections on knowledge and the human subject and suggest that maybe the broader category of manifestation, red or white, has a specific place in physics.

I am aware that my use of words in relation to causal chains may be confusing. I would like to be as clear as possible but I see this in part as due to the very nature of words. Rather than being the rocks on which we build ideas I see them more like stepping stones in a river that get us from one idea to another. In fact I see them as almost like logs we happen to find at hand to throw in a stream in the hope that they will provide enough support to allow us to jump across. The white star world is for most of us most of the time hidden behind an illusory stepping stone ‘physical’ world of ‘how things are’ that functions to link ideas of what is really going on, but which is not in itself real. With care we can infer a story that requires only the dynamic arrows and red stars but it is very hard to talk about! Equally we can believe white stars have the same significance as red stars, but we can never be there to see!

Key point: although we can only be sure of the existence of the manifestations of our own experiences there is a sense in which all causal interactions can be considered to be describable in terms of both dynamics and manifestations.

3.18 Reality check: do we need manifestations to …?

One way of trying to avoid a complementary account of the world of the sort I have developed is to deny that we need any talk of manifestations at all. Perhaps the ‘materialist scientific’ position is that all that we need to concern ourselves with is dynamics – of instances of operation of causal dynamic patterns. If we try and describe what is going on in the world in a reliable ‘objective’ way that is all we seem to need and all we end up with. However, as I have indicated, I think this misses Russell’s point that even if we say that dynamic goings on are reified (given real status) by the fact that they come in instances we have no ultimate definition of what they are instances of unless we use our experience at least as an example of what they might be cashed out as – what they are dispositions to. Perhaps the very reason why manifest experience does not seem to come in to the explanation is that it is what is being explained.

Put differently, causes are meaningless unless they cause some effect and the manifestations of experience are ultimately the only effects we know. That is not to suggest that causes and effects are different sorts of thing. What is here, now, is presumably going on here, now, and what is going on over there is presumably in some sense here and now for something over there. It is just that the world, as we understand it, has these two aspects and you cannot pretend to describe it just in terms of one aspect. You can build a form of words that brackets out one aspect for most ordinary and scientific purposes, but you cannot claim that it is does not exist, any more than you can dance on a moonbeam.

Part of the motivation for denying the need for manifestations in science seems to be a fear of postulating ‘two sorts of stuff’ or two sorts of ‘properties of stuff’. However, that suggests to me a confusion with the lay concept of materialism, a world made of ‘actual physical things’, rather than dispositional packets. For reasons given, I see the concept of materialism not as a new insight of science, but rather as a retreat into a naïve position recognised as such two thousand years ago.

Another way to put the case for manifestations is that we all agree that there is such a thing as the colour blue. As discussed, even if this is a sign of some goings on in the world the same sign is used for at least three quite different sorts of goings on. Moreover, although modern educated adults have some understanding of them, historically, most humans have not. It is not implausible that in fifty years time ‘reflection of light’ will be seen as an illusory concept, but nobody is going to deny the occurrence of blue.

There is another sense in which manifestations perhaps ought to be closer to reality than dynamics. Dynamic entities are inferred putative dispositions described as if without specific context of operation. They are only half the story of an occasion. In a sense science works by abstracting half of a story so that it can be used to make predictions in another context. Manifestations are a full story of an interaction. But they are also in another sense only half a story because they are from a specific asymmetrical view and so do not allow generalisation. Neither is more real than the other, they are just different senses of real.

As to exactly why a full account of the world should need both of sorts of account I am not absolutely sure, but I strongly suspect that it is an inherent aspect of any form of knowing, as I will discuss in the relevant section.

A rather different scepticism can be directed at the idea that manifestations have to be to some entity. There are various motivations for denying that manifest experiences need to be based on what is received by something. One seems to be the concern of ‘the spectre of a Cartesian subject’. To avoid this, it may be implied that manifestations are some sort of event or occasion as defined from an outside, third person point of view. This seems odd since the key feature of experience is that it belongs to an individual, who is for this experience the first person.

Russell asks if a person blind from birth who regains sight comes to know new physical facts. Jackson raised a narrower argument about a scientist, Mary, who had been brought up with no experience of colour, but exhaustive knowledge of physics textbooks. I think the more interesting argument is a wider one. If we progressively remove all sorts of sensations from a person’s history until we have someone who has never had signals from the outside world, could they have a detailed knowledge of physics? As I shall come to under the section on knowledge I think that if we look at the mechanics of this situation we find that no knowledge would be possible because all knowledge builds on comparisons. You can cut one bit out of the meshwork of knowing paths and still keep most things knowing, but if you take out too many paths the whole thing collapses. The key point here is that like all paths, knowing paths have to be to somewhere, and so if we allow it to convey knowledge I see the idea of a manifestation that is not to something as being an oxymoron.

An example of this potential division of approaches relates to the philosophers’ use of reference to C fibres in discussion of the experience of pain. When we have pain it is often as a result of a process that includes the stimulation and consequent firing of nerves known as C fibres. Although it is generally agreed that the experience of pain relates to rather later events within the brain, since these are not as well understood C fibres are used as a model candidate for a domain of experience.

Kripke was probably the first philosopher to refer to C fibres here, in terms of pain corresponding to C fibre stimulation. Subsequently, this has sometimes been modified to pain corresponding to C fibre firing. Stimulation and firing might seem to be no more than two consecutive ‘physical events’. However, there is an important difference between the two concepts. Stimulation implies a relation in terms of an entity having an input. It also may have connotations of that input being ‘felt’ or sensed. Firing implies no particular relation and, if anything, relates to the output of something. To my mind firing is the sort of event that might lead to a manifestation to something other than the entity doing the firing whereas stimulation is the sort of event that might lead to a manifestation to the entity being stimulated.

There is no doubt that postulating an entity that a manifestation is to, raises a problem of just how we construct and justify an account of that entity in terms of its ‘receiving’ dynamics and the domain of those dynamics. Section II will focus on this problem. The problem seems to have bothered philosophers interested in logic, such as Russell, and may be the basis for his apparent desire to dispense with entities that manifestations might be to. He suggests that experiences are the intrinsic qualities of events. The idea would seem to be that the world consists of events, or, in Whitehead’s terms, occasions, defined apparently from some neutral outside viewpoint presumably in terms of interactions of dynamic patterns, and that experience is the intrinsic nature of such events.

To me this runs contra to an intuition that an experience will be based on some sort of input to something, and thus will relate to an event as viewed by one particular protagonist. It also seems at odds with the fact that each of us has a different experience unavailable to others.

Moreover, it seems to run contra to the dynamic structure proposed by modern physics. In quantum theory all the elements of the world that affect the evolution of a packet of causal dynamics like a photon do so in an indivisible way. That is to say that the limitation of the possible patterns of evolution that make up the photon’s history do so all together, not one by one. For the photon there is only really one event – which is its total life history, or indeed ‘sum over histories’. If we then consider any one of the elements in the world that affected that history something similar is likely to apply. That is to say that discrete events in the world of modern physics are as defined in terms of a particular dynamic protagonist in a set of interactions that will form the causal inputs to that protagonist’s life history. Although Whitehead’s idea of preceding occasions being ‘satisfied’ by being taken up into subsequent occasions makes sense in this context what his account does not address is the fact that for every causal interaction there is ‘taking up’ in two opposite directions and the ‘occasions’ may be quite different for the two different directions. In a sense all occasions are autobiographies of dynamic entities. The domains of these occasions are the domains of interaction of one particular ‘subject’ entity – close to what Leibniz considered the monad’s domain of apperception.

To my mind this sort of approach shows why it makes sense to link pain to the stimulation of something, not to the firing of something, unless that firing is considered from the point of view of its effect on some other entity further down the causal chain. It also generates a problem for Russell’s claim that experience is somehow an intrinsic property because an intrinsic property is non-relational. I find it hard to see how a manifestation can be non-relational for the reasons given.

Key point: although it has been popular recently to try to deny the need for appearance to appear to anything, representations to represent to anything or manifestations to be manifest to anything there are reasons for thinking that there is such a need if we are to have a causal account at all.

3.19 Are manifestations causes?

A key reason for separating dynamics from manifestations in the approach developed here is that the two aspects are incommensurable. A description of the taste of salt cannot substitute for a description of the dynamics either of salt ions or of whatever dynamics in a brain underlie the experience of saltiness. In this sense manifestations cannot be dynamics and vice versa.

However, the assumption is that this incommensurability is in some way a function of point of view and that ‘all that is’ can be described both as dynamics and as manifestation. As suggested in the last section it seems that maybe manifestations are half of a story in one sense and dynamics half of a story in another. Moreover, this is entirely consistent with a general ‘common sense’ position. If John says he is in pain we think his claim is caused by him being in pain. If Mary turns her head we may assume this is caused by her hearing an unfamiliar sound. The manifestations of experience do seem to be causes in normal parlance. Yet philosophers wade in and ask ‘How is that possible?’. The answer seems to be that in day to day language use our brains are programmed to make the necessary switch of accounts unnoticed. But if we impose rigour on usage, the incommensurability of the accounts seems to pose an insurmountable problem. There may be an insurmountable problem, but not an ontological one: rather one of how we match accounts. Unfortunately, this problem has historically got mixed up with some genuine ontological puzzles.

Concerns about the ability of ‘physical’ and ‘mental’ events to interact are often traced back to Descartes. Descartes claimed that the mechanical dynamics of bodies like levers and billiard balls could not account for the dynamics of human subjects (souls). What Descartes appears to have correctly identified here is that there is a level of dynamics that cannot be made to fit with the dynamics of familiar large objects described in a (‘classical’) way that treats dynamics and manifestations as commensurable. In simple terms mechanical interactions are seen as having local effects on parts of objects. Descartes argued that souls do not interact by parts, but have one indivisible interaction with the world. It may be that Descartes’s only mistake was to think that this sort of indivisible interaction was peculiar to human souls.

Leibniz realised that it would be more consistent to suggest that this amechanical immediate dynamic applies to everything and is devoid of any sense of ‘bumping into things’. ‘Mechanism’ tends to imply mediation, or steps in between and I see one of Leibniz’s key insights as the idea that at some point mechanism becomes meaningless. The most intimate interactions can, by definition, have no mediation or steps in between. As I shall discuss in the section on meaning there are reasons for thinking that manifestations must be felt at the most immediate level. Interestingly, modern physics holds that the account of immediate interaction is very different from that of more distant, mediated, dynamic relations. Individual dynamic packets just know how their rules of progression relate to rules operating around them, without any ‘bumping into’.

Leibniz calls this immediate relation a progress in harmony based on the perceptions and appetitions of each monad. Thus he saw the world as populated with dynamic packets, each progressing according to its appetition in the context of its perception of the rest of the universe. There is no distinction between physical and psychological here. The popular myth that Leibniz believed in ‘psychophysical parallelism’ is not supported by the text of the Monadology. There is no issue of how physical connects to mental because they are always the two sides of the same coin.

Leibniz’s metaphysics is a form of panexperientialism, with all elements of the universe having a mental aspect, or being the subjects of manifestations. As Skrbina points out, panpsychism, or panexperientialism, is more prevalent in the history of philosophy has been recognised in recent years. The simplistic view that rocks might have thoughts and emotions like us, is clearly unnecessarily absurd. Mental aspects may come in many levels and forms. Leibniz himself suggests that for monads other than human souls manifestations will be through a simple perception, which, unlike the clear apperception of the human soul, will be indistinct and perhaps devoid of any ‘meaning’ in the sense we recognise.

Panexperientialism is often rejected out of hand by contemporary philosophers, for reasons rarely given. I will discuss later how I think one can accept the basic principle of panexperientialism without getting too exercised, and indeed by being agnostic about, what it may be like to be a rock. The advantage of panexperientialism is that it assumes that the analysis of the human subject should follow the same principles as that of any other goings on in the world. It allows room for the human subject to have unique features but it holds that the two features of the world that we recognise, dynamics and manifestations, relate in the same way throughout. That seems to me to be the only workable approach, and I shall follow it.

So, whether or not an instance of transfer of causal information that is a manifestation to something is considered mental or physical, the question is whether or not it is causal. The literature frequently contrasts causal powers, or dispositions, with their manifestations, as if the manifestations were not in themselves causal powers. But this is the illusory ‘public manifestation’ raising its head again. The public manifestation causes nothing; it is just there. Real manifestations pass on information so do cause something.

How can the dynamics that determine manifestations and the manifestations dynamics determine both be causal powers if they are so different? The answer, as I see it, lies in the asymmetrical relation discussed in section 3.17. Causality is the constraint on possibilities for the future that arises from the possibilities generated by a dynamic packet or mode operating in the context of the possibilities manifest to it in the pattern of the rest of the universe. This pattern may be considered a field, in the strict sense of an array of values in space and time, in contrast to the mode, which is a pattern of dynamic progression distributed in space and time. Both the mode and the field constrain the possibilities for the future, so both are causal powers, but they contribute to the particular ‘causal occasion’ in a very different way.

There is a strange irony here. Heil (and indeed Leibniz) indicates that our intuitive distinction between active and passive is illusory. The vase and the hammer are equal protagonists in a smashing event. However, at the fundamental level there is an asymmetry, at least in the way that modern physics formulates things. The irony is that in a sense the ‘active’ dynamic mode is what receives the causal influence from the ‘passive’ field it progresses through. It is the passive part of the occasion that ‘does something’ to the active part. I guess that this might mean that our intuitive concept of ‘action’ does after all have some basis in the fabric of the world. However, if it does it suggests that we usually get it upside down. The action of contemporary philosophy remains a pseudoconcept, as Heil implies.

Having unpicked causality at the fundamental level, there is still a practical need to be able to discuss causation at the level of more familiar aggregate goings on, particularly when it comes to two major topics to come later: meaning in language and knowing. In this context I hope it will be reasonably clear to distinguish the manifestations of experience from ‘dynamics’ in an aggregate sense, used to mean whatever causal chain may lead up to something manifest in experience, where we are interested in the experience per se, rather than any further causal dispositional powers it may be imbued with.

In summary, I do think that manifestations can be said to be causes. One reason why they may seem not to be is that they cannot serve as full accounts of causation because for a given occasion they are only part of the story. Equally, the dynamic component is only the other complementary half. Thus we have a reason why dynamics and manifestations should be incommensurable aspects of the same real world.

Another reason why manifestations may seem not to be causes is that, as indicated in section 3.17, our language for causal chains is built such that any manifestations along the way drop out of the story. An aggregate or overall story of change is built by linking smaller changes with no reference to intermediate outcomes. This is only to be expected if the outcomes are entailed by the antecedent change.

Aggregate causal chain stories of change, or dynamics, are essential for practical description of biological phenomena like language, thinking and knowledge. We have the strange situation that it is reasonable to consider that manifestations are an essential causal ingredient of all causal chains, but at the same time reasonable to consider that they add nothing not accounted for by the dynamics. It looks as if two contradictory statements are true. I do not think we should be too worried about this, as long as the non-verbal ideas being dealt with make sense – and I think they do. We have to accept that changes in context of word usage can do this. Too bad for analytical philosophy if it finds that hard to stomach!

Key point: a description of a pattern in the world in terms of how it is manifest does not lend itself to use as a cause in an explanation. However, there is every reason to think that every manifestation occurs as part of a causal dynamic interaction and in this sense manifestations can be seen as causes.

3.20 Dynamic and manifest time

Having already made some remarks about space and time I need to come back to time to discuss the ways it may differ from space. There has been a suggestion in the philosophical literature that our experience of time may be more direct or veridical than our experience of space. Like William James, I think this may be ill-founded. In later sections on knowledge I will argue that the temporal and spatial frameworks of our knowledge of our own experiences, qua experiences, are very different. However, when it comes to our knowledge of the time and space of external events, which is what I will concentrate on now, I think the differences have more to do with the intrinsic differences between dynamic time and dynamic space, rather than the way we form signs for them.

If this sounds confusing, I apologise, but I think it may be inevitable. I find the task of trying to put into ordinary language the complex contextual variations in meanings of the words time and space bewildering. If the above is unclear I hope it may be clearer on re-reading in the light of the illustrations in this and further sections.

The space and time of our experience are distinct from the spacetime of physics, as indicated above, and in that sense there is a similar disconnect in both aspects,. On the other hand there is no doubt that time is not just a dimension, or perhaps it would be more precise to say that the time component of spacetime is not all there is to our concept of time. Dynamic goings on describable in spacetime also have a sequence. They come before and after. This is often referred to as the asymmetry of time. However, it is perhaps better considered not as an asymmetry in the dimensional framework, which appears to be symmetrical, but as a separate feature of the way reality utilizes that framework. Most other ‘asymmetries’ can be expressed in mathematical terms; I am not sure that sequence can.

Einstein’s special relativity, and the concept of a single metric of spacetime raised the issue that the universe is not changing, but rather exists as a ‘block’ of spacetime, which somehow we only appear to migrate through. However, there is a danger here of suggesting ‘what the universe is really like’ in a way that should be avoided. It is not that the universe is really sitting there like a preformed block, to be observed all of a piece. Observing involves a now. Moreover, it bypasses issues like the nature of information, knowledge and sequence. It might be argued that the machinery of human knowledge should not be allowed to interfere with our conception of the way the world really is. On the other hand it may turn out that the rules of knowledge are very much tied up with the rules of the universe (we are, after all, just bits of universe), and that trying to take a ‘God’s eye view’ of a block universe very much misses the point of what is really going on. Sequence may be as important to everything as it is to us.

Modern physics might seem to change the unidirectionality of causal sequence. The experiments sparked by the Einstein-Podolsky-Rosen paradox suggest that if the spin of one of a pair of entangled photons is measured then this affects the spin of the other photon apparently through a mechanism that involves the whole process of evolution of the photons in spacetime back to their point of emission. However, there is a crucial limitation to this apparent action back in time. It can only apply within the domain of a single quantum system, and one of the tenets of quantum theory is that the entire spacetime domain occupied by that quantum system cannot be divided up into independent dynamic elements in a causal account. That is to say that no element of the dynamics is ever ‘done and dusted’ within a quantum system, just as no element is ‘somewhere else’. What this suggests to me is that we should not so much think of a block universe as of a block domain for each ‘event’ in the universe. These block domains are then connected in an asymmetrical way as causal sequences. I use the analogy of a garment made of sequins each with hooks around the top end and eyes around the bottom end. Each sequin is a block, with no ‘direction’ in it, but all the connections to other sequins can only go one way. Thus the concept of dynamic time is not just a metric. It is two ideas rolled into one – metric and sequence, the first symmetrical but the second not.

For some reason, the dichotomy between dynamic and manifest time has seemed to be more difficult for people to grasp than for space. This may relate to the popular myth that Descartes said that our experiences are not in space (which as far as I can see he did not, in the sense people think he did) but left it to be presumed that they are in time. McTaggart identified a dichotomy in time but appears to have considered this distinct from any dichotomy in space. McTaggart defined an A series time and a B series time, the former with a now and the other without. As far as I can see this distinction is chiefly that between dynamics and manifestation and equally one could have an A and a B space, one with a here and one without. The difference is the issue of sequence, but that applies to both A and B.

It seems to be easier for people to accept that the space of experience is not ‘in the same place’ as what it is about (a tree in the distance for instance) than it is to accept that the time of our experience is not ‘at the same time’ as that which it is about. Even if it is accepted that it will take a little time for signals to travel to and across the brain the thought may be that this just introduces a trivial offset. The trouble is that for things like stars the offset may be millions of years and different for every star. Experience just isn’t ‘at the same time’ as what it is about.

Nor does the intuition that we experience from a point in time hold up. ER Clay famously described our sense of ‘now’ as a specious present, that must have duration, despite seeming like a point. Thus, to sense a melody we must sense a series of notes over a few seconds. To sense the mounting tension in a symphony we must sense the shifting of harmonies over many minutes. This seems to most people paradoxical, although it is the same for space – we sense the pattern of an extended scene. ‘Here’ is just as much a specious vicinity.

It seems likely that there is an important reason for us being surprised by the specious present and not by the specious vicinity. It seems that sensations are presented to us more or less in the form of spacetime ‘sliced by time’ – giving slices with spatial but not temporal extent. The reasons for this probably relate both to the relation between the velocity of light and that of most gross biological processes and to the survival value of having sensations that tell us about sequence, in a way that informs us about the causal dispositions around us.

Another reason for being surprised by the idea that now has a duration is that this would seem to require that our experiences should be four dimensional. For each point in space we should be aware of a series of appearances, all in one experience. Yet we feel that we do not. As I will come to in the next section what I think this illustrates is that the time and space of experience are not in fact dimensions at all. They may be determined by local brain events in spacetime but experienced time and experienced space are not distinct dimensions. If one accepts that causal dynamics and manifestations are not in any way analogous, this becomes easier to grasp.

The situation is , of course, complicated by the fact that for any given experience three time spans are involved. There is the dynamic time span of the scene that the experience is ‘about’ – maybe a batsman playing a cricket ball. There is the dynamic time span of the local brain events that are involved in signs of this scene being manifest to some part of the brain. Then there is the ‘span’ of manifest time of the experience. The second has to follow the first but the third is not time in the same sense. It is a manifest sign or idea, that may include an expectation of a shot to the boundary or anticipation of tea or even a sense that summer is nearly over. It is not a temporal dimension.

This latter point raises the issue of whether our experience is a continuous flow, at least during the waking day, or rather a series of separate ‘frames’, one for each of a series of evanescent subjects (the string of pearls view of consciousness). The arguments for and against will be considered in Part II. One of the arguments against the frame account is that experience appears to be continuous, with no frame changes. However, if one takes into account the idea that experienced time is not a dimension but a set of signs of a dimension the objection recedes. If the signs manifest to the subject do not include a sign for ‘frame change’ then no frame change need be sensed. Within each frame the signs can relate to as narrow or wide a band of world dynamic time as one likes, just as they can for space.

Key point: both ‘space’ and ‘time’ are words that refer both to dynamic metrics and aspects of manifest experience. These are completely different meanings. This is especially difficult to appreciate in terms of time, which is more intimately associated with experience than space, for reasons to be discussed later.

3.21 Manifest movement and motion

I suspect that one difficulty in grasping the incommensurability of change and manifest actuality stems from our apparent ability to ‘see actual change’. We think we see the motion of a squirrel or a flickering flame. Manifestation appears to include dynamics. However, this ‘seeing motion’ must in fact be the receiving of signals, signs or symbols that indicate to us that there has been change. Change itself cannot be carried to the subject down a nerve fibre. What goes down a nerve fibre is a signal that indicates that there has been a change. There may also be ‘top down’ signals that infer change even when there is none, as when we step off a pitching boat. These signals then give rise to an experience that we think of as change; we think we truly experience the change that is the hop of the squirrel but in fact we must be experiencing the reception of a signal that is not itself a change from one state of affairs to another but a single manifestation.

This situation is well grounded in neurophysiology, because we can measure the single impulses from the retina that indicate the changes in light reception that provide the information that we decode as movement in the world. Signals in the visual sensory system as early as the optic nerve fibres do not just encode ‘here now, then over here’. A single signal encodes ‘going left’. There is no reason to think that this is ever retranslated into a series of signals over a period of time further forward in the brain, because that would give rise to the true homunculus fallacy – that you will then need some machinery to re-compare the earlier and later signals and re-infer a movement that was already inferred at the retina.

This underpins the basis of Clay’s specious present and shows that an experience is never of a point in time, nor is it the alternative of a four dimensional ‘movie’. It is a pattern of signs that are in some way ‘compresent’ to the subject, as Whitehead describes it, that can signify across dimensions. A single sign (probably encoded in space as I shall discuss) can encode a velocity or a frequency of flickering or something getting nearer or further away.

The upshot of this is that, as for time and space themselves, motion itself cannot be ‘like anything’, in the intuitive sense. It is hard not to think that swifts in the sky are not really ‘shooting about the place’ in the way that we sense they are. But our sense of ‘shooting about the place’ is merely the way we receive a set of internally generated signs that are not in themselves shooting about the way the swifts are, even if they do have a dynamic history. One might want to reserve the possibility that there really is something that swifts shooting about the sky are like, that God could know. However, if so, then God’s knowing would have to be some supernatural relation so unlike anything to do with us that ‘knowing’ would seem to be an inappropriate term.

On the other hand, although dynamics in the world cannot be like anything in experience, that does not in any way stop us from knowing all there is to know about them. If dynamics are instances of operation of dispositions according to definable mathematical laws and our brains have the ability to infer these mathematical laws, we can know all there is to know. Conventional quantum theory makes the bold claim that its equations are a ‘complete description’ of a pattern of dynamics with no ‘hidden variables’. This may turn out to be optimistic but the idea that such a claim could be reasonable needs taking seriously. The only limitation is that, at least at present, we do not understand the local rules for the ‘feel’ of the manifestation that the immediately associated dispositions determine.

It is interesting to compare the perception of movement with that of ‘even shiny blueness’. In both cases perception involves the receiving of manifest signs of the operation of a disposition. One is to travel in space, the other to reflect certain wavelengths of light in certain directions. On the intuitive view that we ‘see what is there’ the perception of movement is the more puzzling, since motion in dynamic terms is never in one place – it is a change of place. On the complementary view it is the perception of even blueness that is the more difficult to explain, unless we grasp the fact that blueness is a dispositional quality.

The perception of ‘something moving at a place’ also emphasises the error of thinking of perception of even blueness as consisting in receipt of pattern of many signals of blueness laid out in some sort of dimensional array. For movement there is no array of signals ‘it is here, then here, then here etc’, just a signal ‘it’s moving left’. Similarly for blueness it seems likely that there is only one signal received for ‘this is evenly blue’ - that nothing in this signaling is distributed in space in the way that pixels on a computer screen are. Again, it shows that the space of experience is not a set of dimensions within which signals for spatial arrays are laid out in arrays.

Key point: the distinction between metric dynamic time and manifest time is highlighted by the experience of movement, which crosses the dimensions of space and time.

3.22 No empty space inside an atom and no ‘unobservables’

In Newton’s time, the atomic or subatomic grain of the universe was well outside the range of the great majority of physical experiments. Maybe an astronomer might have recorded the presence of a star on the basis of a single photon reaching his retina, but the empirical demonstration of the grain of light had to wait for the development of photoelectric devices. Thinking of dynamic accounts of the world as if they could be thought of in terms of a continuum of manifestation raised few problems. By the late nineteenth century the grain of the world was beginning to be probed. Not only had evidence for atoms been found, but these had been split into fundamental subcomponents, even if it was not until 1910 that the size of the subatomic grain became known (by isolating single electric charges and calculating the associated mass).

For a period around this time the assumption that manifestation and dynamics could be described as if analogous continued. Experiments on collision of subatomic particles led to the idea that atoms consisted of a central nucleus surrounded by ‘mostly empty space’ within which a few sparsely scattered electrons orbited. To this day many schoolchildren are still taught that atoms are ‘mostly empty space’. However, I suspect that there is no sense in which this is a justifiable statement.

If we think of space as part of the spacetime metric of fundamental dynamics it is now clear that the space around an atomic nucleus is completely full, in the sense that all of it falls within the dynamic domain of the associated electrons. An electron orbital is the operation of a dispositional pattern that sits essentially immobile all around the nucleus. One of the paradoxes of the ‘solar system’ model of the atom was that if electrons actually moved around in the electric field of the nucleus they should emit electromagnetic radiation and fall inwards. The quantum mechanical solution to this is that electrons do not move – they are more like standing waves (but not quite), which are dynamic but never change position.

Thus in dynamic terms, the space around a nucleus is chock full.

If we then consider the space in terms of the manifest space of our experience we have the immediate problem that this manifest space is always inferred from visual, auditory or tactile inputs to the nervous system. All of these inputs have a grain much bigger than the domain of electron orbitals. When we look at a bunch of close packed atoms we see what we call ‘solidity’. If it is a piece of glass we feel the solidity. We are used to using magnifying glasses and even robotic micro-instruments, but even these have a limit above atomic grain. However much you refine your micro-instruments you cannot poke them through a lead atom. Moreover, however much you magnify, no light comes through a sheet of lead. The only manifestation to us relevant to tests of the fullness or emptiness of space in an atom is one of fullness – in other words solidity. You can fire neutrons through atoms and sheets of solid material but our sense of emptiness is not a sense of where you can fire neutrons. The ‘empty space’ that neutrons might go through is merely analogy taken to a scale that does not merit such an analogy.

The problem is, of course, that we tend to assume the we can extrapolate our sense of space down to an infinitely small scale and thereby get an idea of ‘what an atom really looks like’. But this is absurd, in the context of any physics that has a finite grain, in the same way that it is absurd to ask exactly what route a knight takes on its way to a new square on the chess board.

Strangely, the idea that atoms are mostly empty space is often presented as an important insight into reality. What I think it is is an illustration of the jungle of confusion you get into if you begin to address the paradoxes of the childhood view of the world without getting through to the other side. You create new problems. In many ways it was a very good thing that Rutherford created these new problems, so that Bohr could tackle them and come to the conclusion that a complementarity based approach was the only possible option. What is perhaps ironic is that if Leibniz had been alive he might have pointed out that this approach was the only thing that made sense anyway.

Another myth that has circulated in twentieth century philosophy is that science postulates ‘unobservables’ like electrons, to explain ‘observables’ such as elephants. The idea seems to be that elephants send elephant signals to us, may be via light, whereas electrons cannot send electron signals. Yet this is nonsense since seeing an elephant is entirely due to light interacting with electrons, i.e. seeing electrons. It is true that we tend not to be able to discern individual electrons but that is not the point. When we stand on a beach we say we can see the finely granular substance we call sand, but we do not distinguish grains. The point is that nothing sends ‘elephant signals’. All our sensations are generated by inference from collation of signals that individually would be meaningless and about nothing. We infer the presence of an elephant and we can infer the presence of a single electron just as well from the click of a Geiger counter.

Key points: the confusing claim that matter is mostly empty space is not the reason for rejecting the intuitive concept of matter as ‘stuff’. The reason is much more fundamental and the empty space claim is probably misleading anyway. Electrons are as observable as elephants.

3.23 Rainbows and flashing cats’ eyes

I have mentioned some of the problems associated with the concept of an object and the associated problem of objects having properties. Rainbows are perhaps the most intriguing example of the way ordinary language can lead us into muddles over this.

A young child will probably think that a rainbow is an object at least in some ways like the sun or the moon, or perhaps a cloud. They may realize that there is something a bit ethereal about a rainbow but at least they will tend to think that there is something particular going on in the sky specifically where the colours seem to be coming from. They may also think that it must be something that other people can see as well, and so a ‘public’ viewable object.

No doubt what a particular child thinks will depend a lot on the age they first see a rainbow and their familiarity with other ‘tricks of nature’ like echoes that we gradually get used to understanding as not quite what they seem. Nevertheless, our language about rainbows sounds much like the way we talk of objects. Rainbows are red on the outer side and violet on the inner side (except for doubles and other exotica). They are curved. They are far away.

If we unpick what these statements mean we find a number of anomalies. To be red usually means to preferentially reflect light of long wavelengths. It can also mean to preferentially emit such light, as for a fire. A third related case is preferential transmission, which applies to wine. Then we have a quite different meaning of red when we talk of red light: the disposition to do something to our sensory apparatus. It may also be taken as implying having a wavelength of the sort that tends to give a sense of red for most people under normal circumstances, even if someone taking digitalis might see it as yellow.

None of these senses of red applies to rainbows. There is no dynamic structure in the sky that has an outer border that does something to red light that the inner border does to violet light. There are water droplets in the sky refracting light but they are neither red nor violet. Moreover, the water droplets creating a sense of red for me will not be the ones creating the sense of red for someone standing a hundred yards away. Mine will be invisible to them and vice versa. There is no coloured structure visible to both.

These problems arise because, like an echo, a rainbow is a dynamic pattern that by definition includes its observer and their relation to the sun and some rain. If there is a rainbow at all, I have to be part of it. It is hard to think of such an entity as an object. On the other hand, if we call it a pattern of goings on associated with a particular type of manifestation and allow that pattern to be partly defined by its observer, at least we have a workable way of describing the world.

An even more peculiar phenomenon along these lines is that of flashing cat’s eyes. For reasons I do not fully understand, over the last decade or so, I have found (like many others) that driving at 100kph at night along motorways the cat’s eyes marking the road lanes flash rapidly like strobe lights at a dance. The flashing is at maybe twenty times per second and is only seen at a particular angle. Cat’s eyes viewed at other angles do not flash. Moreover, if the road is also illuminated well enough to see the surface the cat’s eyes do not flash even if they are easily seen. They only flash if it is too dark to see the road itself.

These new cat’s eyes are apparently LEDs that emit light cyclically at 100-250Hz. Our visual apparatus cannot normally distinguish cyclical changes at this rate. My initial assumption was that what must be happening is that some sign generating mechanism in my ocular physiology has a minimum cycling time of roughly 20Hz and that this gets imposed on signals that suggest to our visual apparatus that flashing is going on but at a speed too fast to signify. This unusual situation is likely to be dependent on the way the bright spots move through the visual field in the absence of any local contextual information and at the same time cycle in intensity.

On further introspection I began to notice that what I sensed as flashing at twenty per second was sometimes replaced by the sense of a static string of about six bright dots …… for each LED. Then I had the peculiar sense that the sensation of flashing and the sensation of a string of dots was the same sensation. It was as if the dimension-crossing signifying in my head had rumbled its guess at 20Hz for 250Hz.

Situations like this are of course typical of a number of standard optical illusions but this one brought home to me just how complex the role of the knowing mechanism of the observer may be in becoming ‘part of an object’. It could be argued that seeing flashing cats’ eyes is an aberration or illusion that should not distract us from ‘normal function’. However, I see this as a very dangerous road. Seeing the flicker of a flame is in a sense just as much an aberration since there cannot be real motion at one place at one time. It may be more useful than seeing flashing cats’ eyes, but ‘illusory’ echoes can be useful in finding one’s way about in the dark (certainly for bats). Seeing close things double is a good indication to hold them a bit further away. The sense of dizziness has the useful function of making you sit down before your balance apparatus is so confused you fall over, and so on.

What I think these examples illustrate is the fact that although science mostly deals with ‘public observables’ it does not always do so and there is no hard and fast boundary at the surface of the human body that divides the human subject from ‘perceived objects’. The surface of the human body often roughly equates to the interface between world and ‘knowing machine’ but this boundary can be shifted either way, in or out (the latter when a blind person ‘feels’ with a stick). This confirms the idea that an actual ‘knower’ within a human body that actually supports a human experience is deeper within, beyond the collating mechanisms that in cases like the flashing cats’ eyes generate a confusing signal.

Key point: most of physics deals with patterns in the outside world but in a few cases we are familiar with dynamic patterns that seem to be external but have the internal workings of the observer as an intrinsic component. The surface of the human body is not the interface between the human subject and causal dynamics, we need to look further in.

3.24 Pragmatism and Positivism

The approach I have been advocating might seem to have some commonalities both with the Pragmatism of William James or Charles Sanders Peirce and the Logical Positivist movement. Both branches of philosophy tended towards the idea that all that was worth discussing was what we have empirical evidence for. Pragmatism held that the distinction between what existed and what we could know about was in practical terms not worth making. Human knowledge was seen as constrained in certain ways such that it was simply unproductive to try to break out of such constraints and propose metaphysical laws that could not be verified.

Logical positivism took the same approach to a further stage, suggesting that unless a statement could be verified it was simply meaningless. This was closely linked to the view of Carnap and others that logic was the underlying fabric of the world.

Positivism has been rejected in recent years by most philosophers. That is not necessarily a reason to do the same, but there do seem to be a number of features of the positivist line that overstep the mark. There is, for instance, a reasonable argument for saying that statements about what is going on inside an astronomical black hole do have meaning, even if those goings on can never be observed directly. On this basis I would support the view that at least some statements that cannot be verified, and thus can never be given a definite truth value, are meaningful.

Nevertheless, I think there is a strong case for saying that there was a grain of truth in the rejection by the Pragmatists and Positivists of statements unverifiable in principle, but neither on the basis of pragmatism nor the requirements of logic. Pragmatism sounds as if it implies a practical approach in terms of ‘the best we can do’ carrying with it the implication that there may be more to the world but we will have to do without it. Logic-based approaches, which tend to assume that the words used in logical constructions have unique context-insensitive meanings such that every statement is true or false seem to me hopelessly unrealistic, at least in the context of natural language. Even Frege and Carnap recognised the pitfalls of applying logical theory to ordinary language.

I think the reasons for thinking that statements unverifiable in principle are at least likely to be meaningless is much more to do with issues raised earlier. If knowledge and verification are usable concepts at all, they are probably very different from what they are generally assumed to be, but I suspect that consistent concepts of this sort relate very closely to the only ‘fabric’ of the world we have reason to think we have any dealings with. The only fabric of the world we have reason to think we have dealings with is one of dynamic dispositions to interactions that at least for human subjects appear as manifestations. Aspects of world fabric not of this sort would be entirely alien to anything we have come across so far. The fact that we think we come across this other sort of ‘material’ fabric all the time needs to be discarded as evidence for any dealings with such fabric. If it was there we would not know.

So when I say that any statement that is likely to be unverifiable in principle is likely to be meaningless I mean that most such statements are likely to refer to kinds that we simply have no reason to think we have ever had dealings with – objects as naively conceived, or pseudoconcepts. If Leibniz had suggested the existence of black holes three hundred years before there was any chance of even inferring their likely existence, let alone making observations that are best explained by their presence, it is fair to say that his suggestion would not have been meaningless. But what Leibniz would have been doing was suggesting the existence of dynamic patterns of the same basic kind as the commonplace dynamics that described everyday life. Black holes would be clusters of monadic dynamic units progressing in harmony with the universe just like planets, fruit and nerve cells. We have every reason to think that there are aspects to the world that we may not infer the existence of for decades or centuries: not even infer that we might ever need to make an inference about such things. Nevertheless, the important point is that these are not aspects of the world that once suggested anyone would suggest are uninferable on principle.

A statement about an aspect of the world would need to be unverifiable in principle on one of two counts. Firstly, it could be about some pattern of causal dynamics that never interacted with any of the causal dynamics we are familiar with in such a way as to affect our experiences. Although one could just about conceive of the existence of dynamic patterns that made no difference at all to our lives even in the most indirect way, what we know so far of the universe is that all the dynamics are in some way or another interconnected. Patterns that did not connect at all with those we are familiar with might arguably be said not to belong to our world. Note that if there were no interaction it would become unclear whether or not it would mean anything to say that the novel patterns were in the same spacetime domain, since spacetime is the metric of causal connections.

So statements about dynamic causal patterns that we can never infer are just about possible within our world. However, the point is that these are not what philosophers would argue about. The second situation, which applies to philosophical objections to a purely dynamic account of the world, is where a statement is made about an aspect of the world that is not a causal dynamic pattern, i.e. is something we can never ever have reason to infer on the basis of evidence, because evidence can only come to us through causal dynamics. In other words, if it is positivist to say that there are no material objects in this world with non-dispositional qualitative properties then I am all in favour!

Key point: talking about things we cannot gain empirical evidence for is not to be rejected as meaningless or useless but it should rightly be treated with scepticism unless it has a clear justification on grounds of e.g. parsimony.

3.25 Parts and wholes: mereology

Having considered the general category of objects, I want to explore the nature of the entities that populate our world further, specifically in relation to the issues of parts and wholes and of what is termed emergence. This exploration may be of most interest for the question of what subjects, rather than objects, are, as will become clear in section II. Nevertheless, there are also reasons for trying to pin down at this stage the most justifiable and parsimonious account of the causal dynamic entities that might be known to a subject (‘objects’ in the popular view).

There is a long tradition in philosophy dealing with the relationship between parts and wholes – so-called mereology. Leibniz dispenses with this in his Monadology in one fell swoop. He claims that the only entities are the universe and indivisible monads. Everything else that gets called an entity will be an aggregate of monads. That does not mean that this aggregate is a whole with parts because it is not in any true sense an entity, just a convenient chunking together of monadic entities.

At first sight modern day science would appear to sit fairly comfortably with Leibniz’s view if monads are fundamental particles. Everything else is just an aggregate of such particles, which we might conveniently call a horse or a quartz crystal, but which is nonetheless just an aggregate. The justification for this can be given on the basis that if an aggregate can be seen as a pattern of dynamics consistent with the summation and implicit interaction of the dynamics of the fundamental units then there is no need to regard the aggregate as some additional entity or whole.

Two lines of thought have been developed that call this simple ‘reductive’ view into question. The first is the ‘systems’ view. In this view it is argued that complex aggregates of interacting elements show patterns of behaviour that cannot be seen simply as the sum of the behaviours of the parts. The occurrence of dominant patterns within systems that obey chaotic rules is often given as an example. Living systems are given as another.

I am personally not convinced that systems of this sort deserve to be regarded as new entities that are wholes with parts. The patterns of dynamics seen in chaotic systems can be modeled using computer programmes that mimic the individual dynamic elements. Despite the overall dynamics seeming surprising, what this demonstrates is that they really are a summation of the dynamics of the elements.

A particular problem with this sort of ‘system whole’ is that it is often very unclear where the margins would be. If we take a computer to be a system then is the mains supply part of the system? If the computer is operating in conjunction with a CD, is the CD part of the system? Are two computers connected by WIFI one or two systems? I will return to the problem of systems as entities in section II, but at this point I simply raise the query as to what extra element makes them a new entity in the form of a whole with parts.

The second line of thought relates to a trend associated with modern field theory and the general concept of asymmetry. It might be thought that the simplest universe would be an empty space containing one fundamental dynamic unit (‘particle’). However, such a universe is in a sense quite complicated in that every point in it will belong to a spherical shell at a distance from the domain of this unit that will differ dynamically from all other shells. If we require there to be at least something going on in a universe, a simpler one might be one that is homogeneously chock solid with indistinguishable dynamic units or just one ubiquitous symmetrical dynamic unit, if such a thing is meaningful. The point is that it may be best to see a dynamic entity as something that can be defined in terms of an asymmetry between its pattern of dynamics and those of the rest of the universe. The corollary is that every asymmetry in the universe is associated with some dynamic pattern or entity.

Much of the discussion of asymmetries in field theory relates to the existence of dynamic units like photons, which are seen as associated with asymmetries that arise when charged dynamic units change their domains in relation to one another. Asymmetries that seem to be more directly pertinent to everyday life are those such as the asymmetry between what we call a solid object and the air-filled space around it.

A solid object differs from air-filled space in its dynamics in that it has a defined spatial domain that can become a parameter of certain energy-bearing patterns of dynamics. These include rotation around a centre of mass, or spinnings, and acoustic modes of vibration, or ringings, whose wavelengths and frequencies derive from the dimensions of the object. These dynamic patterns have a certain amount of energy associated with them and modern physics also requires that this energy is quantised.

The implication of this seems to be as follows. In a domain in which all the dynamic patterns are distributed symmetrically, like an infinite volume of homogeneous hydrogen gas, the only possible dynamic patterns are those based on the laws of familiar fundamental particles. There will be patterns we call protons, or their constituent quarks, and patterns we call electron orbitals. These patterns involve complex harmonic oscillations with notional frequencies determined by the rules of traditional quantum theory. In a domain with an asymmetry, like a nursery roomful of air containing a humming top, new dynamic patterns exist which have frequencies that are determined by macroscopic boundaries in space. The top maybe spins at 20 Hertz and emits a hum at 500 Hertz.

In comparison to the total thermal energy in the room, the energy in these dynamic modes is small. Nevertheless, each mode has its own energy and this is reflected in the fact that such modes are associated with subtle changes in the amount of energy you have to put into the domain to change its temperature. A grosser form of this effect was utilised to make ice cream before refrigerators. If salt is mixed with snow you go from an asymmetrical mixture to symmetrical brine. Because modes that hold energy, which strangely turns out to be negative, are lost, the brine gets even colder than the snow without any heat being lost. If the brine is in a bucket into which a glass of fruit juice has been lowered the fruit juice will be cooled below freezing and form sorbet.

The long and the short of all this is that if the world is not so much populated by objects as by dynamic patterns of goings on then familiar entities like vases and chairs include not only electrons and quarks, but also much bigger energy bearing patterns like acoustic vibrations. Since energy and mass are equivalent and these patterns contain energy, we have no reason to think they are any less real or even ‘physical’ than the quarks.

As alluded to earlier when considering a fragile vase, this suggests that quite apart from its quarks and electrons a vase includes entities that occupy the entire vase. But note that these are not wholes with parts. Vibrations have no parts. They are other entities, some of which may be patterns of disposition that traditionally we called properties of the vase.

CB Martin makes the point that when we say an apple is red and juicy we are in fact describing aspects of two different things. The skin is red and the pulp is juicy. If we follow this line of argument to its logical conclusion we can see that ‘apple’ as a name for an object is merely a convenience that allows us to use one word for all sorts of associated aspects of the world. If I say the apple is heavy everyone knows I am not talking about the aspect that is red. One might think that scientists would tidy up these ambiguities but the fact is that the need to cut corners is just as bad for scientists and if anything they tend to be even worse with multiple usages of words.

In summary, both from the modern science point of view and the philosophical point of view, there are good reasons for thinking that rather than having an idea of wholes with parts it may be more useful in the construction of a coherent world view to stick with Leibniz and just have dynamic units, sometimes associated in common domains. Moreover, we need to consider whether or not in thinking about the relation between the human subject and the world we ought to consider some dynamic patterns that do not fit simply into traditional categories.

If a whole is the sum of its parts it is nothing more than an aggregate of parts. If it is more, and thus other than the sum of its parts then these are not in a legitimate sense parts. The whole is something in itself and as such does not warrant the name ‘whole’ if that implies constitution by parts. If we look at the world in terms of patterns or packets of dynamic disposition then there is no problem about these apparently occupying the same ‘body’. The disposition that is the tendency for a magnet to fall on the floor is different from the disposition to attract iron filings. The former is almost entirely an aspect of the combined dispositions of a number of nuclei. The latter is an aspect of an order state of electron orbitals. Even where dispositions are apparently inseparable there may be an argument for considering them as separate entities. One of the strange things about quarks is that they do not occur in isolation. The reason for this seems to be that so much energy is needed to pull two quarks apart that if applied to the pair it would turn in to two new quarks so that there are now two pairs.

In conclusion, I would suggest that the traditional conceptions of wholes and parts do not deserve to be elevated from everyday talk to the language of metaphysics. Like so many ideas under discussion here, they seem to fall short of usefulness when we want to formulate a coherent comprehensive account.

Key point: the traditional conception of wholes with parts is probably always misleading. Leibniz’s conception of fundamental indivisibles and aggregates of these is much more coherent.

3.26 Emergence

An interesting implication of the above arguments on the existence of large dynamic patterns associated with spatial asymmetries is that modern physics describes the world in a top down rather than bottom up way. These dynamic modes are not summations of small modes. The nature of the small modes may be irrelevant (a flute will play D whether wood or metal). This is another nail in the coffin of the misconception that science is about reduction of large to small. It may be the opposite. The true agenda is to find the minimum necessary set of explanations for what we observe. If an explanation requires another explanation at a smaller scale to underpin it and the smaller scale explanation entails the larger scale explanation then the larger scale one is redundant. However, if the reverse is true there is equally a case for considering the smaller scale explanation redundant.

Philosophers have spent a lot of time wondering whether or not there are properties of entities in the world that ‘emerge’ in the context of complexity. It has been argued that the ‘wetness’ of water cannot be seen simply as the summation of the properties of all the water molecules involved, that a new property has somehow ‘emerged’. This argument is often seen as relevant to the properties of minds, with the suggestion that mentality somehow emerges from the complexity of the nervous system. The wetness of water is probably a poor example, since I am not sure that there is much consensus about what wetness means. If it is liquidity, then that seems to be fairly predictable from the known behaviour of the molecules. If it is something more to do with what water feels like to us, then, as I have indicated above, I do not think we should think of that as a property of the water itself.

A difficulty I have with the concept of emergent properties is just what emerging might be. If elemental units do not have a property but an aggregate of them does then I am not sure why one should see this property as emerging or arising from aspects of the elements. In the top down view of things it is simply that at different scales different rules apply. One might equally think that aspects of dynamic units that only apply where there are, for instance, spatial asymmetries as in a vibrating violin string, and disappear for single atoms are cases of ‘immergence’ rather than emergence.

More importantly, I think it is relevant that emergent properties are often attributed to collections of dynamic elements. As I have argued, I think rather than talking of new properties any additional dispositions not being a summation of those of the elements should be regarded as new emergent dynamic entities that occupy the same domains as the elements. Rather than seeing dispositions as properties of things I think it is more appropriate to see them as dynamic patterns or packets in their own right. The vibration of a violin string is not an emergent property of a lot of catgut molecules, it is itself a dynamic entity that occupies the same domain. It may not have mass but it has energy, so it is just as real as the atoms it cohabits with.

Interestingly, this view of new emergent dynamic entities appears to be incompatible with the idea that systems of interacting units might be new entities. By definition, if units interact, they are separate dynamic patterns and so cannot be parts of a single dynamic packet. If new dynamic patterns emerge in the presence of a system of interacting units then it is the new packet that is the new entity, not the aggregate of the pre-existing dynamic units. This for me becomes crucial in terms of trying to identify the entities that can be subjects. If a subject has experience it must have its own dynamic interaction with the world of which the experience is in some (rather complex) sense a narrative. I see an advantage in having a definition of an entity in both object and subject terms as a necessary dynamic ‘story’ of this sort that is not accounted for by the dynamic stories of other entities.

A number of philosophers, including James Ladyman (Every Thing Must Go) and John Worrall, have argued that we need to move away from a view of the world as populated by ‘things’ and towards a view of patterns of dynamics. Much of what I have said so far is in keeping with their views, although my final conclusions may be rather different from theirs. It can be argued that our classification of entities in the world is to a large degree arbitrary and that as long as we have an accurate model of how things relate it does not matter how fine one divides the world up. The limit seems to a large extent to be determined by a sense of parsimony – of just as many divisions as one needs in order to give an account that fits with experience. This does, nevertheless, seem to give a legitimate basis for defining the minimum number of discrete entities in the world. To my mind, if we believe in the existence of instances of operation of a dynamic pattern that cannot be accounted for simply through summation of other patterns we need to think of this pattern as a discrete entity. Moreover, we may have some guidance about whether this really is the ‘finest cut’ because quantum theory claims that discrete quantised dynamic entities cannot be subdivided. The evolution of causal disposition (using ‘evolution’ in the odd sense that QM leaves us with) in part of the spacetime domain cannot be disentangled from the evolution in other parts of the domain. Perhaps the most salient example of this is the result of the experiments Bell proposed to test the Einstein-Podolski-Rosen paradox.

This discussion of emergence, or coming into being of new dynamic patterns with asymmetric arrangements of matter will be of crucial importance in section II on the nature of the human subject. However, I think it is also an important part of the way one thinks of what is really going on in the world. It emphasises just how little physics has to do with the common man’s idea of the ‘physical’ and how little science has to do with the philosophers’ idea of ‘reductionism’.

Key point: the concept of new properties emerging with complexity of interactions is unjustified but dynamic patterns based on asymmetries may be considered as emergent entities.

3.27 Fermions, bosons and Descartes’s space.

I have kept away from technical issues in physics as far as possible because many would need provision of extensive background. The idea that several real entities rather than one could share the same domain within what we call a vase or an apple may be puzzling and it may be worth explaining in a bit more detail, particularly as I think it is relevant to Descartes’s early attempts to define the scope of physics.

Descartes had the interesting idea that space did not exist other than as a property of matter. He felt there could be no empty space: that even what appeared to be a vacuum must be occupied by subtle matter. In some ways Descartes’s view seems to have been confirmed. Just after the Big Bang it is thought not that only a very small amount of space contained matter but rather that space itself was small and limited to where the matter was. Moreover, the dynamic dispositions that we now see as what constitute matter, despite having localised domains of operation in some respects, technically extend throughout space. (And the Higgs field is probably everywhere.)

Where Descartes was wrong was in the belief that any one patch of space was the property of just one entity – what he called its extension. He described extension as a tendency to exclude other entities from a domain. This works well for rigid solids and can just about be made to work for compressible gases made up of atoms. It did not really work for the vacuums of mercury barometers that Descartes tried to explain and it does not work for electrochemical or nuclear phenomena or light. Nevertheless, modern physics has an equivalent to extension that may make Descartes’s distinction between matter and soul appear more reasonable than is often implied.

The Pauli exclusion principle and its extension to all dynamic elements that have non-integral (e.g. 0.5) spin values states that no two dynamic units with identical dynamic characteristics can occupy the same dynamic domain. These dynamic units are fermions, which are, roughly, the mass-bearing components of matter: electrons and the quarks of the nucleus. Note that they only exclude from their domain other units that have exactly the same characteristics. In atoms, two electrons can occupy any orbital domain, but to do so they must differ in the direction of their spin. If we plug the general fermionic exclusion principle into the dynamic properties of the units we find we can explain why composite matter has the tendency to exclude other matter from its domain. This sort of matter behaves much as Descartes described, in a ‘mechanical’ way. Note also that this mechanically behaving matter is by nature aggregate in that it is composed of a large number of different dynamic packets interacting with each other. An important implication of this is that mechanically behaving matter can be divided up into dynamic parts such that some interaction with the left hand end of an ‘object’ can be considered independent of some interaction with the right hand end.

Modern physics also describes dynamic units, with integral spin, that do not exclude each other from their domains. In fact they have a tendency to share domains. These, the bosons, include the units of light and other electromagnetic radiation, and, if there are any, of gravity. They are material in a sense but they do not behave mechanically because they do not exclude other units from their domain and thus do not form extended aggregates. Instead if bosons ‘combine’ in any sense it is to share the same dynamic disposition or mode. In fact rather than saying that two bosons share the same disposition it may be more appropriate to say that there is a mode of disposition comprising two quanta. The mode rather than the quantum unit seems to be the dynamic packet.

Descartes got several things wrong about his account of the world, but he may have got more right than he is given credit for. He emphasised the unreliability of our intuition about what the world is really like. He emphasised the relative reliability of mathematical rules and suggested that physics should be built around mathematical dynamic concepts. He also pointed out that experiencing human subjects, or for him thinking souls, together with light, did not seem to behave in the mechanical way of ordinary matter with extension. Human subjects seemed to have a single relation to the world that could not be meaningfully divided into right and left hand relations. They also seemed to be located in the same domain that was apparently the extension of the matter of the body. Descartes is often considered to have for this reason treated the soul as ‘outside the physical realm’. However, it is not clear that he would have wanted to endorse such an interpretation. To my mind he was dividing the natural dynamic world into two categories, both of which one might call physical if one wanted. One type behaved in a mechanical extended way and the other did not. The modern distinction between fermions and bosons seems very consistent with that. I will return to this issue in section II.

When it comes to fragile vases, again, the suggestion of additional dynamic entities occupying the same domain as the atoms of the china relates to cohabiting bosons. The bosons in question are of a rather more esoteric form than the photons of light. They are known as phonons and carry the sort of energy that one finds in a sound wave. What is rather confusing about them is that they are considered quanta of energy with dispositions described by versions of the usual equations of quantum theory that involve complex harmonic oscillations with both ‘real’ and ‘imaginary’ components, but their behaviour can also be described using more traditional classical terms. It is unclear how the imaginary aspects of their quantum descriptions should be interpreted. This makes their bona fide reality as real entities open to doubt in some circles, but if we are thinking of a world of dynamic dispositional packets there seems little doubt that they are as real as anything else.

So when talking of emergence of new dynamic entities with the formation of aggregates of atoms, the entities concerned will be comprised of bosons, and very likely phonons.

Key point: Descartes’s division of substances into two types needs revision but the basic insight of two types of dynamic pattern holds up in certain ways; the insight may still be fertile.

3.28 Man as part of his world

Up to this point I have gradually moved towards a position in which the human subject can be seen as a part of his or her world, describable in the same terms as his world. The outside world has things going on and so, we have reason to think, does our inside world. Our inside world involves the manifestations of experience, which appear to be determined by the dynamic goings on. Although we cannot assess manifestations in the outside world in the same way (or indeed in other humans) we can suggest causal junctures, where dynamic dispositions are ‘cashed out’ that can be seen as similar in functional terms.

Treating the human subject as just a rather special case of a general category of entity in the world, with dynamic and manifest aspects is very much what Leibniz’s Monadology is about. Descartes felt that a distinction needed to be made between the human subject and other entities in the world. Ironically, although Descartes suffers a lot of criticism for his analysis, I suggest that it may have been more insightful than some contemporary analyses that strive to find a way to make the human, or perhaps the higher animal, subject uniquely different from the rest of the world. Most contemporary accounts of the human subject base it on some sort of concept of a system of interacting parts. Descartes was of the view that this made no causal sense, and I think he was right, for reasons that I will enlarge on in section II when addressing the nature of the human subject specifically. A human experiencing subject seems to have one relation to its world, not lots of interlinked relations between parts. Some argue for a non-causal account of the human subject but I cannot see how non-causal accounts can be tested for validity. Moreover, since the whole point of causal accounts of the world is to explain the patterns of our experiences there seems to be something badly adrift here.

What I expect to worry a reader at this point is how we can expect a human subject to be a single dynamic packet of the sort described by modern, or indeed traditional, physics. Leibniz gets away with this because nobody knew enough about dynamic packets to challenge him. Now we do know enough to challenge such a proposal. Moreover, many people are resistant to the idea that the elemental packets of quantum theory could be relevant to the sort of biological dynamics we expect to underlie human thought. Quantum theory is generally associated in people’s minds with esoteric microphysical events of a sort that we can confidently exclude from relevance to neurobiological function, even if such micro-events may be relevant to individual molecular interactions, including those involving quite large protein complexes.

There is certainly a major problem to be overcome in bridging a gap between neurobiology as we know it and the idea of elemental dynamic packets having human experiences. Nevertheless, certain misconceptions may make this more of a problem than it need be. One such misconception is that the world obeys two different sets of dynamic laws, at different scales. It seems to be thought by some that the dynamics of small things are quantum dynamics and those of large things ‘classical‘ dynamics. This is not the case. Modern physics holds that all dynamics are quantum dynamics, even at the cosmological level. Some aspects of dynamics at large scale will approximate to the ‘classical’ laws of Newton but others will not.

The term ‘classical’ is very confusing here. It can mean a ‘traditional Newtonian or Maxwellian’ description or it can mean a description in which everything is in one place at any one time. The latter is (arguably) a feature of Newton’s dynamics but that does not mean that quantum theory brings in Newton’s dynamics at a measurement. The use of ‘classical’ in quantum theory refers to the fact that although the dynamics are always distributed in spacetime, manifest measurements or observations yield single values – so that everything is ‘in one place’ at a measurement.

Thus, the ‘quantum/classical divide’ people talk about is not a divide between rules at different sizes, but a divide between dynamics and manifestation that applies at all scales even if at large scales it is sometimes possible to ignore it. So there is no escaping the conclusion that the dynamics of the human subject, if they obey the laws of physics, will be quantum dynamics. As indicated in the previous section, this would seem to have a major potential advantage in that quantum dynamic interactions have a quality of indivisibility that was not part of previous theories and which would provide a potential basis for explaining Descartes’s puzzlement over the dynamics of the human soul.

The difficulty is in finding a dynamic packet that would qualify as unitary in quantum dynamics and would plausibly function as a human subject. That it might have to be massless does not seem to much of a problem since we lose no mass on being anaesthetised or indeed at death. Other attributes seem more difficult to square with the idea of a single dynamic mode in the context of what we know about neurobiology.

I introduce the issue of the nature of the human subject here only to confirm that I do indeed want to find a way of giving an account of the world that treats the human subject like everything else within a single framework of dynamics and manifestations. How far one can go towards pinning down what the dynamics of the human subject could possibly be I will leave to section II.

Key point: attempts to relate the human subject to the world made in the seventeenth century may bear reappraisal for several reasons. Physicists seem to have been much closer to the key problems at that time than they are now.

3.29 God and the God’s eye view

I have always considered myself an atheist. However, on reading Leibniz I was tempted to wonder whether if he could talk of God within a metaphysics that otherwise looks like a near ideal basis for the sort of thinking I prefer, perhaps I could find a meaning for the word God within my own view.

Leibniz makes the slightly surprising claim that the monadic units that populate the universe cannot in themselves have the power to do anything. In order to progress in harmony they must be guided or informed by the laws of God. Thus, without God nothing would happen at all. If, as was common for many natural philosophers in the seventeenth and eighteenth centuries, we take Leibniz to equate the idea of God more to what we might call ‘Nature’ than might be the popular view today this does not seem too unreasonable. In a sense we have to postulate that monads are guided by laws of nature if we are not to have a distinctly inharmonious universe with every monad doing its own thing.

We tend to take it for granted that it is only to be expected that the elements of the universe should all obey the same laws. However, as pointed out by Einstein, this is in a sense the most magical and inexplicable of all aspects of our world. We have absolutely no way of thinking about how it might be that every electron in the universe ‘knows’ how to behave according to exactly the same rules as all the others. The most powerful objection to the concept of free will is that it is very difficult to know what to make of the concept of ‘uncaused cause’ or ‘primary agency’. Yet, as Descartes pointed out, there is a sense in which the universe ought to come to a grinding halt unless there is a God that can keep telling all the bits how to carry on by constantly injecting causation.

At the end of the day, my feeling is that this is an area where we just have to find a convenient form of words and accept that the continuous harmonious progression of the universe is just what does go on in our world and we cannot expect an explanation. So in this sense I can go along with Leibniz’s terminology of ‘Laws of God’.

There is a problem here, nevertheless, if we want to think of God as in any way analogous to the human subject – as is apparently suggested by the Christian concept of man being made in God’s image. In a previous section I suggested that the individual dynamic patterns within the universe can be seen as asymmetries within the totality. An electron is an instance of a pattern of goings on in one domain, with the implication that this is different from what is going on outside the domain. A single electron pattern is not the equal operation of a set of dynamic rules throughout the universe in space and time. If there were no asymmetry between a dynamic unit and the rest of the universe there could be no dynamic interactions between units describable in terms of spatial and temporal location.

The situation becomes even more apparent when we consider manifestations. Manifestations have the aspects of here and now for the entity to which they are manifest. They present themselves to a ‘viewpoint’, even if this may not strictly be a point but rather a limited spacetime domain.

This is where any analogy between the monadic subject and God breaks down because any God that is commensurate with the totality cannot have a viewpoint limited to part of that totality. God is not going to a be an asymmetry within itself. Thus, there can be no ‘God’s eye view’. This is, I think, of relevance to the whole case that I am trying to establish, in that our naïve childhood view of ‘what things are really like’ effectively entails that there should be some ideal, neutral view of things that can be equated with what an all-knowing God would see. The trouble is that a God that is the totality itself is not the sort of thing that could ‘see’ or ‘know’. The problem, as I will discuss later, is that I think knowing requires not only a viewpoint but also a ‘knowing machine’, which for us is a brain, which recombines signals coming from the dynamic pattern to be known in very specific ways. No traditional conception of God is going to know in that sense. If it is proposed that the totality knows itself in some other, magical, way then it would probably be better to use quite a different word, and it would be interesting to have some idea what that word was supposed to mean!

In this context I shall just mention briefly here that I do not see any possibility of a human subject knowing itself. Whatever knowledge is, there are good reasons for thinking that it can never be reflexive in the sense of an entity knowing itself. I will discuss this in more detail in the section on knowledge.

There would seem to be a similar problem in relation to God having any power to ‘intervene’ since this would presumably mean changing the dynamics from the laws of God to something else. Since most religions seem to see the laws of God as being inviolate this does not make much sense.

The central problem then, for a God either as a viewer or an agent, is that these properties would seem to imply relations across asymmetries that in the case of God do not apply. It may be better not to call Nature God.

Key point: if we want to call nature God then that may do no harm, but the idea that such a God might have a viewpoint on the world is probably incoherent and a tendency to assume that a ‘God’s eye view’ is meaningful causes untold trouble.

3.30 A sequential universe

In summary, I would like to suggest that our universe is best conceived of in terms of a network of instances of operation of dynamic dispositional rules. The interaction between one of these instances and the rest of the universe is how I would see what Whitehead called an actual occasion.

The dynamic pattern of each instance will be an energy-bearing mode that can be described from without by quantum theory in terms of a wave function. For a mode encountered in everyday life there will often be more traditional (‘Classical’) descriptions. The pattern presented to this instance by the rest of the universe I will call a manifestation, which, at least for an instance that is a human subject, will be an experience. Thus each occasion is both an episode of dynamic interaction and an experience and is the asymmetric relation between an evolving dynamic dispositional element and the manifest (and thus determinate) presence of other dispositional elements. (See figure 1.) In quantum theory I suspect that this relation more or less equates to the unfolding of the probabilities dictated by the evolving element in the context of the possibilities dictated by the pattern of surrounding elements. However, the concept of causal interaction gets so abstract at the immediate level that Leibniz’s claim that it can only be conceived as ‘progress in harmony’ looks to me very reasonable.

Importantly, in this view both components of the occasion are causal, or dispositional. The asymmetry is between the dynamic and manifest within the occasion. This might seem to be an asymmetry between ‘active’ and ‘passive’ but it is not. All elements are dynamic within their ‘own’ occasions. It is a symmetrical asymmetry. The chessboard analogy may be help again here. A pawn is a packet of dynamic disposition to move a certain way. This disposition limits the future to certain possibilities. The pattern of other pieces on the board also limits possibilities. This might seem to make the pawn active and the other pieces passive. However, in the real world we need to consider something more like two jousting knights. Each is equally ‘active’. That is the symmetry. In a sense there clash is two occasions, one is a dynamic Sir Lancelot interacting with a manifest world that includes Sir Gawain, the ground underfoot and the galloping of the horses and the other is the version for Sir Gawain. In both cases both the dynamic and the manifest component is a causal power and each is both dynamic and manifest in one or other occasion, but in a given occasion only one takes the dynamic role. That is the asymmetry that leads to the misconception that dynamic dispositions and their manifestations are somehow different entities rather than the same entity in different occasions.

Note that there are no ‘things’ which go from potential to actual in this scheme and there are no gaps of potentiality between actualities. What are actual are the occasions, which entail both manifest and dispositional elements. Every element is an instance of operation of a dynamic disposition in its ‘own’ occasion and an aspect of the manifest universe in the next.

A crucial aspect of this model of the universe is that it dissociates sequence from the aspect of time that is part of the metric of dynamics. The dynamics of each occasion has a spacetime metric within which there is no dynamic ‘before’ nor ‘after’ in the sense we normally understand. The domain of the occasion really is a block, both in terms of the way it must be considered from outside and in that the whole of it is a single ‘here’ and ‘now’ for the subject, even if these are ‘specious’ in the sense that they include a sense of distance and duration. The apparent direction of time relates not to this internal metric but to the connection of occasions in a sequence. The directional nature of sequence is inherent in the asymmetry of the manifestation.

I have previously mentioned the analogy with a fabric made of sequins (the etymology is unrelated) joined by a hook and an eye. Within the sequin there is no directionality. Between sequins there is a directional connection. What may confuse in this analogy is the apparent mutual exclusion of the spacetime domains of sequins. This is not how I intend the analogy to be taken. A large number of occasions can operate within the same spacetime domain. The number of mobile phone radio signals passing through the same urban space can be enormous. So the model does not chop spacetime up into bits. (It does not ‘quantise’ spacetime per se.) Rather it says that spacetime is used in bits. (Maybe that means the use is quantised.)

It might be argued that if occasions use bits of spacetime and several occasions can use the same bit of spacetime that space and time must exist prior to anything going on in them. I am suspicious of such an argument if it is a reflection of the intuitive desire for there to be some ‘way things really are’ or something that the universe is really like. As indicated, all I think we need to postulate is that some things are really going on and that these goings on lead to manifestations. I think it is reasonable to say that we should not want a concept of dynamic space or time other than the metric of things that are going on.

There is also the argument that unless the metric was prior then the goings on could not match up in their use of it. We certainly have the inexplicable fact that instances of operation of dynamic rules are all calibrated to match, but we do not want to have to invoke some cosmological rulers and clocks sitting around in the background for goings on to refer to. Perhaps a better way of resolving the issue is to see the dynamic entities that we come to know about as ripples or asymmetries in an otherwise universal ‘dynamic spatiotemporal fabric’. The fact that what was once thought to be empty space is now known to be populated by a variety of distributed dynamic fields may make that easier to contemplate.

The problem is that, ultimately, any attempt to envisage ‘how we should see the world’ must to some degree fall foul of the false intuition that there is a way that the things really are in the childhood sense. A purely dynamic view of underlying reality, reified by instantiation and a relation to manifestations that are undeniably real in another sense has to stand on the basis of the logical arguments presented for it being parsimonious and all that we have reason to think we have ever had evidence for.

An attraction of the accounts of the world given by both Leibniz and Whitehead is their simplicity and global applicability. For Leibniz everything is a monad, for Whitehead an occasion. The schema given so far and illustrated in Figure 1 is in a similar vein. (This is the upgraded version of diagram x earlier.) If we look at the detail of how it might be applied to an account based in modern physics there are, however, questions to be raised about the need for a more complex, heterogeneous model. As mentioned, quantum theory divides fundamental energy-bearing modes into those governed by Fermi statistics (fermions such as electrons) and those governed by Bose statistics (bosons such as photons). Should the ‘autobiographies’ of interaction with the universe of both fermions and bosons equally be considered as Whiteheadian occasions of experience, or are they too different to be treated this way? Do we need further subdivisions for different types of dynamic package? Would, for instance a single occasion of experience for an electron last the billions of years that electron mode may last. It is hard to see why a photon should ever have more than one occasion of experience, but would this be different for fermions?

Life History Ellipsoids

Figure 2. The relationship between instantiated dispositions as elements in the world. A is a dispositional element in a spacetime domain with time vertical and spatial dimensions reduced to one horizontal example. During the evolution or instantiation of the disposition that is A the possibilities in its domain are determined by a pattern of manifest outcomes of the instantiated dispositions e (with spacetime domains shown with time horizontal). The dispositional element A also generates a manifest outcome that will determine possibilities for further elements e (as at top). Physics provides a third person account of how the rules of disposition unfold and how their outcomes alter possibilities for further dispositions evolving. Manifestations are the first person account, as here for A of the pattern of possibilities determined by the outcomes of other elements e. The various e may be considered dispositional partners for A in Martin’s sense but note that all interactions are asymmetric and directional. Influences are from outcomes of dispositional elements on the evolution of other elements.

There is also the issue of what an instance of operation of a dynamic rule might include. In one sense quantum theory does not deal with quanta or modes so much as with changes that determine values for specific variables. Perhaps an occasion of experience is the autobiography of the momentum or spin of a mode. In Bohr’s terms quantum theory is all about setting up experiments that compare final values for a particular measurable variable with initial conditions and thereby allow inference of a dynamic relation. The superposed dynamic domain that yields a determinate value for spin may not be one that yields a value for momentum. This raises the possibility that the subject of an occasion might be not even be an energy-bearing mode but rather some ‘piece of information’ associated with a mode. This idea is unattractive as a basis for a human subject, in terms of taking us even further away from our intuitive idea of entities than energy-bearing modes might. That may not matter, but it also seems to tie things too much to practicalities of laboratory physics that would not seem to be relevant to ‘what is really going on’ in the world outside of a physics lab. My current feeling is that occasions should be associated with modes.

A further point is that, depending on what we consider occasions, the ‘weave’ of the causal fabric may look different from different angles. A photon only ever passes on information at one juncture – where it is absorbed. An electron is continually contributing to new causal interactions. In a sense the life of a photon involves many inputs determining one output, but for the electron things are quite different.

Although I think these caveats are important, there is also a sense in which quantum theory adheres to general principles that apply across the board of dynamics. It seems reasonable to hope that the idea of sequentially connected ‘sequins in spacetime’ can be generalised. There may still be crucial differences in what such a sequin or ‘occasion’ might imply for, for instance, fermions and bosons, evanescent or long lived modes, or modes with contrasting domain geometry, like ‘free-travelling modes (photon, free electron) and ‘standing wave modes (phonon or electron orbital). Moreover, it might turn out that a sequin or occasion of a sort compatible with a human experience could only possibly be associated with one or the other, and indeed only a very specific subtype of whichever category it proves to be.

I have already introduced the suggestion that this approach to the world provides a way to unify concepts of physical and mental. Section II will focus on how one might make use of the approach to move towards identifying the nature of human subjects. However, before moving on to that I need to deal with some further layers of complexity relating to meaning and knowledge. Meaning has to be tackled partly for methodological reasons but also because there seems to something about the relation between dynamics and manifestation that inherently gives rise to multiple meanings. A clear concept of knowledge is also essential to an understanding of why manifestations ‘look’ the way they do. Moreover, it may be that knowledge is intimately related to the fabric of the universe itself.

Key point: separating the concept of spacetime from the concept of sequence and considering all patterns in the world in terms of elements in sequences that can be considered in terms of both dynamics and manifestations provides a framework in which modern physics can fit with no ontological paradoxes.

4 Meaning

4.31 Language as an internal process

Meaning is not purely a matter of language. In a broad sense everything we perceive means something to us, and this will be relevant to what follows. However, language takes meaning to another level, allows us to analyse it, and also brings its own issues of complementarity. It naturally forms the focus of discussion.

The twentieth century philosophical literature seems to have made a particular mess of the analysis of language. Paul Grice introduced some interesting insights into the complexities of communication acts but, otherwise, a string of philosophers, including Frege, Wittgenstein, Quine, Putnam, Fodor, Brandom and Burge have, I believe, generated little other than confusion. The key mistake seems to have been a focus on abstract logical, rather than causal dynamic, relations. As Chomsky points out, language is a causal biological process, like digestion, and needs to be accounted for in terms of what is going on in users‘ brains. The remaining sounds and scripts that link brains together are adequately covered by known physics. And even Chomsky’s internalist programme lacks any commitment to detailed biological mechanisms, which is, I think, why its fruits have been modest. Much of that detail may be, and some may always be, unascertainable but without at least outline proposals of how language might work at the level of nerve cells I see great difficulties in building the sort of causal naturalistic account that we need.

There is a widespread view in philosophy that a causal biological account is inadequate to deal with issues like meaning. Putnam has, for instance, said that ‘meaning just ain’t in the head’. I will argue that if one puts the complementarity of dynamics and manifestation at the centre of the analysis this objection loses force. The interesting aspects of meaning are in heads. And if you look outside heads you take your eye off the only ball in play. What I will argue is that:

1. Language and meaning are internal matters.

2. Each person has their own language and meanings.

3. Meanings are highly sensitive to linguistic and other context.

4. Many words have at least two quite different meanings buried within syntactical structure.

These premises make analysis difficult, but they may also make it possible; without them we are, like Prometheus, enchained to an eternal cycle of empty tautologies!

The rationale for writing this text is entirely dependent on the idea that my words can have meanings such that what they mean to you matches in some way what they mean to me. On a simple view of a language as a set of words that have a single agreed meaning in terms of referring to things in the outside world and an agreed set of rules for joining them together to form sentences, also with meanings, this should not be too difficult. However, as indicated from the outset, one of my main reasons for thinking there was a point in writing was that I believe that a major obstacle to forming a clear workable view of the world is the fact that language is not like that at all. The interesting aspect of language consists of what is going on inside individual users’ heads and that will be different not only for every individual but on every occasion.

There are those who argue that if this is how we must look at language then we cannot possibly explain how we communicate usefully. However, no justification is ever given for this objection, which I think chiefly demonstrates that the objector has not bothered to stop and think how language might actually work. I see no difficulty when it comes to simple practical conversation. When we learn language what matters is that the practical external implications of the ideas we associate with words match those for others, not that the ideas are based on the same internal mechanism, nor that they entrain similar personal experiences. When it comes to discussion of more complex and abstract concepts I think there is every reason to think that communication is indeed a very hit and miss affair. Debates with people talking at cross-purpose are commonplace. However, that does not preclude frequent enough hits to allow us to agree consistently on many things.

On the popular view of language as some public externally defined system of common signs with common referents, when Mary says to Jean ‘John kicked the ball’ what these words mean to Mary not only should be, but, for some, by definition is, what they mean to Jean. Each word in the language brings with it a certain defined meaning and any apparently inappropriate response merely indicates that the hearer does not fully share the language.

The absurdity of this, for me, is illustrated by Tyler Burge’s claim that someone who has a concept of arthritis that does not fit with the concept as defined by experts in arthritis to whom the community defers if in doubt, will, when he uses the word arthritis, mean not his concept but the experts’ concept. I have particular reason to see the absurdity having, as a professor of medicine, spent many years telling students that there is in fact no agreed definition of the word arthritis. The idea that words have these precise definitions held by experts, like holy scrolls in a box, is rubbish. Even in biomedical science we muddle through using words in a cavalier fashion, hoping that, most of the time at least, the context will make the meaning clear. The more one becomes an expert the more one becomes uncertain about the way words are best used.

There is moreover, an absurdity in the idea that one could define what it would be to be enough of an expert to really know the sacred meaning of a word. Inasmuch as arthritis has meanings, they refer to collections of claimed facts spread out over thousands of scientific papers. The validity of these facts is constantly being reappraised. There is no ‘thing’ that one could keep in a box, like the international standard metre, that one could get out and say ‘this is what “arthritis” refers to’. Every time a new person has a new pain in a joint the epidemiology of arthritis, which is part of what the word refers to, changes just a little bit.

I see no alternative to the Chomskyan view that there are no common externally defined languages, only groups of people with internally defined personal languages that may or may not allow useful interpersonal communication most of the time. What I mean by arthritis may fit in functional terms well enough with what you mean by arthritis for you to be able to discuss your pain in the knee with me in a useful way. However, as someone who has many thousands of times tried to explain to individuals whether or not it is sensible to think of themselves as having arthritis I am acutely aware that the meaning of the word will be different, in a variety of respects, for every single individual. To ignore that is, apart from anything else, to be a useless doctor.

The externalist philosophical view of meaning appears to be driven in part by a sense that words and their ‘referents’ have some sort of real but non-causal relation – what is often called the relation of intentionality. I hope to show that this can be accounted for entirely on a causal account, as long as one accepts that causes have manifest effects and that manifestations must figure in the account. The basic problem with any non-causal account is that we have no way of testing its soundness through predictions. In a field bristling with concepts that turn out to be either conflations or no concept at all testing for soundness seems to me to be essential.

Key point: language is largely a matter of internal brain dynamics and the patterns involved will be different for every individual.

4.32 Meaning by, meaning to and other meanings

Before going into a detailed analysis of what I think is interesting about meaning I want to touch on a wider range of usages of ‘to mean’ and ‘meaning’ to clarify certain distinctions made in the past. The diversity of approaches to meaning is illustrated by the entry on Meaning in the online Stanford Encyclopedia of Philosophy. I have a strong sense that this diversity reflects in part the inappropriateness of some of the basic assumptions often made in current philosophy of language. Nevertheless, there are some practical distinctions that are worth going through because they clarify the context of discussion.

Grice developed an account of what it is for someone to mean something by some language-based utterance. This sort of linguistic or ‘non-natural’ ‘meaning by’ was distinguished from ‘natural’ meaning of which an example would be that a sudden chill breeze from under a black cloud means it is just about to rain. Grice showed that for a speaker or utterer to mean something in his non-natural sense involved a complex set of assumptions on the part of both speaker and hearer about the intentions or beliefs of any counterpart in a communication act. Grice’s theoretical interpretation has been challenged but I think there is no doubt that he identified some interesting aspects of our communication faculties.

While Grice’s work is fascinating, I think its interest is outside the scope of my objectives here. I am concerned with what might be called ‘meaning to’ rather than ‘meaning by’. This might seem to account for only half the story (the hearer’s half). However, there are strong reasons for thinking that meaning to is relevant both to hearer and to speaker. Part of the process of uttering is monitoring ones own output to keep track of what it means to oneself. In fact it seems likely that beyond strings of maybe two or three words we usually build utterances up from what we have just heard ourselves say in a way that sometimes can be substituted for by another person with hardly any sense of incongruity. As Kempson has pointed out in the context of the theory of Dynamic Syntax, we often hear this sort of thing:

Elderly woman to security guard “I left the keys on the …”

Daughter “…mantlepiece in the hall…”

Mother “…at the left hand end where I usually do.”

The variability in the way the verb to mean is used is also illustrated by the fact that we talk of a person meaning something and also of a word meaning something. This seems to be a bit like the way we use blue to describe aspects both of things that reflect light and the light itself. It presumably relates to the fact that we take on board words as children without necessarily relating them to any very clear concept, just a cluster of associated ideas. It is only too easy in a discussion of meaning in the abstract to slip from one usage to another without noticing.

There is also the issue that when we ask about meaning of words we are often given other words. However, it seems reasonable to suggest that when we say that filly means young female horse we really mean that the word filly means what the words young female horse mean. The meaning of a word is not another word. An account of meaning that does not give more than a set of rules for translation (rather as for Tarski’s account of truth) from word to word does not seem to address the interesting aspect of meaning, merely the logistics of convention.

These various aspects of the meaning of ‘to mean’ and ‘meaning’ suggest that we should be wary of trying to pin down a single account of how to use these words and what might be the ‘right’ way to interpret them. That does not, I think, mean abandoning any attempt to analyse what meaning is about but it does mean that as one proceeds with the task one needs to be aware that one may be treading on thinner and thinner ice and that every effort needs to be made to support the context of usage from as many sides as possible.

I also think that there is a wrong approach to meaning, which underlies much of the endless argument illustrated in the Stanford Encyclopedia entry. This is illustrated by the extract “Most philosophers of language these days think that the meaning of an expression is a certain sort of entity, and that the job of semantics is to pair expressions with the entities which are their meanings. … the entity corresponding to a sentence is called a proposition …” In the model that I am proposing the only ‘entities’ that we are entitled to think we can know of must be defined either as patterns of operation of dynamics or as manifestations. In both cases they have to be considered in the context of the human brain in which they are instantiated, or if generalisation is possible, in terms of instantiation in brains in general. The goings on in those brains may in turn bear historic causal relations to goings on in the world that we might describe as ‘It is raining.’ or ‘The chair fell over.’ but I see no legitimacy to the idea that the words ‘The chair fell over.’ relate, in some way independent of brains to some real entity called a ‘proposition’. Trying to build theories around entities that are not defined in causal dynamic terms seems to me a recipe for nothing other than disaster and confusion.

Heil gives an interesting critique of what he calls the ‘Picture Theory’ of language – roughly the idea that subjects and predicates in natural language have an uncomplicated and constant relation to objects and their properties in the world. It seems that many philosophers like meaning to be constant and public because it legitimizes the sort of syllogistic arguments they build around concepts like particulars, properties, and natural kinds. Unfortunately, brains do not follow rules that would keep meanings constant in this way.

Key point: the term meaning has several meanings, which are not always adequately distinguished in the literature.

4.33 Why meaning is never constant

One of the most interesting recent insights into language for me is that of the radical contextualists, like Charles Travis, who point out just how context sensitive meaning is. Combined with internalism this suggests that the one thing one cannot assume is that when trying to convey new ideas, i.e. to use words to generate new meanings, that the ‘agreed meanings’ of words will simply do the job as they are. Clearly they cannot. Trying to generate an alternative world-view with words tied to another view is a struggle. Yet we routinely get words to do this sort of work in other contexts, even if the shifts involved are often less dramatic.

As I see it there are very basic reasons why there are no constant meanings for words, even for the same individual. There can only possibly be internal or private languages, and even these are in flux. To my mind, what we mean by saying that a word means something to someone is that when that someone hears the word something goes on inside their head that involves certain ideas being manifest to some subject in the head, and a tendency for the person to behave in certain ways. This may sound like Locke’s view and I am aware that many philosophers think that Locke has been superseded. However, if we want to use language to help our understanding of what is really going on in the world we need to think of what language does in terms of what goes on inside the heads of users. What goes on outside is of trivial interest. Putative relations that are not even goings on are worthless.

On this basis, for there to be one meaning of a word, every brain would have to be set up such that the same things went on when the word arrived. If we take an analogy with an Apple Powerbook computer, a word typed in to the keyboard would have to meet a machine set up in exactly the same way as others, straight from the factory, never used. If one Powerbook was set up with Word 1998 software and another Word 2004 software we could not expect the input to lead to exactly the same goings on, even from the very beginning. It is highly unlikely that any two human brains are wired up exactly the same way. Not only is there genetic diversity, but if the nervous system is anything like the immune system random factors are likely to play a significant part in the way interaction pathways are set up. The outputs to the world may be similar but we are interested in how these outputs arise, not in what they are.

And even this is not the really important issue. From the very first day that a child hears human voices it will be modifying its language software. It seems likely that this occurs to some extent in the way that hard discs get filled up with software and data. Space is allocated on a first come first serve basis and retrieval will involve arbitrary tracking routes that just happen to get to the right address on the disc for that machine. What goes on when a word comes along for a computer or a brain is dependent on a vast number of previous inputs, each of which has in some way added to the way the system is set up to receive the word.

Thus, as soon as a computational device like a brain or a Powerbook has processed some input its internal configuration has changed, its responses have changed and so by definition its internal language has changed. There are new patterns to relate further signals to. The meanings of our words change for us every time we use them. The change may have trivial transient implications as in when I use the word dog to include the implication of ‘the instance that is nearby’. The meaning of Friday changes daily, with shifting differences around the globe. It may be a major permanent change in my private language. The word Florence now refers to my great-niece.

On top of this we have the fact that these issues also apply critically to every other word in the utterance being heard, or the input being typed before the Enter key is pressed. The meaning of the word cheese to me, in terms of the goings on in my head that it invokes, will depend entirely on not only every other word in the utterance it comes in but also what is on the table today or what someone else just said in some other utterance. It might mean the cheese on the table or the cheese in the fridge or the cheese a particular chef recommended for a recipe we might enjoy tomorrow.

One way of looking at this problem is to say that words themselves do not have meanings, but only sentences. The problem with this is that we do think words have meanings in some sense. There is also a view that sentence meaning must be built out of word meaning, a phenomenon known as compositionality. I think the right way to look at this is that word meaning and sentence meaning are different but related dynamic patterns that compose rather in the way that a poker hand composes. A single card has a value but cannot win a round of poker. A hand of five cards can win in a way that is entirely built from the values of the cards and their relations.

The meanings of words are heavily dependent on context. The human brain takes in about 100 million signals per second. Any one of these can alter the meaning of words being heard. One argument for Chomsky’s idea of a built in universal grammar is that children appear to learn language rules too fast for them to be inferred simply from the speech they are exposed to; the so-called ‘poverty of stimulus’ argument. Yet, this does not mean that language is not learnt in an extraordinarily rich sensory context. Moreover, it is learnt in a way that embeds context into structure. We learn to relate contexts not just to words but to sentence structure. To my mind this suggests not only the existence of inborn rules of the sort Chomsky envisages but more specifically a pre-programmed proactive relation-testing system of the sort given in Dynamic Syntax. On hearing a word the system will look for contexts, either verbal or non-verbal, that might relate in certain ways to the word, until a sensible meaning has been established. The sensory systems must also do this in order to gather information from which dynamic patterns can be inferred.

In summary, it is ridiculous to suggest that words have fixed meanings because the whole point of language is to change the meanings of words by using them. Inasmuch as the point of using a word is to engender ideas, it is also to change the ideas to be entrained on subsequent occasions. In real language use a statement is not valued as a truth but as an increment in information. It changes ideas and thereby changes all subsequent dispositions to entrain ideas, i.e. all meanings.

The meaning externalist might argue that adult users of English gradually come to respond to a word like doorkey in the same way, so in some meaningful sense ‘doorkey’ means a doorkey. It does in this trivial sense but the fact is utterly uninteresting. This is meaning in the sense of ‘referring to’ or reference, which is beloved of some philosophers but which does not seem to be definable in terms of any clear account of something going on in the world. Reference is not a sort of goings on. It is another of those pseudodynamic concepts that seems somehow to be a relation in the world but does not seem to have a candidate to fulfill its nature. What we want to know is how this ‘meaning’ works in heads. The fact that it gives the same result is no indicator that it works the same way in each head. An analogue and a digital clock may both read 3.30 but this does not indicate any similarity in the way they work. If we want to understand the disposition to indicate the time or the disposition of words to mean something to people we need to know what is going on not which arbitrary conventions happen to hold.

Key point: inasmuch as language is intended to inform, or generate new ideas, and meaning is the entrainment of ideas, meaning changes with every use as the ideas available change.

4.34 Meanings as dynamics and as manifestations

Locke argued that the meanings of words relates to the ideas they gave rise to in the mind. I think this remains the most useful starting point.

One objection to Locke is the behaviourist point that we can only frame testable theories about words’ meanings to someone in terms of how they change that person’s behaviour, most obviously in terms of the content of their verbal response. ‘Ideas’ in the sense of manifestations to a human subject are, arguably, only understandable by other humans in terms of their dynamic role in behaviour.

One problem I see with this is that I can only define the dynamic roles played by Jack’s ideas in terms of my manifest ideas of entities in the world. (In general, for these sections on meaning, I shall use dynamic in the aggregate causal chain sense discussed in section 3.19, to be contrasted with a final manifestation of experience.) So we need ideas after all. Moreover, functional role in terms of entities of the sort typical of the intuitive conception of ‘objects’ looks to be highly questionable if such concepts have no real basis. This brings us back to Russell’s point that all our explanations of the world are anchored in manifest ideas and that doing away with ideas is not a viable option. Nevertheless, that does not mean that we should only consider meanings in terms of manifest ideas. A definition in terms of behavioural dispositions is just as valid. We want both if we want full complementary accounts.

This discussion reinforces the basic thesis that everything can be considered both in terms of how we might describe it as dynamics and how we might describe it as manifestation. All meanings can be considered in terms of the manifest ideas they engender in the subject and also in terms of the dispositional changes they engender in the individual. Manifest ideas can only be described in terms of the dynamics they relate to and dynamics can only be described with reference to manifest ideas. But that does not mean that they are not both essential facets of reality. In a complementarity-based approach this poses no problem. However, the dichotomy needs to be recognised if one is to avoid all sorts of misunderstandings.

Defining people’s ideas in terms of their behavioural functional roles is all very well; the behaviourist doctrine is perfectly valid for a certain type of problem solving. However, we have to be very careful what we mean by function. Function can mean action, effect or purpose and these are often conflated and shifted to and from in ways that obscure crucial issues of interpretation. A heart contracts. It also pumps blood. In addition it keeps us alive. All of these could be called functions but they are not in any way interchangeable descriptions. The basic problem is that every event or process in the world is bound up in as many ‘functional’ causal chains as there are ways to divide up the history of the universe: a vast number. One might conceivably be able to argue that a functional role defines a (vast) subset of the vast set of historical descriptions but, as indicated later, even this seems to be founded on sand.

To illustrate the most relevant point I will take what I call an SUUS: a screwdriver for unscrewing un-unscrewable screws. A security lock on one of my windows broke one day and I discovered that I could not replace it with a spare I had because it was held in by un-unscrewable screws. However, we had an excellent locksmith and I knew that he had in his bag an SUUS. I have never seen this instrument and have no idea how it can do its job, but I am 100% confident that it exists and does its job well, because he replaced the lock in an hour without drilling anything away.

The SUUS case shows that we tend to describe something in terms of its functional role in the world in the context of what is effectively one of CB Martin’s dispositional partners. An SUUS deals with un-unscrewable screws. However, we have no way of predicting what it would do to anything else. I cannot even be sure that it will deal with Yale as well as Banham un-unscrewable screws. The SUUS functional description has no generalisability as a dynamic description. A major aim of science is to convert dispositional partner-specific descriptions into generalised dispositional descriptions; a tendency of an apple to fall from a tree is replaced by a tendency of any mass to attract other mass.

The relevance of all this to meaning is that if we come to think that we need to see the patterns in the world that we take as ‘referents’ for semantic goings on in brains in a different light then we may find that our definitions of meanings by functional role collapse. If we want a full account of what is going on in the world and how that relates to the manifestations of our experiences we need to have an account of the goings on in brains that correspond most directly to those manifestations. We want to know what the local dynamics of the SUUS inside the ‘black box’ of the brain actually are. Thus we want to be able to describe what a meaning of a word is not just in terms of referents, that may turn out to be illusory, but also in terms of the internal dynamics that mediate the ideas that the word engenders in a brain. In practical terms we want to know what an SUUS looks like.

There is a justifiable practical resistance to defining meaning in terms of internal dynamics associated with ideas. In practice it is extraordinarily difficult to find out what these dynamics are and there are reasons for thinking that at least some details of the dynamics are indeterminable. However, I think the mistake is to think that meaning has to be seen either in external referent terms or internal dynamic terms. There are two complementary meanings to the word meaning, one for each aspect, and we want to use both.

Within philosophy this sort of problem has been described in terms of ‘external’ or ‘internal’ content. The ‘content’ of a mental state, or idea, can be defined either in terms of what is referred to in the outside world or in terms of the ‘nature’ of the local mental state or idea itself. The debate starts with the intuition that each idea must be determined by whatever is going on locally in the brain having the idea. This is then criticised on the grounds that different people, and perhaps more tellingly, different animals, like humans and octopuses, might have quite different goings on in their nervous systems when they are in pain. This suggests that not only ideas relating to outside world referents but also ideas that we think of as having purely internal functions, like pain, can be ‘multiply realisable’. As a result many philosophers have claimed that internally defined content is a misconception – that the content of ideas is entirely related to functional role. However, this is where the SUUS argument about functional role becomes a problem. We find we have a description entirely in terms of function in the context of specific dispositional partners such that the description has no explanatory power at all. In effect the externalist seems to be saying that mental content, as necessarily determined by external functional role, is defined by external functional role. It becomes unclear why there is any need to call this ‘content’ at all since we have an apparent tautology. Moreover, it does not in any way alter the fact that the sensations octopuses have when they look as if they are in pain are quite different from ours at least in certain respects.

The way I look at the problem is that we need to relate meanings to external functional role to have a full description, but the external aspects of the problem are trivial ones dealt with by traditional scientific accounts of cause and effect in the outside world. What we are more interested in in an analysis of how language works is the tricky aspect of the problem, the analysis of the relation of dynamics to manifest ideas inside the brain. This I see as the issue of the local internal determination of content, which is a valid, if seriously difficult enterprise.

Key point: the meaning of a word can be thought of in terms of the ideas it entrains in a hearer and those ideas can be described both in terms of further internal dynamics and of the associated manifestations of experience.

4.35 Dynamic, manifest and intermediate meanings

The complementarity-based approach I am pursuing holds that aspects of the world that in the childhood view are ‘the way things are’ need to be broken down into two complementary conceptions – dynamics and manifestation. These appear to be conflated in our language, although there also appear to be means within the language to disambiguate where it seems important. Thus we understand what it is for something to look red but be yellow.

One way of looking at this is to suggest that the absence of any explicit indication as to whether a word is to be interpreted in terms of dynamics or manifestation is simply one of the many ways in which language use assumes a predetermined context of interpretation and only provides explicit indicators of how words are to be interpreted when this is not preset. Thus, in ‘Tom arrived and John shook his hand.’ the dynamic relationships between people and hands has a default interpretation given by the sentence structure (but overriden in ‘Tom arrived and John shook his hand in the air to signal.’). In other languages word order may not preset interpretation in the same way and words need to be modified, perhaps with suffixes, to indicate roles and relationships.

But this explanation seems too simple because many language users would be surprised by much of what I have said about the differences between dynamic and manifest meanings. At least if this is the way to look at it one must assume that any difference in dynamic handling of words by the brain (to play dynamic or manifest roles in an informative statement) is not overtly recognisable in a difference in the train of manifest ideas evoked. People do not generally admit that they ‘know that in this sentence space has a dynamic meaning and in that an manifest meaning’ even if they may admit that they get a different sense of what is meant by red in ‘it looks red’ and ‘it really is red’ or time in ‘what time is it?’ and ‘how much time do we have?’. Even if our language faculties are built to cope with complementarities, it seems that much of the time our manifest ideas do not always keep up. Since ideas are only signs of dynamics and, like blue, can be used for more than one dynamic disposition this lagging behind is not too surprising.

Difficulties in seeing complementarities at work in language may stem from a traditional teaching that puts too much emphasis on meanings of individual words and on the idea that words are used to build sentences that are evaluable in terms of truth. I suspect that if we look at language in a more dynamic way as an incremental process of informing through the relations between words then the embedding of a system for implicit sorting dynamic from manifest makes more sense. To an extent I am suggesting that like other sensory inputs, language is routinely used as a basis for inferring dynamic aspects of the world. One might say causal aspects, except that, as indicated previously, our conception of cause may need some updating.

I am attracted by Kempson’s theory of Dynamic Syntax, in which hearers assign words in sequence to available roles constrained by innate rules (universal to all human speakers but operated in different communities via a range of different syntactic pointers) in an anticipated truth evaluable unit, or proposition. However, my suspicion is that the innate rules for interpretation of language involve not so much an expectation of completion of a truth-evaluable proposition as completion of a datum from which some dynamic property of the world can be inferred, with the potential for an increment in knowledge.

In this context I shall take a simple sentence as an illustration of how context can determine the roles individual words play in completing this datum, as a prelude to suggesting that the conflation of dynamic and manifest meanings may in part arise from a need for further intermediate meanings entailed by the way we have to use language.

Consider the sentence ‘Mars is red.’ This might be truth evaluable but, ironically, the sentence is in general only much use to someone who cannot evaluate its truth because it is for them a new piece of information. What is important about such a ‘complete sentence’ as an effective piece of communication is that it provides a datum, albeit an indirect one, from which some dynamic pattern in the world can be inferred. What the datum is, and what can be inferred from it, will depend, however, on the context of the statement. Consider it as a response to the following questions from a child.

‘What is Mars like?’

‘Which one of those bright spots is Mars?’

‘What is that bright reddish star?’

In each situation the child starts with certain dispositions to think certain things. After the reply this changes in three different ways. Firstly, a child who has already acquired a disposition to associate the name Mars with some entity, and probably some astronomical entity, conceived in terms of associations with material things and skies perhaps, becomes disposed to attribute this entity with the disposition to reflect, or perhaps emit, ‘red light’, however understood. Secondly, a child looking at the sky who already associated Mars with an astronomical entity will come to attribute to the only reddish bright spot in view all the dynamic features learnt about Mars (orbiting 141 million miles from the sun etc.). Thirdly, a child viewing the sky may come to associate a bright spot identified as reddish with the name Mars and the disposition of other people to call such a spot Mars, whether or not further planetary dynamics are implied.

The point I would like to make is that the words Mars and red are used in these sentences in ways that evoke both dynamic and manifest aspects of the world in ways that provide a datum from which new dynamics can be inferred and that depending on what the hearer knows and the context, the new inferences can be legion. Moreover, they need not involve a simple complementary pair of a dynamic pattern and a manifestation. Association may be between one sort of manifestation and another. The older child will learn in the third case to associate the red spot with a palpably hard rocky surface rather than a searing mass of incandescent gas.

This concept of meaning constantly shifting to suit the dynamic context of each sentence has implications for the text I am writing here. I am acutely aware that I have had to use words in ways that could easily be misconstrued and that without access to the hoped-for set of ideas in the reader’s brain that the result may be unintelligible. But my feeling is that a bit of corner cutting is better than interminable qualification of every word in every sentence. At times I refer to a ‘dynamic account’ or an ‘experiential account’ to distinguish the complementary meanings. My suspicion is that if an account is a full propositional sentence then it is likely to consists of a relation between a more dynamic meaning for some words and a more manifestation-related meaning for others. So things are more complicated. Nevertheless, there is no doubt that some statements seem to be about dynamics and some about manifestations. I see no incompatibility here. If we think of language in dynamic terms as always involving a stepping stone from one state of disposition to another the lack of any central anchor point to meaning is to be expected. I think the mistake in the past has been to assume that propositions have some sort of eternal static meaning that is true or false. I think that approach is closely related to the childhood view of ‘the way things really are’, and I think it needs to be firmly consigned to the waste bin.

Key point: the ideas entrained as the meanings of words can themselves be of dynamics (goings on) or of manifestations (experience). A meaning in each sentence is implicit in the context

4.36 Two complementarities

Following on from the last section, I think it may be useful to consider some of the historical background to the concept of complementarity in the twentieth century. The term is specifically associated with Niels Bohr’s approach to quantum theory, although recognition of the need for complementary aspects in a world view goes back much further and was touched on explicitly by James and Gibson in Bohr’s era. Bohr himself recognised both a narrow technical meaning for the term and a wider application, as did Wolfgang Pauli. (see Atmanspacher and also Folse) Both were probably attracted to the idea that the technical sense of complementarity was an example of a wider principle that was relevant to many aspects of our description of the world, including the dichotomy between ‘mental’ and ‘physical’. This idea may be helpful, but it may also lead to an oversimplified analysis based on false analogies.

My current view is that a general concept of incommensurable complementary aspects to our world description is useful as a way of liberating ones approach from an insistence on a single ‘material’ account. The Presocratic insight that no single account can cover both change and existence is a good starting point. Nevertheless, rather than trying to explain all the difficulties we face in building a world view in terms of a single sort of asymmetry, I think it may be more appropriate to treat each on its own terms – in essence to follow the principle of liberation and not to force all problems into a single Procrustean mould.

A specific reason for thinking along these lines is that my impression is that our description of the world needs to be divided into complementary incommensurable aspects in different ways for quite different reasons. The primary complementarity may be seen as that between dynamics and manifestation. However, in order to be able to relate these two in an explicit way, and certainly in a way that involves language, it seems that we need additional bridging concepts. Most obviously, it seems that in order to talk about goings on in the outside world of common interest, independent of what is going on in our own brains, we want to have a language not just of external cause but also of external effect. We like to have an account of ‘how things are’ or ‘what the world is like’. What we may be describing are dynamic aspects of the world that we expect to give other people as well as ourselves a sense of something like blueness or that the clock stands at half past three but we want to talk of them as effects rather than causes.

The need for such an intermediate concept certainly seems to fit with a need to share concepts and communicate through language and I shall explore its relevance to meaning here. However, when it comes to an analysis of what knowledge might be in the next section I will also suggest that such an intermediate may have a much deeper relation to the way we think about the world that is likely to have preceded any verbal communication in evolution.

One way of looking at the role of an intermediate account is that it divides the primary complementarity of dynamics and manifestation into two subsidiary complementarities: that between dynamics and ‘determinate measurements’ and that between such measurements and the experience we have of them. I think this is useful but the two sorts of complementarity generated are probably not in any sense analogous to each other, except in the sense that they are in some sense incommensurabilities.

My reservation about this way of dividing the dynamic/manifest complementarity is that I think it may be a mistake to think that there is any single clearly defined intermediate. Rather, I suspect that all intermediates are arbitrary, depending on the needs of the situation. (In medical and neuropsychological contexts one is often treating internal events as if external. The way we talk about pain is often anomalous.) Moreover, I suspect that complementary pairs can exist within the intermediate ground such that the primary complementarity can be broken down into an unlimited number of intermediary complementarities. For these reasons I am cautious about suggesting that one should try and define specific sub-complementarities to cover all situations.

I have indicated that the existence of intermediate concepts is probably an inherent requirement of both language use and knowledge-based thought. In addition I think there may be an important reason why we seem to start with the intermediate account and only with time come to distinguish the purely dynamic and experiential aspects. As children we learn by associating ideas, at a stage when these ideas are very poorly formulated. In fact a ‘complete‘ dynamic account of any aspect of the outside world is something that we never actually attain and in childhood we are a long way away from it. Mummy and daddy are distinguished at a stage when the different natures of these individuals is barely known. An experiential account is probably not formulable until quite late in childhood when a clear conception of one’s own identity and status as observer is understood.

Thus, to begin with, the intermediate account of the world is probably conceptually very limited. As time goes by it becomes more sophisticated in its relation to ideas of dynamics and experience. Along the way inconsistencies are likely to be constantly ironed out, with the intermediate account being reformulated. Eventually we may come to realise that the intermediate account is itself nothing more than a stepping stone, of no real ontological status, between what is going on and what we experience.

These issues provide an important reason for rejecting the idea that we should define the meaning of a word in terms of the ‘referent’ it picks out. Words are acquired at a stage when the referent is poorly defined and it never becomes completely defined. Moreover, a clear view of meaning may require the recognition that the elements of the world we thought of as things as children have to be replaced by a completely different set of referents, if indeed the term ‘referent’ is worth retaining.

It might be thought that children would start with an experiential account – just raw sensation. However, as will become clear in the section on knowledge, we may not have access to raw sensation quite in the way we like to think. The sensations of our experience are signs of causal processes in the world and are presented to us as if they were those processes. They are already ‘about’ something.

Key point: there seems to be a practical need in the use of language to bridge dynamic and manifest meanings of words with intermediate meanings that reflect a ‘public’ view of ‘what the world is like’. This figures prominently in science.

4.37 Do words and their meanings exist?

In the public, or externalist, view of languages, words and their meanings have an abstract quality that makes them difficult to analyse. If words are seen purely in terms of private, or internal languages, however, they appear to fit happily into the complementarity-based view of reality being proposed. We do not need to ask if ‘the word “knife” exists, any more than we need to ask if the species Tyrannosaurus rex exists. All that we need is for a pattern of goings on that we identify as an instance of use of a word to have occurred, just as patterns of goings on that we indentify as instances of T. rex have occurred.

This pattern does not have to have some special status as a ‘natural kind’, any more than T. rex. Neither is it an atomic or monadic entity. Both are pragmatically defined aggregate patterns of goings on that will have no clearly defined boundaries (one of the reasons why the principle of bivalency could never work for propositions about objects). There is no answer to the question when did a dinosaur first become a member of the species T. rex, just as there is no answer to the question when did an instance of the word ‘cool’ become an instance of a modern English word. All that we want is to be able to identify a pattern of goings on as a reasonable instance of a use of the word ‘knife’ in terms of its disposition to determine other goings on and/or manifest ideas. An instance of some goings on of this sort is all that we can mean by ‘physically real’.

The goings on that we call instances of word usage come in quite distinct forms, as marks on paper, sounds in the air or even embossed patterns in Braille or hand movements in sign language, but they are all real instances. This highlights the fact that the categories of goings on we call uses of a word are dynamic patterns very much defined in terms of how they influence a dispositional partner. Like the SUUS, they are defined in terms of what they will do to certain other patterns, in this case nervous systems with private languages that consider ‘knife’ to be a word. However, because these nervous systems teach each other to develop a consensus on what qualifies, we can also give pretty reliable definitions, without reference to dispositional partner, of the sorts of marks or acoustic vibrations that will qualify.

As indicated earlier, the external mechanisms involved when words or objects entrain ideas about knives, or whatever, such as sound waves or reflection of light, are all covered to our satisfaction by physics. What are of interest in understanding language are the internal mechanisms that allow an instance of usage of the word knife to entrain ideas that overlap with those entrained by presentation of a knife to view. Of particular interest are the mechanisms by which the specific ideas entrained by a particular instance of usage of the word can be directed by immediately associated words and other non-linguistic aspects of context. Theories of syntax have identified rules for the ways in which ideas entrained by words depend on word order but these are rarely considered in terms of internal mechanisms.

Although philosophers such as Wittgenstein have suggested that it is not until we start using language that we start to think thoughts with logical content I find it hard to see how noises or marks on paper can bring into being ideas that could not be brought into being equally by all sorts of non-linguistic sensory stimuli. Indeed, the role of words would seem to be largely if not entirely confined to triggering ideas already triggered by non-linguistic aspects of the world or at most to co-triggering such ideas in a way that might not have been likely to be triggered by simply wandering about in the world.

If meanings are the dispositions to trigger ideas then we need to consider what sorts of ideas our brains seem to be innately set up to entertain. As indicated in the opening of the introduction, these seem to fall into categories like stuff, things, animals, people, being, changing, doing, desiring and believing. As Hinzen and Uriagereka have pointed out, we have categories of words that work in different ways that reflect these basic concepts. We have mass nouns for stuff that do not need any indication of the presence of individual units. We talk of water, without needing to say a water or this water. We then have nouns that imply individual things, like table. Next we have nouns that imply some sort of inner animacy or agency, like horse. Finally we have nouns for human agents. Similarly we have intransitive verbs that need no object, like shine, and transitive verbs that imply an interaction, like hit. In addition to these there a verbs that imply some sort of completed sequence, like disconnected, and verbs that imply the completion of some goal, like overcame. These nouns and verbs are associated with distinguishing features described by adjectives and adverbs that relate to the modalities of sensory experience and a framework of space and time.

This list seems to be based on a certain range of basic concepts, that look to be innate: material stuff, individual entities, plurality of entities, classes of entity, animacy or agency, change, causal sequence, will and purpose or goal. Words appear to be used as labels or signs for these aspects of the world. The use of signs in language-associated thinking might seem to be a major innovation. However, as argued previously, there are reasons for thinking that all manifest ideas take the form of signs of dynamic goings on in the world. The sensed blueness of my cooker is a sign of a dynamic pattern in the world. What may be different about language may be that it is built out of signs of signs, as for the words’ blueness of the cooker’. Maybe this has something to do with the apparent conflation of dynamics and manifestation not just in language but in thinking.

Very little is known about the code in which signs of goings on in the world or signs of those signs are made manifest in a brain. As will be discussed in Section II, until we have a working model of where manifestations are cashed out it is difficult to build a theory. It may, nevertheless, pay to consider what might be required for such a code to function.

Key point: words can be considered to exist as long as they are seen in terms of individual instances of dynamic patterns within human brains associated with manifestations.

4.38 Manifest meanings must be immediate

Going back to my previous discussion of manifestations, my conclusion was that a manifest Lockean idea triggered by the use of a word must be manifest at some particular causal juncture or nexus. In the case of an idea as the manifestation of a meaning this will be a manifestation with the overt ‘phenomenal’ aspect that we recognise in the experience of a human subject.

Descartes was of the view that the way things ‘feel’ to us must be determined at the point of reception of signals from the world by the human subject (soul). There is no way that we can construct a mechanism that would determine how the signals are interpreted as feel or concept prior to the signals being received. This would seem to endow the subject with remarkable powers of interpretation of signals that Descartes saw in terms of tiny pulling motions of nerves and we would more likely see in terms of chemical or electrical signals at synapses. However, the alternative, that the interpretation in terms of feel is made in advance, seems to be no less mysterious and poses serious paradoxes.

In order for the interpretation of a signal to be made prior to the signal being received it is necessary to postulate that the signal somehow brings something more with it than just its physical dynamic properties (e.g. electrical potential). There is nothing within our conventional understanding of dynamics that can make any sense of this. It would seem to require that at some point in the causal chain, perhaps from a blue sky to the subjective experience of blue, the sequential electrical and chemical signals in the brain acquire a ‘blueness’ tag. That would not be expected to occur in the retina because a sense of sky blue results from the relative pattern of stimulation of red, green and blue cones in comparison to what the rest of the visual field is getting. Sky blueness is not just a passing on of some blue tags from blue cones. It is encoded later in the brain after collation of signals from a large number of receptors. Stimulation of blue cones can give rise to a sense of pale yellow or pink if the pattern of stimulation of other cones is right.

This suggests that the interpretation, if not made at the point of manifestation, must be at some closely antecedent juncture where it has acquired full significance after preceding collation of signals. The paradox here is that there is no causal requirement for such a signal to have a tag of blueness because, as far we know a particular pattern of signals arriving at a point of reception will always signal blueness, in dynamic operational terms, whether or not it has any tag. In other words, as a cause a signal does not need to bring an effect with it to cause what it causes.

The conclusion is that blue feel is encoded in the site of arrival of the signal or the sites of a pattern of signals, not in the nature of the signal. A particular pattern of signals between nerve cells in a brain always has the same significance because of where and when those signals arrive, not because of some other quality of the signals. Each pathway in the brain can only carry signals of one type.

To illustrate this, consider what would happen if pathways carrying signals in a brain A where given a side branch so that whenever they carried signals these would also be carried through a wire to another brain B. If the signals meant ‘red tomato’ in brain A, what would they mean to brain B? The answer is that they could mean absolutely anything to brain B, depending on where the wires were plugged in. They might entrain the idea of a white horse, or a ring of a bell. Nobody may have done such an experiment but experiments on the component events (on outputs from brains and inputs to brains) have been done which allow us to be sure that is the case.

This analysis has implications for the nature of such a subject that will be explored further in section II. I do not intend to open that particular can of worms just yet. However, I introduce the issue at this point to emphasise the need to think of meaning in terms of quite precise causal pathways within brains if we are not going to fall into all sorts of false assumptions. We need to have a reasonably clear idea of what human subjects are. At present there is no consensus and rather little recognition that there is even a question to be answered.

Another way of looking at this issue is to see the brain as an apparatus for asking a series of questions of the world. All it can get from the world is a pattern of ‘yes’s and ‘no’s. No qualities come in from outside. The only qualities it can attribute to the world are ones that it has set up itself as its way of signifying a yes or no answer to a question. If those qualities are manifest they must be manifest at the point where the combined external and internal causal dynamic pathways give rise to an effect in the experiencing subject. This suggestion may seem to be outside the standard ‘materialistic’ view of the world but at least it does not conflict with it. To suggest that in addition to the causes we describe in physics there are manifest effects is merely to fill in a gap that people like Descartes, Newton, Liebniz, and Russell have indicated exists. To suggest that qualities can somehow ride piggy-back on casual chains, or in other words for causes to bring effects along with them is to make a nonsense of the distinction between cause and effect.

The interpretation of the sentence structures of language in terms of meanings is often seen as very different from the interpretation of ‘raw sensory inputs’. A distinction between conceptual and perceptual content of experience is often made in this regard. However, as indicated in the account of the blue cooker, we have reason to think that ‘raw sensory’ experience is in fact the result of an unconscious inference from comparison of current and past input patterns to a sign of some dynamic disposition. The rules of inference from the sounds of the words in a sentence coming in to the ears to a ‘propositional concept’ will be specialised in important respects but may not be as different as is often assumed.

One of the distinctions between what ‘raw’ sensations seem to mean to us and what words seem to mean is that raw sensations tend to be more closely associated with what is sometimes called ‘thick’ experience. Thick experience is the red of the tomato actually in front of us or the chime of the bell reaching our ear. Thin experience is the sense of red or a chime we get when we hear or read words referring to these. Nevertheless, the more one introspects the less clear cut the distinction made here seems to be. When I looked at the most typically blue parts of the cooker they ceased to be blue. When I see a tomato the thick sense of a globe seems to need a thin sense of the far side not in view. When I hear a tune how much thicker are the notes in the immediate past that together with the note of the immediate present form the melody than the tune in the head of a conductor reading a score in bed or the parlour scene in the head of someone reading Henry James?

I mention the distinction between thick and thin experience here in the context of language meaning because I think it raises seriously difficult questions about the homogeneity or heterogeneity, and also of the unity, of any model of manifestation to the human subject. It somehow seems more certain that the green of a leaf is manifest than the idea that capitalism is on the brink of collapse. I can personally see no alternative to considering both being manifest to ‘me’ but I am very prepared to discover that there is a false premise lurking here.

Key point: since manifestations are associated with the final immediate causal interaction in a causal chain, so must manifest meanings, as recognised by Descartes.

4.39. Pseudoconcepts and pseudodynamics

I have mentioned at various points the idea of a pseudoconcept, or a pseudodynamic concept and it may be useful just to take stock of some of the illustrations of this idea that are relevant to the analysis.

If our manifest ideas are signs of dynamics, caused by internal brain dynamics that are in some way usefully appropriate to some dynamic aspect of the world then we can expect all concepts to be to some extent arbitrary, incomplete or ill-defined. However, a lot of them seem to work well, without leading to serious contradictions or mysteries. The concept of a tree being blown over by the wind seems to work pretty well, despite the fact that there are all sorts of ways in which such a conception could be considered only partially valid. In contrast, when I talk of pseudoconcepts I am interested in manifest ideas that we have a sense of belonging to some dynamic aspect of the world but which on closer inspection there does not seem to be any dynamic aspect of the world that could be signified.

Perhaps the paradigmatic pseudodynamic concept is agency. We all grow up with a sense of some event being an action of an agent. We think of ourselves as agents. We have an idea that ‘if I had not hit the ball it would have bounced past’. Agency is mixed up with the idea of ‘choice’ in which we can ‘change the way things would have been’. An essential aspect of our sense of agency is that we do not think it applies to events like rocks falling down a hill on to cars. On the other hand we might be ambivalent about whether or not a wave washing away a house, or a gust of wind felling a tree in a storm were agents.

There is no doubt that the causal pathways involved in human activities like hitting a ball are different in important respects from those of a rock falling down a hill. Complex tracking of external events by internal brain events is involved and self-sustaining ordered states are involved. However, these do not legitimise the central felt idea that somehow agency is different in that it changes what would otherwise have occurred. The reality is that an ‘action’ is an event that was going to occur just like any other. The certainty that it was going to occur might not be absolute by any means, because the world is not totally deterministic in that sense, but this is not something that singles out actions from other events.

Put another way, free will has no coherent meaning as a concept, in the sense that even if there was some strange exceptional causal path that was somehow free of natural rules there does not seem to be any way in which it could be of any use. Certainly the random component of quantum physics would be of no use to an ‘agent’. To be any use agency, or choice, ought to follow reason – i.e. to follow sensible rules.

A whole range of verbs seem to cover action concepts that are pseudodynamic in a slightly less irredeemable way. Seeing and knowing are particularly relevant to the analysis here. ‘I see John.’ is not about something I do to John. Slotting see and know into our verb machinery seems to be handled well enough for most practical situations. However, at the level of metaphysical discussion, as here, it is clear that we need to back off any sense that to see something is somehow to get direct knowledge of the way it is. You can define it as such, as a direct realist might, but there is no causal dynamic account that allows such a definition to lead on to a useful logical analysis. As a result philosophers construct syllogisms that may be valid but lead absolutely nowhere because they are unsound.

Similarly, our concept of an object works well in everyday life, but starts to fall apart if we are engaging in metaphysical discussion. Moreover, it probably only works in everyday life because the shakiness of the concept is compensated for in the way we learn to use language. If an apple is red and juicy then there ought to be something that is both red and juicy. Yet the skin is red and not juicy and the flesh is the reverse.

Almost everything in the childhood view might be considered a pseudoconcept to the extent that there is a conflation of dynamic and manifest aspects. Thus colours need to be understood as involving two quite different sorts of concept, one at least of which may be a rag-bag of diverse concepts rather than one. In general we all come to realise this double-aspect nature of colours. What seems much more difficult is to appreciate the similar double-aspect nature of concepts like space and time. Everything has two complementary aspects, including the concept of ‘reality’ itself. Maybe the mistake made by the Presocratics, and also by Plato, was to think that there could be only one ‘real realm’ and therefore only one meaning of the term reality. I think we need to take on board the idea that although there may be only one real realm, that realm has two incommensurable complementary aspects that are both ‘real’.

Key point: many of the concepts we acquire during childhood turn out to have no possible coherent dynamic meaning. Much of the job of science and philosophy is to expose these pseudoconcepts (or pseudodynamic concepts) and replace them with coherent ideas.

4.40 Aboutness

Philosophers in the twentieth century have been much exercised by an idea they call intentionality. The word is used in all sorts of different ways with different implications. That in itself may be no bad thing but it is often very difficult to establish from context which way the word is being used on each occasion. It sometimes gets mixed up with the idea of intent, and when it does I usually find it hard to grasp what is being meant at all. In contrast to this, the ‘purist’ use, which is of interest here, has nothing to do with intent but just with what has been called ‘aboutness’. Our sensations appear to be ‘about’ goings on in the world, as do our words.

If I could find evidence for at least some philosophers using the term intentionality in an explicit and consistent way I might be tempted to follow. However, the more I read the less I am convinced that anyone uses the term consistently. I think it is high time it was abandoned.

Leaving aside the troublesome word intentionality, I do agree that the feature of aboutness that we seem to find in our manifest experience and in our language does deserve some pondering, if only because it shows how complex things are. Fodor has claimed that the standard materialist view of the world simply does not seem to have any place for aboutness. Attempts at definition of meaning, including Fodor’s, founded on conceptions of aboutness, all seem to fail to give a reliable account.

I think there is an argument for being almost as cynical about aboutness as intentionality that provides part of the answer to Fodor’s puzzle. ‘About’ often means ‘in the region of’ and I suspect that aboutness actually means little more than ‘seeming to be in the causal region of’ or ‘often seeming causally close to’. We tend to think that our intuitive grasp of using the word in all sorts of different contexts must be based on some common underlying concept but I doubt it does. There is only a pseudoconcept loosely holding together lots of different patterns of causal relation. I am fairly sure, for instance, that the way ideas are about external world patterns has nothing to do with the way that a dynamic disposition is about its manifestation. (One of the few places where I firmly disagree with John Heil.)

That said, I think there are several interesting causal relations worth exploring that can fall under the heading of aboutness. Where I think analysis of aboutness, as indeed for meaning, tends to go wrong is in the notion that it is some sort of relation, of a non-causal sort, between an idea or word and some object in the world. As indicated previously, I am not at all sure that we should consider objects in the world as things real enough to have any sort of relation. Moreover, I see in philosophy a tendency to assume that relations can be described for universals, when the only real relations will be for instantiations within systems that are so complex and individual that no analysis in terms of universals has any chance of holding up. Sometimes I think philosophers want to answer the sorts of questions scientists address without having to do any hard work!

In terms of the way ideas are about patterns in the outside world, I think we need to break the relation down into two, a relation between patterns of goings on in the world and patterns of goings on inside a brain and a relation between those goings on in the brain and the manifest sense of some external pattern.

The first of these will be a standard causal dynamic relation. However, it is hopelessly optimistic to think that this relation could be described in some simple formula such as instances of the word cow are caused by instances of cow objects. Like a modern personal computer, the brain is enormously complex and any attempt to give an account of the relations between external and internal dynamics involved in aboutness without reference to such complexity will not get very far. Some broad principles might be manageable, but a description that covered all anomalies and exceptions is unrealistic. We want a causal account but one that takes note of the immense combinatorial complexity of any causal history, which may include the entire linguistic history of an individual. Moreover, the aboutness relation involves a specific causal complexity that underlies what we call knowledge. I shall return to this. Suffice it to say here that it requires multiple causal paths organised in a very precise way. An inherent aspect of this multiplicity of paths is that we should not expect the relation between an object and a word to be an ‘A causes B’ relation. Moreover, the causal contributions of the large number of cows one encounters, not to mention a few mistaken horses, to any use of the word cow will vary in all sorts of ways. Otherwise, the relations involved, at this level, are pretty unmysterious.

The second aspect of aboutness, the relation between internal dynamics and a sense of an experience of something, is an immediate relation across the incommensurable complementarity of dynamics and manifestation. As such we cannot expect it to be ‘reducible’ to further microdynamics. As Newton pointed out, it is so far known to us just as a ‘brute’ relation that looks to have rules but nothing like the rules of dynamics themselves. Certain brain dynamics are associated with certain manifest experiences and that is all we have to go on. Even if these rules are tractable they will never take the form of ‘explanations’ and to that extent they provide a reason for considering aboutness mysterious, but not alien to naturalism.

Even if we accept the bruteness of this second relation and its compatibility with a naturalistic view there is, a further puzzle about aboutness. Why should a pattern of events in a brain seem to give us an experience of something outside in the world rather than an experience of goings on in a brain?

The answer to this question is not easily given but I suspect it has a lot to do with the relational nature of the world and qualities within it. As indicated earlier, all the things we might once have thought of as intrinsic qualities must be relational. To experience something is to relate to it. To have a rich experience is to relate to a pattern of relations. Our brains are set up to set up patterns of signals that will not resemble the outside world but will have some rule-based correlation with patterns of goings on in the outside world. This may involve some transform akin to the Fourier and Gabor transforms, and also reference to a specific set of subject centred co-ordinates. Nothing in these relations provides information about the qualities of the neural substrate in which they are instantiated as considered by an outside observer whose experience again reflects relations to relational patterns. And those qualities, as considered by some neutral God’s eye, simply do not exist because without being part of a relation they are nothing.

A further complication is that our experience includes a sense of ‘belonging’. It is easy to think that belonging is a dynamic feature of the world like boiling or falling but it has absolutely no dynamic meaning. It is a classic pseudodynamic concept. Our experiences have a sense of belonging built in through the way we chunk evidence for dynamic patterns into ‘objects’. The blueness belongs to a ‘cooker’. Aboutness is itself a version of belonging and to that extent is part of the spectrum of our manifest sense of things. It too is a pseudodynamic concept. The mistake is to think that experiences will be like what they are ‘of’ when experiences, as manifestations, are incommensurable with any dynamics that determine them, whether direct and internal or indirect and external.

Put differently, experience consists not of what the world is like but a pattern of manifest signs for patterns of dynamics. All experience consists of signs for goings on. Words act as signs for other non-verbal signs and to that extent their aboutness is different in an important respect, and I will explore some aspects of possible mechanisms shortly. Nevertheless, the mysterious part of semantic value, or aboutness is not between the word and the idea so much as between the idea and a pattern of goings on in the world.

Key point: we interpret the manifestations of experience not in terms of the proximal dynamics they are immediately associated with but in terms of inferred distal dynamics. This seems puzzling but there are reasons for thinking it could not be otherwise.

Need to explore the de dicto/ de re issues of something being ultimately about what it seems be about. A thought of a cow precipitated by visual stimuli from a horse are about a cow, not a horse, maybe. The referent of aboutness is generally an internal disposition in terms of idea rather than putative external object?

4.41 How meanings might be set up.

If meaning is never constant and the complexity of causal pathways in the brain so extensive, it is reasonable to ask whether it will ever be possible to give a causal neurobiological account of language. Nevertheless, I think one can expect to be able to construct and test certain general principles. A considerable amount of effort has gone into building models using networks of computational units in computers and with at least some success it would seem. The difficulty is in knowing whether such models really mimic the workings of a human language faculty or whether they simulate our linguistic behaviour, but using quite different computational means. I do not intend to review such models but, rather, want to focus on certain aspects of experience and language that relate the dynamic and manifest aspects of meaning.

In terms of general principles, I would suggest five of particular relevance to the complementarity-based view. This is not intended as a comprehensive list, but rather as a set of illustrations.

Firstly, I think there are strong reasons for thinking that our brains use selective rather than instructional mechanisms when it comes to using verbal and non-verbal signs for world goings on. Secondly, I think we need to postulate what I call a ‘mordant loop’ in which a pattern of signals can be ‘fixed’ for re-use and association by positive feedback mechanism. Thirdly, the brain is likely to make use both of both high and low specificity responses to local input patterns to allow categorisation both in terms of unique individuals and in terms of general or as yet unidentified types. Fourthly, words must be able to trigger ideas both directly through associations set up through mordant loops and through entraining dynamic subroutines that simulate external dynamics in space and time using general mathematical principles. Fifthly, there must be internal cross-connection between pathways that allows the brain to draw inferences about its own internal dynamics. Such cross connectivity is likely to be one of the factors necessary for the ability to use language.

Like Gerald Edelman, as a one-time immunologist, I am very aware of the advantages of a selective rather than an instructive theory of neural memory and its retrieval. An instructive theory would suggest that when we see an object we wish to remember our brain is ‘instructed’ both to construct some apparatus that will respond to this object again in a specific way and to link that apparatus to various triggering pathways. In a selective theory when we see the object our brain is already set up to respond to it in a specific way and all that is required is that the links are set up.

It might seem absurd to suggest, rather as Jerry Fodor has done, that our brains come pre-fitted with a carburettor-recognising apparatus, a Tickell’s leaf warbler-recognising apparatus and even a Dalek-recognising apparatus. Yet this is exactly how the immune system works. When we encounter a novel micro-organism we mount an immune response to it not because we custom build antibodies that will bind to it but because we already happen to have some antibodies that will bind, and we expand our store of these. We improve on these antibodies but not by customised modification. We just randomly shuffle the genes for the antibody protein about and then select the good variations for further expansion.

I strongly suspect that the brain uses a similar strategy, setting up a mass of alternative response pathways and reinforcing ones that prove useful, with others constantly being jiggled about so that the next time round an even more useful pathway has been set up. Such a system might seem wasteful. However, if a single recognition unit takes up no more than a billionth of the system mass and most of the system mass is involved in functions other than simple recognition then it need not be.

Of note, a selective approach like this puts a different cast on the role of experience seen as so dominant by Locke and other empiricists. Receipt of signals from the world is needed in order to develop associations with words or other signals, but the ideas entrained by experience are already in place, at least in the sense of pathways available for use. The pathways may be plastic and therefore open to diversification and further selection with repeated use, but in an important sense there is a contrast with the simple Lockean concept that all ideas come from outside.

An important implication of this sort of mechanism is that in a sense it appears to reverse causality. It does not actually reverse causality but it provides a way of ‘capturing’ a cause for a given effect. Since that would seem to be what we most need for understanding our world it would seem likely to prove a very useful tool if at came into existence. And of course this is how evolution works (hence Edelman’s term neural Darwinism).

Thus we have every reason to think it plausible that our brains are programmed to develop with machinery ready and waiting for generating any idea we are likely to find useful.

Key point: we do not yet have any detailed understanding of how neural connectivity relates to our use of language but certain general principles may be expected to apply to whatever mechanism is involved.

4.42 Mordant loops

In very simple terms, what the meaning of a word needs to involve is a disposition to set up certain dynamic patterns in a brain that overlap with those set up in the context of some sensory datum A, like seeing an apple, in such a way that these routines can be re-triggered both by the word and datum A and make available manifestations of a type associated with both word and A. There needs to be a rule-based system that will make use of both word and datum A in terms of both inputs and outputs. By analogy, in the immune system, we have a signaling mechanism that takes levels of both foreign proteins and self-made antibodies as inputs and has modifications of the levels of both of these as outputs. The dynamics of the two systems are very different in a number of ways but the immune mechanism provides insight into how a biological system can build and remember a repertoire of useful responses to outside stimuli by linking these stimuli to specific ‘identifiers’. The mechanism is complex and subtle, and has taken 50 years to unravel, but the key principles of operation are elegantly simple.

The key challenge for the brain in the setting up of meanings for words would seem to be the ‘marking’ of a certain pattern of sensory input in a way that allows linkage to another pattern of input that is the word. Marking in this sense will have a much wider relevance than just to language and so we should expect it to use a mechanism widespread in parts of the brain involved in learning. As an example, a penguin chick in a large colony rapidly marks the pattern that is the sound of the parent’s call, distinguishing it from a thousand other very similar calls.

The sort of mechanism that would achieve this can be described as follows (see figure). Inputs from sensory organs will be distributed through axon branches to arrive at each of a bank of cells, each of which responds to a different spectrum of patterns. The outputs from this bank of cells may be sent on to further banks of cells, responsive to a ‘higher level’ spectra of patterns. Thus, at some point the visual input from an orange will lead to responses from a bank of cells A1, A2, A3 … AN, with AX sensitive to evidence of a spherical shape (outline and shadows), AY sensitive to orange coloured patches and AZ responsive to patterns of brightness indicating a rough pitted surface. Outputs from each of the cells in bank A are then fed via branching axons to each of the cells of a bank B, again preprogrammed to respond individually to different patterns. Let cell BO be responsive to the combination of all the features of an orange. We then have a third bank of cells C that mediates the marking function. At any one time only a tiny proportion of cells in bank C are ‘active’, perhaps with each cell only active once in a lifetime. Outputs from B feed cells in C. Outputs from cells in C feed back on to cells in bank A. When BO fires a cell CO is induced to respond and send signals back to bank A. The synapses linking C to A cells have the property that if they are activated at a time when A is firing they are reinforced such that further signals from C will activate A in the absence of the pattern of inputs from the sensory organs that otherwise make A fire. If a C to A synapse is activated when the A cell is not responding to inputs from sensory sources the synapse loses its signaling power.

The end result of the mechanism so far described is that any stimulus that will induce CO to fire will lead to a ‘re-run’ of the signals from A to B that signify an orange. If we then add to this mechanism a similar mechanism that uses banks of cells A and B to mediate specific responses to the spoken word orange with a cell Bo in addition to BO feeding into CO and with feedback from CO to relevant cells in A as before for A and we assume that CO was activated at a time when the person was presented both with the fruit and the word then we have a mechanism that ensures ‘co-marking’ of the word and object. A variety of ‘filters’ would be needed in this system to ensure that object and word data were assigned a different status in some way but with more complex variations of the basic marking feedback mechanism this should not be difficult to achieve.

The implications of this sort of mechanism for linking words to patterns in the world are complex and will be returned to in section II when considering the subject of human experience. At this point the description given is intended simply as an indication of what it might be for a word to mean something, in terms of the dispositional properties of the instance of use of the word. Note that the dispositional properties of any particular verbal noise are completely arbitrary unless taken in the context of the dispositional properties of a human language faculty and the context of a language learning episode in which ‘object’ and word are experienced together in a way understood to be an instruction about how the word is to be used. This is very different from the sort of dispositions we associate with things like electrons or bricks but even those dispositions are entirely contingent on the dispositional partners under consideration so it is still reasonable to see words as packets of disposition in the same sense as electrons or species of dinosaur.

Key point: building a theoretical model for both memory and meaning seems to require some form of a feedback system that allows output to programme for later replication of input within association networks.

4.43 Variable specificity: from 50% to 99%.

Locke’s view that the meaning of a word is its ability to generate ideas in a person’s mind may have gone out of favour in philosophy. However, as indicated, my feeling is that it is right. Even if we do not want to consider ideas in terms of manifestations then his view remains valid if we consider ideas purely in terms of patterns of dynamics in human brains that determine patterns of behaviour, and in particular verbal behaviour.

The selective approach indicates that all sensory experiences we have of the world and all meanings of words are based on dynamic patterns that fall within a (vast) range of possible interactions between brain cells preset by the existing structure of the brain. True, that structure will change a little with every moment of usage, but essentially our ideas are selected from a range of patterns that the brain has on offer rather than imported in some way. A brain has ready and waiting an optional pattern of signaling that will be what results from viewing a carburetor and another option for the sound of the word carburetor. In a sense we are born with the concept of carburetor, but without it being marked up for association with an episode of viewing or the sound of a word.

It seems that all our brains do is make use of alternative signaling pathway patterns triggered by outside dynamic processes. A sensory input is in a sense a way of dividing the world’s going on into either this way or not this way. If you combine enough yes/no options of this sort you have a very detailed description of the world. All our experiences, or manifest ideas, are constructed out of patterns of yes or no answers. Yes and no have no colour, taste or sound in themselves but patterns of yes and no appear to be manifest to human subjects in terms of such things.

A question that arises in the context of our ability to recognise large numbers of objects and words for objects is how this ability is distributed amongst the individual cells of the brain. Is the ability to recognise an orange or a van Gogh painting allocated to one brain cell, or many? Are we born with one carburetor-recognising cell and how could that possibly have arisen through evolution?

A clue to this may come from the old radio panel game Twenty Questions. With a very few initially given clues (yes/no answers to three type options) in many cases the identity of a word can be derived via fewer than twenty yes/no questions. In the mechanism described above, you need fewer than twenty cells in bank A to have a unique code for every word. This is not surprising since twenty cells gives a million options and English has only tens of thousands of words in routine usage. The interesting question comes with the way you apportion cells in bank C.

Cells of the bank C type are used to establish connections – maybe memorable thoughts of association. How many such memorable thoughts do we have a day? Children at least probably have quite a lot. However it seems unlikely that we have more than a hundred a day on average. That asks for about a million memorable events in a lifetime. If cells of bank C are, for example, in the hippocampus, there will be enough to cover a million events at one cell per event. There seems likely to be some redundancy and we could deal with this if there are 100 million cells.

A discussion of how many cells mediate the association of a word with an object has to consider the relationship in terms of both specificity and sensitivity. It seems reasonable to assume that we allocate more than one cell to such a learnt association. The figures suggest that ten or a hundred cells might often be involved. However, we also want to know if each of these cells is exclusively involved in the association between one word and some single aspect of the world or whether bank C type cells can mediate several such associations. This might of course be the case for indirect associations. However, it is quite hard to see on anything like the simple model proposed how a bank C type cell could be reused for a second association without corrupting the first association.

The immune system may again provide a useful analogy. The cells that mediate ‘learnt’ immune responses use signals to differentiate both on a broad and a highly specific basis. The antibody response to a single foreign protein is not down to a single cell clone, but probably to a hundred or a thousand such clones. Early in the immune response a proportion of these cells are not highly specific and produce antibodies that can bind to several proteins. However, in a mature response the cells produce antibodies that show little cross reactivity with other proteins. A cell making antibody to an influenza protein is very unlikely to have any other significant functional role.

In neuroscience there has been a trend away from suggesting that individual cells have very specific tasks. However, I find it difficult to see how there would be an advantage in using cells that mediate links between words and objects for several different links, unless these were effectively elaborations on one central link. This more or less absolute specificity of cell function in language may be masked by having a significant number of ‘spares’ so that, for instance, loss of the ability to associate a word and a common object is uncommon in healthy early and mid life. Yet this does not mean that it does not apply. Moreover, loss of the ability to link certain words and objects does occur, and increasingly so in the elderly. It seems unlikely that such functions are distributed widely in an overlapping way between cells.

Another piece of evidence is the fact that in the elderly we do not see continual formation of new associations leading to displacement or confusion with old associations. Rather we see a progressive loss of ability to make new associations, with preservation of old ones.

The more we learn about biological systems the more we come to see that they are not like pristine tool sets on offer in a DIY store. They are more like the toolbag of a carpenter who over a period of fifty years has accumulated (or replaced where necessary) everything that is sometimes useful. Spanners come in different sizes and specificities (fixed size, adjustable and mole wrench). The immune system has sets of genes that seem to do similar things but turn out to vary usefully in specificity and domain of operation. The brain also shows a diversity of cell types and connection patterns that might look to be redundant, but are likely to contribute usefully by having different ranges of specificity and domain.

In this context, I think it highly likely that there are well defined neuronal subpopulations with response specificities that differ but overlap. Thus it might be that one population has a response specificity between 99:1 and 99.99:0.01. Another may range from 90:10 to 99:1, another may range around 50:50. Cells for names and episodes are likely to be at the 99:1 end. Cells involved in responses to words like ‘bigger’ will be in the 50:50 range. The loops each sort of word is set up to trigger are likely to be different.

The dispersion of these specificities is also relevant in that it can ensure that in the absence of a 99.99:0.01 specific response there will be some 99:1 specific responders. If less specific responses are slower there is a mechanism for responses to follow a ‘first shout wins’ rule which allows for parallel processing to find the best interpretation of a word (or visual stimulus) very rapidly, at the cost of occasional errors. As the passage of a first shout signal will render a pathway refractory to any immediately subsequent signal this mechanism overrides the fixed time frames of a Turing-type algorithm and replaces it with a ‘competitive’ dynamic in which most computational sequences simply die out. This could help explain why it is so difficult to describe syntactic rules in terms of a single linear logical algorithm.

Key point: rather than seeing all brain cells as ‘grandmother cells’ responding to one specific pattern of goings on, we should expect there to be ranges of response specificity for different functions.

4.44 Nouns and verbs

Any attempt to establish in detail the sorts of dynamic patterns that are entrained in brains by words of different syntactic class meets with formidable practical obstacles. This is presumably why Chomsky considered that his studies of syntax could not be linked directly to neural events with the present level of biological knowledge. As mentioned, there may be an extensive literature on computational models in the field of artificial intelligence but it is hard to know whether these truly reflect what is likely to happen in a human brain.

There is also an important literature on the functional anatomy of dynamic events entrained by words that includes data on syntactic classes. Thus Poeppel and Hickok and also Tyler have discussed the way in which nouns and verbs may be handled in different parts of the cerebral cortex. This work provides an important macroscopic framework that can tell us something about the structure of the pathways involved but it does not address the mathematics of the integrative events in individual neurons that form the basis of neural computations, and, more importantly for this text, it does not address the relation of words to ideas in terms of how dynamics generate manifestations.

I am not in a position to add any new information about how different types of words might entrain ideas to the above studies but I would like to explore briefly certain aspects of the potential differences in handling of word classes, such as nouns and verbs in the context of the complementary view I am proposing. An interesting proposal from Tyler is that although we might expect nouns and verbs to entrain very different sorts of dynamics there is evidence for the patterns being very similar when subjects are presented with single words. However, once those words are built into small phrases the patterns start to change. This makes sense to me in terms of how I would see dynamic subroutines contributing to the determination of manifest ideas.

Building a model for the way in which a word could signify an ‘object’ seems relatively straightforward. If an object is a package of operation of certain dynamic dispositions that tends to remain constant and give rise to a paradigmatic set of manifest ideas (such as a blue cooker) then all that seems to be needed is a tagging system that allows the brain to re-run dynamics that lead to a manifest idea of the object in response to the word. Note, however, that this is already tagging a dynamic pattern to signify an external dynamic pattern, rather than a ‘thing’, even if the manifest idea is of a sense of a thing.

Verbs, commonly being words for interactions or changes, look a bit more complicated in that they tend to relate to dispositions that are context dependent and the ideas we want them to engender will depend on the context given by other words. This distinction is far from hard and fast, however. As in ‘Mars is red’ there are verbs that indicate constant dispositions. There are also ‘light verbs’ such as take in ‘take a nap’ that do no real dynamic work. Thus, in a complementarity based view we would not expect a stark contrast between nouns and verbs in this sense, since we are considering both to be associated with patterns of dynamic disposition. Nevertheless, there are likely to be some important differences.

I suspect that verbs have three distinct dispositional properties that distinguish them from nouns. The first is to indicate what sort of increment of knowledge a statement is intended to generate. Just as in more conventional analysis a verb signals a predication that leads to a whole that can be truth evaluated, I would see it as a signal that other words are to be associated in an informative way.

In the case of the verb to be, the intended association may be essentially nothing more than that – co-tagging. In other cases we have to invoke something more sophisticated than just using tags to re-run ideas. In ‘Rosie fell in the hall.’ fell gives us a novel relation between the dynamic patterns we know as infant Rosie and those we know as the hall. The association is likely to be novel so there will be nothing to re-run. Rather we want to run some dynamic modeling programme that can put Rosie into one role and the hall into another and come out with some ideas based on abstract modeling principles. Essentially we are going to need some form of computation that models space and time and movement of objects. We may come out with the reassuring thought that since the hall has a thick carpet Rosie will probably not have bruised her face too much.

Thus, it seems that verbs will entrain computations of a more complex sort than nouns alone. However, the computations will depend as much on the nouns as the verbs. Where a single word will engage a mordant loop such that a pattern of signals is re-run and sent to a site for being manifest a clause will involve a subroutine in which the re-run patterns are sent off for one to operate on the other in some way, perhaps describable in terms of matrix algebra in which a pattern operates on another pattern to generate a third pattern of a similar form. This then is ready to return to the site of manifestation. This analysis is absurdly oversimplified, but I consider it important to try to build a minimum model of this sort if the issue of the nature of the human subject is to be addressed.

The third aspect of verbs is their tense and I suspect an important part of their disposition is to put associations in a context of episodes. Animals other than humans seem to have a well developed sense of time but maybe humans have a rather unique sense of episode. It seems likely that when we tag words to ideas of things we very often also tag them to episodes. There is evidence for this being not the case for very young children, which may explain why most of us cannot find any episodes in memory for the first 3-4 years of our lives.

Like name tagging, and perhaps even more so, episode tagging seems not to be explainable on some basis of ‘frequent usage’. An episode lasting no more than a second may live on with us in detailed memory for the rest of our lives. It is as if the machinery for capturing a pattern of signals with a mordant loop that allows us to re-run the pattern for ever after is sitting waiting all the time to kick into action instantaneously should something of great emotional impact supervene.

Key point: while speculative, it may be possible to suggest likely dynamic rules for different types of words that reflect their roles in the way statements inform their hearers.

4.45 How can words signify both dynamics and manifestations?

I have suggested that words, like electrons and dinosaurs, can be conceived in terms of dynamic goings on that determine manifestations. I have also suggested that we use words to mark both dynamic and manifest aspects of the world. This gives rise to something of a puzzle. If I consider the word hay in dynamic terms as some goings on that include a disposition to entrain thoughts relating to dried grass, in what sense are these dynamics different for thoughts about the dynamic pattern of dried grass and for thoughts about the manifestations we experience when we see dried grass, if and when they are.

The answer would appear to be that we need to consider the way in which dynamics and manifestations interrelate within our thoughts. I shall return to this in detail in section II but will just make some brief suggestions here. It seems likely that much, if not all, of the time our brains make use of manifest signals as evidence for the existence of dynamic patterns. Very often treating the manifestation as anything other than evidence for dynamics is irrelevant. However, an important element of human, and perhaps more generally vertebrate, intelligence is an ability to recognise the unreliability of our apparatus for inferring dynamics from manifestations. We are able to deduce that our manifestations are evidence for both internal and external dynamics and we use discrepancies that arise from the internal dynamics to infer things about them, just as we use discrepancies or changes that arise from external dynamics to infer those. Even as children, we come to think of a difference between how things seem to be and how things really are. Our brains allocate these different sorts of conceptual feel, but only when it seems necessary. Much of the time the distinction is not invoked.

Further analysis of this situation then hits the problem that our manifest ideas about dynamics are not in themselves those dynamics or even simulations thereof, but rather signs used to indicate those dynamics. Our brains process these signals in appropriate ways much of the time, as required by survival, but at the most fundamental level we have already seen that many of our ideas appear to be signs for dynamic patterns that cannot exist in the real world. The childhood view of stuff and things and doing works much of the time but if we want to know what is really going on we have to revise many of its most basic assumptions.

It seems likely that the human brain has evolved some sort of subroutine that allows us to hive off part of the world into a subdomain that we call ‘myself’. Other mammals and possibly birds may have something similar but not to the same extent. This hiving off allows us to make a distinction between inferred self dynamics and external dynamics. We also have general ideas about cause and effect. These ideas may be equally misleading but they do seem to work even in the context of scientific enquiry where getting the dynamics right is essential. What seems to be a problem, however, is that we do not have any agreed marriage between our ideas of cause and effect and our ideas about the self. It is as if we appreciate that there is a dynamic self subdomain but insist on exporting all manifest effects of causes out of that subdomain into the external world. This inability to marry these two aspect of our thinking is the mind brain problem. My contention here is that if we are prepared to carry through our ideas about cause and effect right to the heart of the self subdomain we will find that this problem is merely a logistic one, rather than a metaphysical one.

Key point: the dual meanings of words may be related to an important division in the function of our brains that will prove significant in later discussion – that of knowing the world and that of knowing our internal goings on.

4.46 Truth

Truth has puzzled philosophers in the twentieth century, but it is difficult to know quite why. Truth is a property of a statement, or a belief that can be expressed as a statement, that depends on the situation the statement is intended to relate to. Today is Monday is true on Mondays. This has led people to suggest that the truth of a statement is the relation of the statement to a situation. However, without a specified computational apparatus such as a brain that has rules for relating statements to situations there is no such relation of interest. Thus, the truth of a statement is entirely dependent on the rules of operation of the brain considering whether it is true.

This makes the truth of a statement a matter of the dynamics and manifestations it gives rise to in the brain for which it is deemed true. There is no ‘proposition’ that relates to the world directly. The brain for which the statement is deemed true will also support dynamics and manifestations in association with interactions with dynamic patterns in the world that our statement might be about. Thus the brain will have developed dynamic patterns as a result of learning that the word snow is used as a sign for a substance that is white, cold, crumbly and, on warming, wet. Whether or not the statement ‘snow is white’ is considered true by this brain will depend not only on the relation of the statement to patterns of brain dynamics but also the relation of white snow to brain dynamics but there is no direct relation between the statement and the snow that bypasses this.

What is it about the relation of true statements to brain dynamics that makes them true? One might suggest that they are true if the brain has encountered the associated dynamics before. But that is neither necessary nor sufficient; if encountered, the brain would have to have logged them as true through some relation that might be just as available to statements not previously encountered. We need a definition of truth for statements that will work for statements never constructed before but can still give rise to brain dynamics of the sort that manifest as a thought that the statement is true because they are entailed by some previously constructed statements.

It seems therefore that truth is not going to be a relation of a statement to some pre-existing statement or to the world directly. The alternative is that truth is an internal relation within a statement – between the words of a statement, as suggested by Hinzen. This is strongly suggested by other uses of the word true. A gunbarrel is said to be true if all sections of it line up in exactly the same direction. I think this analogy transcribes rather satisfactorily to statements. A statement is true if all the dynamic patterns set up in the brain for which it is true ‘match up’ in some way. Thus ‘snow is white’ is true because ‘snow’ will engender dynamics associated with, amongst other things, whiteness, and ‘is white’ will be taken to signify that white is one of the aspects that snow should be associated with. The two parts of the statement are ‘true’ to each other in the way that the opposite sides of a perfect square are true.

Importantly, what are true here are the internal dynamic patterns in a brain that are set in motion by the parts of the statement. Thus the truth of a statement does not exist outside the context of a brain responding to it. Some people will object to that and say that statements are inherently true or false. However, this seems to me to be another example of the childhood view that there is ‘a way that things really are’. Truth, like so many other aspects of our world, like blueness, is a dispositional aspect. A statement that is true has the disposition to set in motion dynamic patterns in brains that are associated with some sort of matching or appropriateness. That disposition is entirely dependent on there being brains that respond in a relevant way, just as the disposition of blueness requires brains with the tendency to infer a blueness disposition through collation of optical inputs.

This approach is in no way inconsistent with the intuition that there are truths about what is really going on in the world. Something really is going on and certain signs, in the form of statements, will be found to be true in the sense that they induce matching patterns for reasons that ‘track’ the patterns of dynamics they refer to in a robust way.

This tracking is a highly complex process that we can only infer in human brains because we do not understand how the computational network operates. It ought to involve some sort of comparison that results in a lack of discrepancy: in effect a 0 value. It also seems to involve a signing of that lack of discrepancy as certain: in probability terms, a 1. In relating to other truths (via And or Or), truths also appear to behave like 1, or ‘open path’ signals. The problem is that for a brain making use of massively parallel pathways with large numbers of inputs to each integrating unit we do not begin to understand how the result is achieved. Unsurprisingly therefore, much of our language is designed to try to cancel out the entailments of the computational detail. However, that does not mean that language can relate to the world other than through its disposition to engender dynamic routines in brains.

Within philosophy there is a concept of a ‘principle of bivalence’ which holds that a proposition about the world must be either true or false. If truth is seen as a relation of correspondence of a proposition to reality this might seem reasonable but if it has to do with matching of dynamic patterns within a brain it seems likely to be highly suspect on two grounds. Firstly, the truth of a statement will depend entirely on the dynamic predispositions of the brain in question. Secondly, there is no particular reason to think that the matching that we associate with truth is an all or none phenomenon. It is easy to think of statements that match quite well, but not perfectly. In fact the vast majority of statements are like this. ‘Horses have four legs.’ is generally but not always true. ‘The earth is round.’ is more or less, but not precisely true. The principle of bivalence seems to me to be yet another version of the idea that there is some way the world is that God will know for certain, even if we cannot be sure. If the fabric of the universe has a lot to do with one thing causing another and knowledge is a subtle form of that then it looks as if such a view is built on sand.

In this context, what I hope to achieve throughout this text, is not to provide absolute truths, but merely to provide ideas that I hope might seem true for the reader. If my words bring together non-verbal ideas in a way that seems to work, that is all I would wish for. As the Royal Society motto has it ‘Nullius in verba’ – take nobody’s word for it.

Key point: truth is not an aspect of the outside world but of internal goings on in our heads involving some sort of matching or consonance of compared ideas.

4.47 Summary

I would agree with Chomsky that at present we do not have the resources to describe how the rules of natural language grammars relate to specific neurobiological events. Nevertheless, I think there are reasons for pushing as hard as one can to build models that might give some idea of how words can act as patterns of dynamic disposition that entrain ideas in users brains in terms both of neurodynamics and experienced manifestations. The ‘semantic’ properties of words are just as ‘physical’ as any other dispositional characteristics. They only appear mysterious because the dynamics that are entrained depend not only on the enormous complexities of the insides of brains but in the constant changing of the responses those complexities entail.

It seems reasonable to suggest that human language is possible because the human brain has the resources to add a ‘meta’ step to a general procedure in thought in which manifest patterns that form signs of dynamic patterns can be tagged and re-used. What may be particularly important to this is temporal tagging in terms of ‘episode’ and an ability to re-run extended acquired sequences.

The converse of this is the suggestion that the elements of all experience, as signs, operate much more like words than the pixels of a display screen or the constituent oscillations of a complex chord. If I see five red roses I think it is much more likely that the signals that immediately give rise to this experience include signs for rose shape, signs for there being five and signs for attributing the same red colour to all five rather than the sign for red being repeated not just five times but for every pixel needed to ‘fill in’ a rose shape. This linguistic economy of integration of signs was suggested by Marr when considering how mental images are finally constructed from cruder ‘stickman’ components. A similar conclusion comes from Barlow when considering the numbers of elements that might be available to encode a visual scene.

This sort of mathematical format again provides a way of unifying what might be happening when we sense that a plate on a slanting table is just about to fall to the floor or we sense the outcome of the same events described in words. A matrix-like mathematical operation in which elements of one pattern operate repeatedly on elements of another pattern seems roughly to fit the bill. This might help explain how something as apparently complex as language could arise in a short space of evolutionary time from an ancestry with spatiomotor skills but apparently no real ability to use grammar. If all that is needed is a meta step, perhaps based on reduplication or branching of a connection path, allowing matrix-like elements to operate with an extra degree of freedom or one more time then the transition begins to look plausible.

An internal causal dynamic approach to language also emphasizes the fact that it is all about increments. It is not about eternal meanings in the form of propositions at all. Truth and falsehood are internal dynamic issues that can sometimes be seen to match up across many uses of a statement but need not. Sentences as syntactic structures are primarily about individual increments of knowledge, not about truth. Increments involve causal steps with a constant interplay between the dynamic and the manifest. Unsurprisingly, therefore, we find words switching from dynamic to manifest roles in different incremental contexts.

The fact that manifest ideas in our brains are not about the proximal events that underlie them but about inferred distal events in the worlds has lead to claims of the existence of some mysterious non-causal relation of intentionality or aboutness. Aboutness certainly makes every point of the analysis more complex. It also leaves plenty of room for the formation of pseudoconcepts, that need to be unpicked before we can get a clear view of things. However, as I will come to in the section on knowledge I think there are reasons for thinking that the sort of aboutness we encounter in meaning is intrinsic to any sort of knowledge acquisition. Without it we may have a kaleidoscopic world to enjoy, but one totally devoid of meaning.

Equally, I return to the fact that the manifest aspect of meaning is ineliminable and in a sense definitive. Meaning is ultimately something sensed, something manifest. We can consider meaning to another person in terms of both internal and external dynamics but to do that we, ourselves, must make use of sensed concepts. Ironically, manifest meaning can only be described in terms of the dynamics that may go towards determining it, but that does not mean that we can avoid postulating that somewhere within the human brain something is manifest to something.

5. Knowledge

5.48 What is knowledge?

From early infancy we appear to have a sense of what is meant by ‘to know’. A child only able to produce a few words is expected to respond to a family photo and the question ‘Do you know who that is?’ with a confident answer ‘Mummy’. Yet what sorts of goings on we should consider to pass muster as instances of knowledge remains very unclear. As indicated in the next section, it is recognised that we have rather different requirements in different contexts. Nevertheless, I think there may be certain central features to any patterns of goings on that count as knowing that, if explored in some detail, may give us a better idea of what our sense of knowing might be about.

The simplest view of knowledge is that all we have to do is open our eyes and the world presents us with ‘information’ about how it is. We can know the world just by allowing information to pour in through our senses. That is how it feels. Although we become aware quite early on that we see by receiving light and hear by receiving sound, that probably does not for most people alter the assumption that somehow red light brings with it the tomato’s quality of redness and buzzy noises bring with them the buzzy quality of a bee’s wings.

At the other end of the scale there is the philosophical view that to know that so–and-so is so requires sufficient competency in language to be able to formulate beliefs about it in propositions. So according to Robert Brandom, parrots cannot know what words mean even if they use them correctly, because they cannot construct propositions.

The former view is the paradigm of so-called non-inferential, and the latter of inferential, knowledge. As indicated earlier, I think even ‘raw’ sensation is likely to involve inference of a sort, but I am also sceptical about language being essential. (I am more attracted by the idea that some human knowledge is of a rather special form that is made possible by some of the features of our brains that also make use of language possible.) There are probably several levels of complexity amongst patterns of goings on that could qualify as knowledge but I suspect that we can identify a minimum structure relating to this central inferential aspect and that minimum looks far from simple.

From the outset I have emphasised the fact that the qualitative features of our manifest experiences are not qualitative features of the outside world but signs our brains use for dynamic patterns. Although to begin with colour information is carried by light rather than sound, once the information has been converted into nerve impulses, there is absolutely nothing about one activated nerve fibre that makes it carry a colour message rather than a sound message. Knowing is not just the world telling us what it is like through ‘quality-carrying signals’. We have to work at it. What we can know depends on the scope and sensitivity of our sensory apparatus.

More importantly, for the issue of knowledge, in order for sensations to have meaning in the sense of being about some specific dynamic pattern in the world and thereby give us knowledge of that pattern, multiple modes of acquisition must lie in certain specific relations to each other.

If I look out of the window a pattern of signals impinges on my visual system. That pattern of signals leads to signs of dynamic dispositions being manifest in my experience. However, this is not a one to one relation. The same pattern gives me many different pieces of knowledge. It tells me that the sisyrinchiums are in flower, that the sun is shining on them, and that my glasses are now clean of the thumbprint I have just wiped off them. An incoming signal is not about any particular goings on in the world. It can be used to know about many things, and it can only be used for knowledge through its relation to other signals, past or present.

In an essay called ‘Misrepresentation’ Fred Dretske makes the simple proposal that to have knowledge about anything we have to receive signals from it through two different routes. Dretske uses this idea to build a theory of how one might ‘naturalise’ aboutness (or intentionality). I share that goal but beyond the simple two-path insight I think Dretske gets bogged down because he thinks aboutness needs to be related to behavioural output and as a result gets caught up in a philosophical debate that I see as fruitless. Nevertheless, the two-path idea seems a good place to start from, even if for reasons rather different from Dretske’s.

If, as I have suggested, all we can know about are instances of operation of patterns of causal dynamics, then knowing has to rely on comparisons. Moreover, the more one considers it the more different sorts of comparisons seem to be needed before we begin to get anything that resembles what our sense of knowing seems to require. We are likely to need multiple alternative routes, and the routes are likely to be alternative in all sorts of ways in time and space.

Moreover, this reliance on comparison has as its complementary implication that we do not get any knowledge of the immediate signals themselves. We have no direct knowledge or sense of photons reaching our retinae or neurotransmitters passing from cell to cell in the brain. Knowledge is always of something other than the signals that carry it. It is in a narrative of signs about something distant from the process that mediates the knowing – a pattern of causal relations. The unexpected indirectness of aboutness should not be so unexpected after all.

[image reference is broken]

But there is a further paradox hidden in Dretske’s little diagram. We seem to need multiple routes for signals to provide knowledge. It would seem that these routes need to converge in order for a sign of some going on the world to be generated. However, there is only any point in generating a sign if something receives it and whatever receives it, and is thereby informed of the goings on, would seem to have to rely on one route for receiving the final sign, rather than many. In a sense it seems that the only thing in a position to know cannot, in itself, know.

As indicated at the outset, I think we need to break down knowing into two parts. One aspect of knowing is a property of a knowing machine that collates and infers from multiple paths, or even, on a wider basis, a knowledge system that includes all the paths in the outside world involved. Another aspect of knowing is the manifestation of a narrative based on what has been inferred about the world to some knowing subject. Many would say that these two aspects should not be pulled apart. However, if we are to try to build a dynamic account of what is really going on in an instance of knowledge, as diagrams like Dretske’s seem to, I see no alternative. It is no good suddenly saying ‘oh it is a property of a system’: not, at least if we want an account that generates a manifest sense of knowing.

It may be that we should accept that there is actually no such thing as knowing about things and try and find some other form of words. In time that may prove sensible but for the time being it would seem more productive to try and unpick what knowing about things might mean in terms of dynamics and manifestations and see if one can come to an account that will work if laid out in the appropriate carefully chosen linguistic context.

Key point: the dynamics of knowing are more complex than our intuition tends to suggest. Multiple relations and pathways are involved.

5.49 What knowledge has been said to be

Before trying to produce an account of knowledge based on dynamics and manifestations it may be useful to review some issues raised in the philosophy literature that help to clarify context. A multiplicity of meanings for the verb to know is well recognised in the literature. One can distinguish knowing that Henry VIII had six wives from knowing how to ride a bicycle, from knowing John, from knowing that his name is John or from knowing how a diesel engine works. A single definition of what it is to know is obviously not to be expected.

One can debate at length how many subtypes of knowing there are, but the main distinction is probably the one that divides off knowing how to ride a bicycle. The acquisition of coordinated motor skills is not really of concern to me here since it is only indirectly related to knowing about goings on in the world – the operating of patterns of dynamic dispositions. Moreover, perhaps in part because it involves developing reflex responses to kinaesthetic data we are barely aware of, it is peculiarly invisible in terms of ‘knowing what it is we know’. Dretske has argued that ‘knowing how to’ … must involve some ‘knowing that’ … and this is no doubt true. However, there is at least a core of another sort of concept in knowing how to ride a bike that falls outside what I want to discuss.

The other examples are all in one way or another knowledge about what is going on (or has gone on, or will …), i.e. about the operation of patterns of dynamic disposition. Linguistic knowledge is about patterns of human behaviour but is otherwise not very different from the others.

Certain types of knowledge have been described as tacit. Thus, when I developed a tear in a knee cartilage I started to limp, apparently in the knowledge that this would be less painful. Yet, finding myself limping, I wondered why I did it when, on deliberately taking several normal strides I had no pain. Not only did I not seem to know that normal strides would hurt, but they seemed not to. It was when, after about a dozen strides, my body weight shifted in a slightly different way on the thirteenth and resulted in a severe twinge, that I realised that I must have learnt of this tendency unconsciously and modified my gait accordingly. Similarly, it was not until I started to learn other languages that I realised that I knew that a rise in pitch at the end of an English phrase meant that a statement was a question.

Tacit knowledge seems to bypass the need for any sense of knowing or any need for the knowledge to be manifest to an experiencing knower. It might seem to lend support to the idea that there is no need to invoke such an internal knower, merely a knowing system. However, there still has to be some point in the system which receives signals that signify the inference of dispositions about the world, otherwise there would be no point in having a system that made such inferences. That point in the system may not be an experiencing human subject in the sense that we are familiar with but it remains true that knowing only means something if it leads to input of such a signal into domain where behaviour is determined either directly or indirectly.

In this sense knowledge has parallels with belief, which may be tacit or overt. My limp might have been considered to be based on a belief rather than knowledge that a normal stride would risk pain. During recovery a limp often persists after this risk has gone, in which case false belief would be an appropriate description. Moreover, both belief and knowledge can be used to imply either a disposition to have certain ideas in certain circumstances and the operation of that disposition in such a circumstance with the generation of a manifest idea.

In this context it is of note that for a long time knowledge was considered to be definable as justified true belief. This was called into question by Gettier, with cases of belief that were true and seemed to have justification but could not be considered knowledge because their truth was for reasons other than the supposed justifications. To my mind the difficulty here is the meaning of the word justification. If justification is taken to be unimpeachable justification then it can only apply to true beliefs and true is redundant in the definition. If justification merely means that the believer has some reason behind their belief then it seems we would have no way of excluding very unreliable reasons, like reading it in a notoriously unreliable newspaper. In fact one popular way out of this debate is the idea that justification has to ‘track’ reality reliably, which, like Dretske’s and my own approaches, means starting to address what is actually going on in a case of knowledge.

We come back to the fact that all knowledge involves the inference of operation of dynamic dispositions of the sort that can generate manifestations and the more unimpeachable the basis for inference the more we are likely to call it knowledge. The remaining issue, relating to truth, is when we can consider inferences unimpeachable. The answer seems clear – we are never in a position to consider inferences unimpeachable. There is no ultimate fact of the matter about the correspondence of two elements of a proposition that might make it an absolute truth. Knowledge involves comparisons just as truth does. There may be no doubt that what is really going on in the world is really going on but as soon as one tries to generate ideas, verbal or otherwise, we are the domain of comparison within a knowing or truth-evaluating system that will be subject to a multitude of sources of uncertainty that can only be minimised, not avoided.

What I think this does suggest is that absolute or unqualified knowledge is a concept too feeble to be worth having. A much more interesting and robust concept would seem to be of knowledge as just an inference about the world drawn in a reasonably appropriate way such that at least under some conditions it is likely to lead to ideas or predictions that match up with further observations.

Key point: many previous attempts to define knowledge have been tied to rigid philosophical concepts without clear reference to causal dynamics. We need an account based more firmly in dynamics with the flexibility to deal with real situations.

5.50 The need for multiple pathways

If we take our mental states to be about ‘physical objects’, conceived in some way that has no already implicit relationship to experience then aboutness can seem pretty puzzling. Thus we have Fodor famously saying that if intentionality is real it must really be something else. If we see patterns of dynamic goings on and the manifestations these goings on determine as two inseparable complementary aspects to everything we can know of, I think things become less problematic. It may mean that we have to think of things in terms of what appear artificial and unwieldy categories, certainly in relation to everyday use, but this is a small price to pay for a coherent overview.

If a pattern of signals s1 arrives at sense receptors from the world the brain is in a position to generate a pattern of internal signs p1 that can stand for that subset of all possible dynamic dispositional patterns in the world d1 that have the disposition to entrain p1. If there is only one path involved, that is to say one place and one time of arrival of s1 at the sensory system, that is the end of the story. In Dretske’s terms p1 cannot be said to misrepresent d1, because it is defined as doing so.

Let us also assume the same arguments for an s2, p2 and d2. If we now consider the subset of dynamic dispositional patterns in the world d3 that include the disposition to present signal patterns s1 under some circumstances and s2 under others the brain is in a position, if it has the computational power, to generate an internal pattern of signs p3 for d3. Receiving both s1 and s2 in the appropriate relations allows the brain to infer an instance of d3. Note that inferring d3 is not in conflict with inferring d1, since it is a subset of d1. This is the beginning of the sort of situation that obtains when the optics of our eyes are working well. We can move from photoreceptor activation meaning that light is around to meaning that the light is radiating out (whether emitted or reflected) from some matter ten feet away. I say the beginning because it is fairly obvious that the number of optical and neurological pathways involved in generating signs from optical signals is very large. Moreover, we can expect the effects of multiplicity of paths to be in some respects combinatorial. Thus we tend to be familiar with a ‘subset of dynamic patterns disposed to entrain a pattern of signs p’ as something more like a specific event at a point in time and space.

I have started with the idea of a pattern of signals from the world and it might be thought that I should start with ‘one signal’. My reason is that I am uncertain that there can ever be a situation in which something receives just one signal rather than a pattern. I do not want to make too much of this in this context, however. Once the argument gets started we can start considering elements of signal patterns that can take more than one form much in the way that binary information theory considers ‘bits’. For every extra element the set of patterns p that can be assigned an internal sign d can be divided up as many fold as there are significantly different ways that element can be. Every extra element of pattern that can take more than one form effectively provides another path from the ‘to be known’ to the knowing machine.

If we then consider how we might have ‘misallocation’ of a sign d to a world pattern p we have a mechanism but it needs a bit more computational power. Let us say we have enough paths, or variable elements, to be in a position to allocate internal signs d1,d3,d4,d5, to subsets of dynamic patterns in the world p1,p3,p4,p5. Each member of the p sequence is a subset of its predecessor. If we also have built in mechanisms that bias the way internal signs are generated on the basis of what might be dangerous, or desirable, or most likely in the context or whatever, then it may be that arrival of signals s3 will lead to the generation of the sign s5. We do this listening to speech or reading words all the time. With standard speech a significant part of the sound patterns that ought to be needed for each word is often missing. We also do it when looking at things. You can look for ages at a nighthawk sitting on a branch ten feet away and just see a branch. Then suddenly your visual system decides to switch signs and you see a nighthawk with all its detailed camouflage pattern.

It might be thought that if our brains work this way the internal signs manifest in experience would just be arbitrary symbols that would give us no real knowledge of what the outside world is really like. The first response to this I would give is that that is must be exactly so, in one sense and that it does not raise a problem because the world is not really like anything except a pattern of dispositions to entrain these manifest signs. Moreover, language definitely uses arbitrary signs and it is sometimes quite remarkable what complex discussions can be had by people using word signs for dynamic patterns they actually have no idea of. However, in another sense the internal signs brains use for dynamic patterns will not be arbitrary to the extent that they will be patterns with internal structure that reflects the way the dynamic pattern subset relates to other subsets.

When I think of a horse chestnut tree the signs manifest to ‘me’ as a sense of a horse chestnut tree include several different types of element. There is the inner speech sound of the near arbitrary ‘horse chestnut’. There is a sense of this being an idea with certain parameters that link to the a concept of a species as a kind with many individual instances, obtaining in the present and in the vicinity. There are also elements of ‘thin’ versions of visual and tactile signs – the delicate colour pattern of the inside of the flowers or the damp smoothness of a freshly harvested conker. These aspects of the sign pattern appear to be non-arbitrary in that they seem to do service in similar ways as components of other ideas about other patterns in the world.

What I think this non-arbitrary component indicates is that when we talk of knowing about something or having a concept of something we tend to assume that we have the capacity to divide the subset of dynamic patterns, say p3, assigned sign d3, into many further subsets (individuals, viewpoints, instances over time) using signs that relate in a non-arbitrary way. This means that by and large we do not think about dynamic subsets that we cannot further subdivide if we wish to. An example of an exception might be the weakly interacting massive particle or WIMP. If asked what we know about WIMPs we might get the response that they are WIMPs, period.

I think this ability to subdivide and cross-reference indicates something further hidden in the concept of aboutness that philosophical discussion tends not to deal with. If I have an idea of something then that idea might seem to be about that something. However, to know about that something we tend to think that we must have an idea that we can relate to other ideas through common features.

The message that I get from this sort of analysis is that what we tend to think of as ordinary examples of knowledge, and even the sort that we might ascribe to ants or frogs probably requires not just two pathways for signals to arrive by, but a minimum of a significantly greater number of pathways, each needed for a rather different reason. I do not intend to try to formulate a precise account of this because I do not think it would be particularly helpful to do so but I would like to spend the next section reviewing what some of the diverse functions for multiple paths might be.

Key point: an essential feature of knowledge of what is going on in the world is that it involves comparisons and therefore multiple pathways.

5.51 Roles for path complexity in knowledge.

It would be nice if one could put together a set of rigorous arguments of the sort found in textbooks on computing that would allow a formal definition of knowledge. Chomsky’s account of types of automata, from the simple two state automaton to the Turing machine, is along these lines. I suspect that a minimal form of knowledge needs an automaton with a push down memory stack, if not the complexity of a Turing machine. However, I have two reservations about this. Firstly, there is no point in defining knowledge in terms of artificial computing systems that may lack several crucial aspects of the human knowing machine (not to mention knowing subject). Secondly, I am sceptical of the idea that there is some magic threshold that allows some goings on to form knowledge. I think it much more likely that knowing is a graded process, a bit like playing the piano, ranging from beginner to the capabilities of the mature Alfred Brendl. As for truth, I think we should move away from the black and white view favoured in philosophy.

Instead, I will itemise what seem to be different reasons for having multiple paths in knowing. Some may turn out to be equivalent in computational terms but if they have a different practical significance that probably means they reflect additional complexity in the real processes of knowing going on in our brains.

1. A single signal arriving at our sensory apparatus can only indicate the total subset of dynamic patterns that entrain the immediately proximal event of signal arrival. It tells us nothing specific about the causal history. Multiple signals from the same source allow us to whittle down the subset of dynamic patterns to those entraining more causally distal events, with A simple example is binocular vision, which allows us to infer that a dynamic pattern is ‘at least that distal’.

2. Multiplicity of signals may also allow us to distinguish specific distal dynamic patterns from yet more distal causes. Thus signals about average intensity and homogeneity of lighting conditions allow us to attribute a colour spectrum to the reflectance properties of a tomato, rather than the pattern of illumination.

3. Constant dynamic patterns like unchanging reflectance can entrain ideas of ‘objects’ with the pathways above but many dynamic patterns reveal themselves as changes in incoming signals. Thus knowledge of movement requires sequential inputs – at least one for before and one for after.

4. In 3 the multiple paths may be repeated use of the same path structure, at least within the primary sense organ. There are, however, further implications for multiple paths when it comes to integrating the sequence of signals during the process of inferring a change.

5. Still considering vision, at least a minimum multiplicity of paths is required for generating relational information in each of two dimensions normal to the line of sight. Since all dynamics are relational it is doubtful whether we can be said to know about them unless we have information about their relations to a spatial frame.

6. Similarly, to know inconstant dynamics in any quantitative sense, as in slow and fast, the extra pathways in 4 will need to include paths that allow comparison with some form of internal clock.

7. The roles listed so far only really provide for knowing current dynamics. Human knowledge, even in terms of simple observations, like ‘that gust of wind is bending the old birch tree’, is couched in terms that imply relation to past, future and specific episode. Our knowledge is tensed. Dynamics are known in at least present, past, imperfect and future terms (not to mention past perfect, future perfect …). We need paths for sequential signals to be stacked up and integrated in a way that allows us to place dynamics in a detailed history.

8. It is doubtful whether much of our sensory input begins to generate knowledge in everyday terms until it has been collated with input from other paths in other sensory systems. We are quite unaware of the fact that vision always works in the context of complex parallel inputs from position sense and in particular position sense as informed by sensory organs in oculomotor muscles. Our eyes move in regular saccades and the input to the retina during these saccades is simply blocked from entry into the sensorium in response to signals indicating muscle shortening. Our ideas of dynamic patterns often fall much less distinctly in the domain of one sense than we think, especially spatial and textural ones. Moreover, very often to know about something is almost synonymous with being able to translate evidence from one sense (like vision) into another (like touch or temperature sensation).

9. In addition to all the above it seems that human knowledge involves pathways that, like those involved in tense and sense of episode, provide a sense of type and of individual (or token). The old birch tree is an individual, as maybe is the gust.

10. Finally, and quite distinct from all other roles, there need to be paths with a sufficient number of options for signal patterns to be able to present a sufficiently wide range of signs manifest in experience to whatever subject experiences them to cover all possible experiences.

The complexity of this list may give some insight into why it is not surprising that there are problems with defining it as justified true belief. Knowledge is a complex, variable messy business. It might often turn out to fulfill the JTB description but is it not more useful to try to pin down what is actually going on? The brain is likely to work on a selective principle such that it is set up with a vast banks of pathways, each with a slightly different responsiveness for each pattern of input, thereby allowing it to divide the world up into fine subsets of dynamic patterns. The diversity of response is likely to be generated at least partly randomly and the reinforcement of pathways with ‘tagging’ to form concepts of kind or individual and to link to words is likely to follow a ‘what is the most salient response at the moment’ form of computation. That seems quite enough to allow us to expect mistakes of inference and to be able to have thoughts about unicorns. Our brains develop with the ability to generate manifest signs in response to anything that might come along, including a unicorn just as our immune systems have already made antibodies (in small amounts) to every protein that a scientist might synthesise for the first time next week. If I pass a cow on a dark night I may have a thought entrained by this sighting which makes use of my brain’s signs for a horse. As to whether my thought is about a cow or about a horse, I think that is just an issue of the laxity of the word about and the fact that aboutness, if not a pseudoconcept is certainly an ill-formulated one for most people.

Key point: even the most basic forms of knowledge is likely to need multiplicity of pathways in several ways and our most sophisticated forms of knowledge even more so.

5.52 Differences between us and animals

Human knowledge has, historically, often been considered qualitatively different from that of other animals. I am sure that there are some essentially qualitative differences in the pathway connections in human brains and the roles they play in knowing. However, I am resistant to the idea that in some way animals do not quite know and we have it taped.

Even for us knowledge must be an asymptotic process of further and further limiting the subsets of instance of dynamic dispositions that we use signs for in order to make predictions, but never defining the relations of those dispositions absolutely. The limiting process is itself limited by the number of paths, which is not infinite. More importantly, if dynamic dispositions are ultimately defined in terms of the manifestations they determine, since we cannot consider, or maybe even conceive of, all the potential contexts for such manifestations we will never have absolute knowledge of anything. To think that we might is to return to the childhood view of there being something that the world is really like, that God might know about.

I think many mammals may have forms of knowing closer to ours than is often allowed for. Thus, I do not accept that mammals only live in the present, although I think they may lack a clear sense of individual episodes in the past. I am also unclear whether we should deny animals a knowledge of their own mortality.

What would nice to know is in what way the range of roles multiplicities of paths perform differs in animals and man. Quantity is not so much the issue. Birds may see more colours than we do, dogs may here more noises and smell more smells. Jays may remember the quality of nuts hidden in 300 places better than we could. However, we can concede these feats without conceding our sort of knowledge.

One provocative suggestion is that animals tend to know a lot more than we do, at least to start with. Birds have remarkable genetic programmes for building nests, singing and migrating. Most very young mammals are better at knowing what to eat and what to run away from than human infants. This has lead to the suggestion that some of the sophisticated forms of human activity, including language, are made possible by a relative paucity of genetically programmed knowledge of this sort. Rather than having an instruction sheet our brains may be left to develop on a ‘see what works’ basis. If, for instance, one of our newly acquired skills is to sort dynamic patterns by instances of episode, then it would be counterproductive to be programmed in terms of when the episodes were and how many there were. It would be much more useful to be able to tag episodes when and where they are associated with learning something new about what might work. On the other hand if the tendency to soft wiring and see what works is very general we might expect some time to be needed to get signs for dynamic patterns collated in a workable way before starting to give them episode tags – which would seem to fit with the absence of stable episodic memory much under the age of three. The infant may be polishing up vast numbers of collated pathways serving the more basic roles so that episodes can be tagged to very well defined dynamic patterns.

Put another way if humans are better at cross collating at a higher level and that depends on selection from a large stochastic repertoire the cost of being able to select cross collations in more ways may be to have a more stochastic development programme at all levels.

Much of the difficulty in attributing knowledge to animals revolves around whether o not we think that one has to know that one knows. This is related to the question of whether an inference has to be associated with the thought (whether or not verbal) that ‘I inferred that’. My suspicion is that most of the interesting aspects of human knowing occur at a subliminal level of inference. But that is not to say that the knowledge is not associated with a sense of something being so and very likely a sense of confidence in it being so, although it is likely to lack any sense of a ‘theory’ of why it is known to be so. I see no reason why animals should not have this sort of sense.

A number of people have suggested that there is an important distinction between mental states that simply involve percepts and those that involve concepts. Conceptual thoughts, sometimes equated with propositional thoughts may be considered only possible for those with a language in which to formulate them. It is trivially true that thoughts formulated as propositions using the syntax of language require competence in that language syntax. However, I can see no reason to think that one needs language to think about what these verbal propositions are about. In other words I do not think we need to propose the existence of propositions beyond linguistic structures.

Philosophers have spent a lot of time on what they call propositional attitudes. These are thought to be attitudes to propositions such as ‘I believe that trolleybuses used to run in Mill Hill’. In my view all this requires is that I should believe in the instantiation of a particular dynamic pattern, of trolleybus running several times in the distant past but not since. Thus if I were to be shown an iPad file of pictures taken by a five year old on their new camera, including ones of the family cat, mum and dad, a flat screen television and a 2012 Olympic mascot, followed by a picture of a trolleybus in Mill Hill, I would laugh and shake my head to indicate my disbelief that this picture belonged in the file. I see no need for language here, nor the suggestion that my reaction reflected a belief in a proposition.

In general I am sceptical about the idea that humans can know complicated things because we have can use language. I think there is likely to be a close link between our ability to use language and our ability to think complicated things but because of some common cause in a genetic change in the programme for neuronal wiring. There is no doubt that writing things down for both oneself and others to read later helps in the building of complex bodies of knowledge like biology or physics. Without language we might know rather little history. However, I doubt that someone coming to know something new has much to do with words. We do not write down premises and then extract conclusions just using language structure, in the way we do for mathematical proofs, and then believe those conclusions. We know that language is so full of uncertainties that we must rely and converting the premises into non-verbal thoughts about the world and then, non-verbally, assessing the implications for thoughts the conclusion raises about the world. Words are, after all, just arbitrary noises and marks that we come to associate with non-verbal thoughts.

I am aware, however, that some people genuinely believe that they always think in words. When we come to consider the nature of the human subject this raises a rather interesting question. I see no reason why it should not be the case that the relative extent to which parts of the brain involved in the sorts of higher level processing that feeds our reporting of experience are involved in verbal or non-verbal types of sign may vary between individuals. Congenitally blind people’s thoughts are non-visual, deaf people’s are non-aural. Maybe the dominance of spatial signs in the ideas I report is just not the case for some people. Nevertheless, that possibility does not alter my suspicion that such variations would build on some common feature of human neural connectivity that allowed sophisticated inferences to be made whether with the help of words or shapes or whatever.

I also have a suspicion that verbal reasoning is not likely to be our best reasoning, or at least is likely to be particularly susceptible to error. If the brain generates signs for dynamic patterns in the world that are based on a framework of pre-established interrelated signs that have been cross-collated through visual and proprioceptive interaction we can expect to have confidence in their reliability. If, on the other hand, we acquire verbal signs that we assign dynamic significance simply by the way they have appeared in conjunction with other words then we are likely to have poor quality knowledge. As Wittgenstein pointed out, just when you thought language was solving a really difficult problem you find it has ‘gone on holiday’. This happens on a day to day basis throughout biomedical science, not to mention economics, politics and of course religion, where words like God with meanings that nobody can possibly know, are in every sentence!

None of this is to deny that language may not have played a large part in the biological success of human kind but perhaps more in the way that Henry Ford made cars available for everyone than the way that Benz or Daimler invented the idea.

Key point: human knowledge undoubtedly has aspects not found in animals. These probably relate to different multiplicities of pathways, which may arise from relatively minor changes in genetic programme. These differences may well be relevant to language skills but that need not imply that language is needed for human types of knowledge.

5.53 Knowing machines and knowers

Most of the multiplicities of pathways mentioned above in relation knowing have roles in generating signs that could be specific for a particular subset of dynamic patterns. To generate signs that give a sense of a tomato we need lots of comparisons of incoming signals and signals from memory-based pathways to narrow down the subset to this one. To know it to be a tomato involves for most people a disposition to have thoughts of inside appearance and taste in response to relevant prompts and to recognise the word tomato as its name. To know that it is Mary’s tomato involves yet more collations.

Irrespective of how the above collations and inferences are carried out they only achieve anything if they can generate signals with a sufficient number of alternative forms to allow the level of specificity required. Thus if we want to divide dynamic patterns up into 100 subsets and specify one, then the systems must have an output that has 100 different options.

This multiplicity of paths, or at least signal options, for the result of the collation process applies just as much to familiar machines as to people. However, there is a crucial difference. For something like a digital camera, all that is needed for the collation process to do its job is for there to be some way in which the specificity of the pattern recorded can be made available to the world. There is no requirement that a pattern of signals fully specifying the dynamics to be recorded be available to anything in the camera. Such a camera is a complete knowing device but with no knower. The two aspects of the idea of knowing come apart.

I think it is worth considering the digital camera in some detail because it provides useful analogies for both the similarities and differences between ‘machine knowing’ and our own. The collation and ‘inference’ procedures modern cameras use are quite similar to those we use. The lens allows the camera to make use of multiple light paths, from which can be inferred dynamic patterns in a plane normal to the ‘line of sight’. The autofocus facility can then infer how distant the dynamic pattern in centre field is from the camera. (In fact this information is not normally then passed on to the recording system but it could be.) The metering system allows inference of average illumination and the white balance facility allows dynamic patterns of interest (in common parlance, objects) to be inferred with a correction for any general skew in spectrum. Image stabilising facilities further improve the precision of the specific dynamic pattern to be inferred. The camera also has a clock that allows the record to include an episode (or ‘event’) tag and a time for that tag. All this is very similar to what goes on in the brain stem and primary visual cortices in our brains, with a bit of temporal lobe beginning to be thrown in.

Once all this collative work has been done the camera generates a stream of signals in binary code that specifies a particular instance of a dynamic pattern. For some cameras this includes further reference clues, like the model of camera, which arguably begins to resemble the cues we use for a sense of agency or the waking/dreaming discrimination. This stream tends to go to a firmware device with a magnetic storage facility that is most typically a card held within the camera. However, for webcams the output goes immediately out of the camera to a cable that may reach any number of sites anywhere in the world. The webcam at the top of the Gornergrat in the Alps (http://www.zermatt.ch/en/page.cfm/webcam_gornergrat) may be a knowing machine with nothing that could be a local knower at all. Moreover, the knowers it does feed are many and widely distributed across the globe.

There is a current fashion for trying to resolve such tensions by attributing knowing or experience or other related concepts to ‘systems’. Even if I look at the picture from the webcam in London it can be said that the webcam and I form a knowing system. However, as I have indicated earlier I cannot see how this works in a causal dynamic explanation of the sort that science is supposed to limit itself to. We cannot meaningfully say that the camera lens knows anything, nor the shutter, nor any individual wire. These are all involved in the collation process that gives rise to a pattern of signs but that to which those signs can be manifest is something quite separate. I would agree that if we use the term knowing in the way it is used in ordinary conversation then maybe we have to conflate knowing machine and knower, but I would suggest that this merely highlights the poorly worked out nature of our common concept of knowing. It is a pseudodynamic concept that conflates two quite different sorts of going on. If we want a coherent causal account we need to separate those goings on.

It might be argued that in the human case the knower cannot be separated from the knowing machine because it provides feedback to memory paths and thereby forms part of the knowing machine. That may be true but it does not alter the distinction between two different domains relevant to two different sorts of goings on.

If we then follow the analogy with the camera and consider whether we can regard a memory card as a knower we start to get concerns about how an array of magnetised elements could be a knower. I will discuss these in more detail in section II but would like to mention some of the issues here.

Being a knower in the sense being discussed seems to involve receiving signs that in some way specify a small subset of instances of dynamic patterns in the world. The first point is that receiving is not storing. If the incoming signals alter the magnetised state of elements one by one, as they do, then when the 1000th signal in a megabyte image arrives the previous 999 signals are no longer being received. Their existence is merely reflected in the sate of the card. Thus the receiving structure getting the 1000th signal includes the states determined by the first 999 signals. There does not seem to be any reason why these pre-established states should form manifest signs for something of which they are already a part. There seems to be a strong case for thinking that what is manifest in experience has to reflect receiving in real time, without an appeal to what are effectively stored signals.

A related difficulty arises with popular theories of mental representations that suggest that a conscious representation of a dynamic pattern consists in a pattern of ‘activity’ or signal traffic within a network of nerves, the domain of the network sometimes being related to ‘re-entrant’ traffic in feedback loops. Such a model would be a bit like a train time indicator on a station platform with light bulbs or LED in a close packed array being used to display words. This would seem to be closer to a storage situation than a receiving situation and as such it is unclear why it should lead to any relevant manifestation.

Another problem that we have with the card as knower is that it does not seem to have any means by which to allocate different sorts of significance to different incoming signals. The binary signals encoding the date the picture was taken will not interact with the card in any different way from that for signals encoding pixels. We think of that sort of task as being performed by the computer we connect the card to later in order to display the image. However, we are not going to find our knower in that computer because all it really does is convert one sort of array into another sort of array, the interpretation of which as a picture taken at a time of a particular object occurs in the head of the computer user.

It might be argued that we also need to have some account of how internal signals can be manifest as signs of distal dynamic patterns at all. We do not seem to have an account of this for a camera card. However, we do not have an account of it for human subjects yet, so we do not have strong grounds for thinking that the camera card is any more inadequate on these general grounds. The more specific difficulties given above are I think more useful than a resort to the common intuition that camera cards are ‘just not the right things’ to be subjects like people.

Rather as Descartes argued, it seems that we need another aspect to the story to make sense of what a knower might be. Again, the suggestion is that we should forget about the idea that anything receives the signs of the dynamic pattern in the world. Yet, even in the camera there is such a thing and there does not seem to be any naturalistic account that can do without it. It is just that we need a camera card that has some extra features. We do not rally want to go with Descartes in suggesting that these extra features belong to a category that is entirely unique to human souls. It seems reasonable to think that a human knower will be unique in certain combinations of dynamic parameters that different even from those in a gorilla. The difference might be considered qualitative, but we do not want to suggest some entirely new category of substance. Nor do I think the suggestion that the human subject is an ‘emergent’ property helps, unless some specific candidate with a causal role is suggested.

These arguments are very much linked to the idea that the knower is something to which that to be known is manifest. I do not dispute the fact that there is a sense of knowing that does not require this, most obviously in what has been called tacit knowledge. However, tacit knowledge seems to be something that we can reasonably understand as existing in a robot that combines a knowing machine with something that uses signs of outside dynamic patterns to determine motor activity. No deep analysis seems to be required. It is the sense in which we think that at some point knowledge actually has to be manifest to some subject that is worth further analysis and thus the main focus of my text.

In this context I would point out that for knowledge to be manifest as a sense of something I do not think we need require any sort of introspective sense of ‘I know that I am knowing”. Most of the time when we say we consciously know something we are referring to a sense that ‘something is so’. That seems to involve have a sense of a particular dynamic pattern and a sense of the scope of its instantiation. The latter sense does not seem to have much ‘phenomenal thickness’ to it most of the time but it is difficult to conceive of senses of dynamic patterns without this context. There are difficulties in knowing just how one should describe things here, but the simple point that I want to make is that we need to be able to give an account of the sorts of manifestations of signs we do have without necessarily getting caught up in introspection (or the spelling out of propositions).

Key point: a dynamic description of knowledge must distinguish clearly between the system that collates signals, or knowing machine, and the subject that benefits from this collation process, or knower.

5.54 The problem of the knower knowing

The separation of knowing machine from knower, as here defined, generates a curious paradox for the knower – in one sense it cannot know anything, or at least it certainly cannot verify that it is knowing. It has to rely on the knowing machine for that.

If a knowing machine generates a pattern of signals that have the right relation to the outside world to be signs of a particular dynamic pattern and passes them on to a knower the knower has to take ‘on trust’ the claim that these signs relate to something truly instantiated in the world. The knower has no means of validating its input in relation to world events because to do that it would need to be a knowing machine. Put another way, it would need another input to check the first input against, and on ad infinitum.

There might seem to be a way around this. It might be argued that a sufficiently complex pattern of signs could only really be generated in response to a very specific pattern of dynamics in the world, thus in a sense having all the checking information included within it. Unfortunately, this does not stand up. If we go back in any causal chain we very rapidly come across a combinatorially expanding number of possible precursor dynamic patterns. That is why we need knowing machines. As long as we know the precise mode of operation of our knowing machine we can be sure what dynamic patterns in the world it is producing signs of. But the knower has no way of knowing the mode of operation of the knowing machine that feeds it. Virtual reality equipment shows this clearly. Even if the part of the knowing machine in our head stays the same, if it is connected to what is in effect a mock dynamic pattern source rather than to the environment in the normal way then the knower is presented with the sense of knowing signs that are not of the dynamic patterns in the world they seem to be of. In dreams our knowing machines seem to be replaced by something uncannily akin to the mad scientist in the brain in the vat story. Our brain seems to unplug the knowing machine and plug in a DVD of what is usually a very macabre narrative of totally unreal dynamic patterns.

This I take to be another good reason for thinking that knowing machine and knower should be considered as separate entities. It also has a significant implication for the relationship between any manifest sense of knowing and its causal antecedents. It implies that the manifest sense of knowing for a knower can only be guaranteed to be correlated with the dynamic pattern immediately causing that sense. As soon as we go further back in the causal chain, even to the first branch in the antecedent knowing machine’s pathways we lose any certainty of correlation because there are an unknowable number of possible combinations of antecedent dynamics relating to an unknowable number of possible knowing machines. An implication of this is that there is no way we can expect manifest experiences to track some ‘functional role in the world’ that does not specify the precise mode of operation of the knowing machine involved. Since at present we have only the very faintest clues about the most general principles of operation of biological knowing machines such a concept of ‘role in the world’ or just ‘functional role’ is devoid of any explanatory power. Moreover, it is not clear that such a concept would have a useful meaning even if we did know everything about knowing machine operation.

Another implication is that although ‘knowing about something’ in a factive or third party sense in which the something is externally defined, as in John knowing that tomatoes are red, requires the multiple paths of a knowing machine, having the sense of knowing, as in my knowing that God is on my side, or indeed that tomatoes are red, does not. The manifest sense of knowing does not have to go with knowing through multiple paths. The origin of a pseudoconcept becomes clear here.

This brings us to the troublesome question of how manifest signs entrained by dynamic patterns in the world can be associated with not just apparently arbitrary ‘qualia’ like blueness but a sense of narrative, relation in space and time and ideas like ‘belonging’ that have no apparent dynamic counterpart. Are all these arbitrary illusory senses that get thrown up during the playing out of the dynamics and which bear no relation to how the dynamics actually work?

I think this is too pessimistic a suggestion. One comforting thought is that we are likely to have evolved so that there is some non-arbitrary link between sense and dynamics just on grounds of that leading to useful regularity. However, I see no reason not to think that at least in our childhood view our sense of what we know relates to the real dynamic patterns in the world in a much more indirect way than we assume. What also seems quite comforting is that we do seem to be able to identify places where the relation seems to lead to inconsistencies. On the other hand there is more than a suspicion that there may be a long road to travel before we even get to understand the true relation as well as we could.

What I think may persistently dog our attempts to understand the apparently peculiar fact that we get a sense of some distal dynamics from receiving the effect of some quite other proximal dynamics is the immensely difficult task of letting go of the idea that there is something that any particular goings on in the world is like. If we can see that to have a sense of knowing about X we have to receive signals that result from a process of inference of the existence of X but to receive these signals through a pathway that does not involve that inference at all (even if it were to involve further inferences) and that all examples of ‘what something is like’ involve this peculiar indirect relation then it may become easier to see that we should perhaps not be surprised about this strange business of aboutness.

Another significant implication of this line of analysis is that although we have no reason to think that in the world as it is at present there might be some inanimate form of knower (as here defined) that might have a sense that it knows that the Henry VIII had six wives, because there simply aren’t the right sort of knowing machines around in inanimate matter, we have no reason to think that anything actually needs to be connected to such a sophisticated knowing machine to have such a sense. If we knew the right signals to feed in there is no reason why the sense should not be generated because generating the sense is something separate form the inferential goings on in any knowing machine. At present we have no idea what sort of receiving would give what sort of sense but maybe certain sorts of biological material are suited to receiving things in an appropriate way and we could build a mock up of this in a lab.

In one sense this looks like an experiment destined to fail, on the grounds that we would not be able to tell whether or not our laboratory synthesised knower was sensing knowledge of the marital status of Henry. There might even be good reasons for thinking that there could be no fact of the matter about this claim. Nevertheless, there is evidence from approaching the problem in a different way that may soon become rather convincing. We are on the verge of being able to build artificial knowing machines, or partial knowing machines that can be plugged in to human subjects. Cochlear implants (maybe soon retinal implants) can give a sense of sound for those whose natural sound-generating machine (cochlear) does not function. This is only one small step towards getting at the final interface between knowing machine and knower within the brain and discovering what signals would need to be passed there to give a confident sense of knowing that King Henry IXth had seventeen wives, or that colourless green ideas sleep furiously, but it is a step.

An important implication of this line of thought is that we have no very good reason not to attribute to animal subjects the sense of knowing things that we consider quite sophisticated or requiring the complex inferential activity of an august academic department. For instance, we have no reason to deny that fish may not be hard wired to have a sense of their own mortality. That need not be provided by a knowing machine with the right inferential powers. It could be genetically determined through a DNA sequence that had evolved through natural selection and which encoded a bank of cells that were set up to give such a sense whenever a predator was nearby.

There is, however, a final twist to this line of analysis, which leaves things in the air, at least until we address some further issues in section II. If the human subject as knower as here defined has no recourse to checking what it knows then can we rely even on the sense of knowing that something is manifest to something in association with ourselves. Certainly there seems to be considerable room for doubt about the richness of what is manifest, with Colin Blakemore suggesting 50 bits and Victor Lamme suggesting that a detailed panorama is just as rich as it seems. I think the broader question of whether or not anything at all is manifest is not quite as fragile as that because what is in doubt is the relation of a manifestation to the real distal goings on in the world, not to its own existence. Nevertheless, this is not an easy area to negotiate and I would not exclude the possibility that we will eventually have to re-introduce more severe forms of scepticism and try to close an even more complex circle of relations to get a working model of our relation to the world.

Key point: the relationship between knowing machine and knower appears to entail a paradox in which there is no means by which the knower can validate what it experiences as known. Knowing is probably still a useful concept but we cannot expect it to fulfill some of the idealised criteria laid down in philosophical circles.

5.55 A return to the problems of systems and functional role

I have already indicated my reasons for rejecting ‘systems’ or ‘functional role’ accounts of our sense of knowledge of the world. I would like to return to these approaches and give some more detailed arguments in the context of the foregoing sections.

The primary objection to any account of a sense of knowing as a property of a system is that any ‘system’ is just one of vastly many possible ways of subdividing the dynamics of the universe to yield a domain that suits the preferences of some interested third party. What system would support the experience I seem to be having? Should it include my kidneys, my hypothalamus, my primary visual cortex, the glucose in my blood or what? Which is the system that is knowing about this puzzle?

There is a way around this that in one way may be less problematic than it looks. We could say that for all possible system domains there is an appropriate sense of knowledge but a verbal report of knowledge will only reflect those systems that include in their functions the production of this report. The senses of the other uncountable number of systems are not available to any output so they remain totally invisible and ineffable. The first problem with this, however, is that it rather destroys the point of talking of a system in the first place. The second is that it looks awkwardly self-fulfilling. But the third and overriding problem as I see it is that we have established that the only thing we can expect to be associated with the sense of knowing is the final receiving event because in every case we can test it seems to be true that the precise nature of antecedent events does not matter. Moreover, any later consequences implicit in a functional role may always be baulked.

The main motivation for giving an account of experience and a sense of knowing in system terms seems to be a perception that at all costs one must avoid Descartes’s mistake of proposing a central receiving soul, for which no candidate can be found. There is also the idea that our superior sense of knowledge must be related to the size and complexity of our brains. And further there is the invocation of the concept of emergent properties of complex systems that seem to have dynamics that simpler systems cannot have. All of these motivations can be called into question.

A specific approach taken by Varela is to link our inner life to our existence as self sustaining or autopoietic, life units or systems. An additional motivation here is probably the sense that having a sense of knowing is confined to biological entities. The trouble with this is that there are a lot of people who are not self-sustaining units whose inner life seems no different from others. There are those who depend on renal dialysis to survive week by week. Steven Hawking has not been self-sustaining for many years but appears to have an immensely rich inner life. Children are not very self-sustaining. On a wider scale there seems to be a strong argument for saying that most people survive more through the combined behaviour of a global social system than their own. And if we are looking for self-sustaining units, the individual organs of the body and the individual cells within them might be considered to have just as good a claim. The autopoietic concept seems neither necessary nor sufficient as a basis for a knower.

Another approach is to consider the system that gives rise to a sense of things as consisting of both the human being and her environment. In the enactive approach a sense of something is seen as arising from the feedback loop of input and exploration of the environment through motor pathways. To sense the sponginess of a sponge is to complete a loop through the act of squeezing to the sponge itself and its compliance to stimulation of touch receptors, inputs to the brain and round again through more squeezing. As a description of a ‘complete knowing system’ consisting of all paths within a knowing machine and those that relate it to what is known about this seems quite reasonable. For some of our sensory pathways, although not for others, motor activity is often essential to the collation process that I associate with a knowing machine. But we come back to the fact that this is a separate matter from being a knower. When we go to sleep the knowing machine and the complete knowing system are unplugged and the knower is plugged into a dreaming machine that produces just the same sense of knowing with no loop to the world at all. Moreover, although proprioception is often integral to vision, vision is not lost without it and smell seems to have no particular dependence on any motor loop. We have a sense that we know what the moon is like to us but we are never in a position to know it through changing our relationship to it through motor activity.

My main criticism of accounts of our sense of knowing in terms of ‘functional role’ is given in the preceding section but I think is worth emphasising. There are a host of problems here. As Seager has pointed out, functional roles can be attributed at all sorts of different levels for any one event, ranging from accounts in basic physics through biology to sociology. It is unclear why the sense of something should track to any one of these. We also have the problem that the only event that seems to be necessary for a sense of something is the one most intimately associated with that sense. There is the further problem that functional roles are always couched in terms of a language of ideas of the world that are themselves based on what things ‘seem to be like’ and frequently follow the childhood view that the world is full of stuff, things and people etc. These seem to be very fragile conceptions to pin a functional role on that is intended to explain the genesis of such conceptions. A further irony is that the invocation of functional role seems to be a way of trying to tie the account to some sort of external causal behaviourist analysis. Yet in doing so it seems to dissociate itself from any involvement of the immediate causal events performing the ‘role’. It requires nothing of them. This very much goes against the grain of what we see in science in general, in that where we have a black box that does a particular external job we are very careful not to presume what things are going to be like inside the black box. To add insult to injury there is a good argument for saying that the functional role of a knowing system is independent of the pattern of signals it may be considered to be ‘knowing about’ at any point in time. If the functional role of a knowing system is to convert an input to an output according to a set of rules R then it seems that this functional role is constant and independent of the input. The last problem, that nothing can know its own output, I will return to in the next section but one.

Finally I will add a few words about evolutionary perspectives. There is a fashion at present for suggesting that difficult problems, such as how our experiences come to have a certain ‘content’ or how we come to know things can be explained in evolutionary terms. As I see it this is a confusion of two quite different sorts of question. The natural selection of the marketplace might seem to explain why lots of people have collections of vinyl Beatles records but it does not explain how records generate Beatles music. Moreover, the generation of Beatles music is no longer the reason why these collections exist since everyone has thrown out their record players. Their ‘function’ is purely nostalgia. In biological evolution function as ‘fitness’ is just as evanescent and misleading an idea since evolution changes the environment as much as the individual. There is no such thing as ‘well-functioning’. The survival advantage of feathers may be an ability to fly for some birds but it almost certainly has nothing to do with their original evolution and has nothing to do with their function in penguins. It seems ironic that evolution is invoked to support teleological ideas when the great achievement of Darwin was to throw teleology in the waste bin!

Key point: despite being popular in recent times a functionalist account of our experience of knowing things appears to be causally incoherent.

5.56 The form of manifest knowings

In my introduction to a complementarity-based approach I indicated that the time and space of our experience are not metrics in the sense that dynamic time and space are. This claim is hard for many people to know what to do with. It seems obvious that the time and space of our experience must be the same sort of category of aspect of the world as the precise versions of physics. However, I think it is essential to an understanding of the manifestations of our experiences that we do not see them as being divisible into repeating contiguous units in the way that metric space and time can.

There is a popular debate going on in neuropsychology as to whether our experience of the world is as rich as we intuitively think it is or whether it is actually very limited. If I look at a wall of Iznik tiles or a Breugel landscape is what is manifest to me visually made up of fifty thousand elements or only fifty?

When this question is posed the case for fifty is usually illustrated by an image with a central area crisp and clear and the rest blurred. This may make the point but is in danger of making the wrong point. A blurred image may contain as many elements as a sharp image, just rather different ones. There is no reason to think that an image with fewer elements would be more blurred. It is not even reasonable to think it would look as if degraded to simple hard edged shapes as found in an infant’s picture book. Both of these misanalogies suggest something that is overtly underexplicit. But there is no reason why a manifest image with few elements should be overtly underexplicit. If all elements are put to appropriate use it should appear fully explicit.

The confusion here relates, as touched on earlier, to the fact that visual experiences are very unlikely to be made of ‘pixel’ elements in a spatial array. Apart from anything there is no structure in the brain with the rigid geometry to provide a framework for such pixels. Presenting images in the form of pixels would be highly inefficient, as recognised by the use of compression algorithms in picture files in computers. Moreover, pixel structures carry information, like their shape and the nature of their boundaries, which are redundant to what the brain wants to signify. There are no pixels out there in the world (except man made ones). There are no pixels in the retina. Pixels are just not relevant.

Hubel and Weisel showed that signals from the retina are collated in primary visual cortex to generate signals for things like edges, which have no pixels. Whatever receives the fully collated output of our visual knowing machine almost certainly receives signals that signify things like ‘completely straight edge’ or ‘perfect circle’ or ‘completely straight edge with one slight wobble’ and more complex versions of the same theme. None of these has any pixels or any sort of subdivision. One might ask how one can integrate elements of experience if they have no substructure. But in fact the more difficult question is how could one integrate elements of experience if they did have dynamic physical substructure. We know from the work of Zeki that colour , shape and movement are all inferred in different places in the visual cortex and then integrated. If the colour, the shape and the movement all had a specific spatial extent, how do you combine them to give an experience in only one place? If you want to combine fast with palomino with horse you cannot do it with fast Lego bricks, palomino Lego bricks and horse Lego bricks. Somewhere behind the scenes you need something to be going on in dynamic space and time but we should not expect the description of those goings on to bear any resemblance to that of the constitution of the experience. After all, the experience is not about the proximal goings on underlying it, but about some distal pattern in the world.

There are very few places in the literature where this seems to be spelt out clearly, but it is implicit in most texts on visual neuropsychology. In Marr’s book on vision he says that the final generation of a full 3D image is too complex to analyse in neurological terms but suggests that it must work rather as adjectives join together in a sentence. Barlow also talks of elements combining with the economy of words. Charles martin also makes the comment that verbal and visual imagery may not be as different as we think. We have no dynamic model for how this ‘would work’ but we should not expect to because the rules of how things will seem to a protagonist in a receiving interactions are not going to be the dynamic rules of physics.

Considering the integration of elements in an experience as being like the integration of words in a sentence provides a starting point for addressing some of the mysteries of what we ‘sense’. If an edge is an edge rather than the boundary between a row of brown pixels and a row of black pixels then it becomes easier to see why it is often unclear whether our experience of an object is visual or tactile and that moving from one modality to another seems to require no translation. It allows us to sense things in what Aristotle called the ‘common sense’, by which he meant one in which there was no distinction between modalities of access.

An analogy with words may also go some way, although perhaps not very far, towards addressing how it can be that some elements of what we sense are ‘thick’, like colour, and some are ‘thin’ like the sense of someone being John, or Susie’s father. This is a very puzzling area, although I suspect much of the puzzle still relates to the problem we have in casting off the conflations of our childhood view. It might be argued that in the sort of complementarity-based account I am giving the knowledge that John is Susie’s father is not manifest to me. Nothing is up on the screen, so to speak. Yet if I am consciously thinking that I know that John is Susie’s father surely some idea is manifest. Something is different from the days when I forget about John and Susie but still know their relationship in the sense of having the disposition to entrain the idea if prompted.

Maybe it is just a mistake to get into any sort of account that starts suggesting something should be up on a screen. But maybe also there are important differences in the way thick and thin ideas are received by the subject from the knowing and dreaming machines. One thing that crosses my mind is that the human trick of language might have something to do with making use of arbitrary thick-seeming data, like grunts and scribbles, to tag on to thin-seeming ideas to help our brains’ librarians to find exactly the right material when needed. If our database hardware has an extra field for ‘library code’ that can use the sort of sorting tricks we use for thick data and an ordering syntax then an obscure but fascinating volume of philosophical ideas can be instantly retrieved under the code P155DU7. The philosophical ideas will still seem thin, but they are easier to get to.

As indicated in the earlier section on meaning, within language there may be an analogy for the difference between handling thick and thin ideas. Marr talked of visual experiences being integrated in the way adjectives are attached to nouns. Thus, big hairy brown dog builds thick aspects together. Thin ideas seem more likely to involve bringing in verbs, as in John is Susie’s father. Nevertheless, much of the structure of language seems to be more concerned with imparting an increment of information and I think this means that syntactic categories cut across the thick and thin divide quite a lot in practice. Verbs get morphed into adjectives and nouns. These effects may get unpicked in formal syntactic analysis but there is no general agreement on exactly how that should be done.

The senses we have that philosophers call propositional attitudes, like belief and knowing, are particularly difficult to form a clear account of. A tomato is an instantiated pattern of dynamic disposition that can, through a knowing machine, give rise to a sense of a tomato. The knowledge that P is an internal dynamic disposition that can give rise to a sense of knowing that P, but it can also give rise to all sorts of other ideas. Knowledge and belief are dispositions that can lie tucked away unused, or can be made use of, but not necessarily to give a sense of themselves. Thus it can be confusing just to talk about belief or knowledge being instantiated.

As indicated earlier, the form of manifest knowledge is perhaps the thinnest thing of all. Most of the time it is almost as if the experiential adjectives and nouns are just presented in the absence of the asterisk linguists use to indicate that a sentence is ‘dodgy’. The sense that something is so, or that its expression in a sentence would be true may be a sense that it passes muster in some way but in what way seems too thin to catch hold of and is mostly just implicit. However, it may be true to say that the thinner the ideas we handle the more explicit the sense of needing to pass muster is.

What determines whether an idea comes thick or thin? One possibility I can think of is that thick ideas are signs of dynamics that we have ways of cross checking through other pathways. At least some of that cross checking is across different sensory modalities and this is where the ‘enactive’ aspect of our knowing machines is highly relevant. These are ideas that often have a distinct relation to episode and probably need to be stored in memory in episode context. This ideas may tend to be at the highest level of inference and not so obviously related to episode. We may have no further cross collation that we want to store in a spatiotemporal episodic form and so although they may act as centres to which other ideas are co-tagged they do not require any internal structure in their manifestation themselves. That of course begs the question as to what the relationship might be between the seeming ‘structure’ of manifestations and the proximal dynamics that are associated with them being manifest and at present I do not think we know enough to try to address this.

How some proximal dynamic interaction in a brain can give rise to the general sense that ideas form part of a narrative really is hard to handle but I keep coming back to the fact that we have no business to think that it is the job of manifest seemings to seem like stuff, since stuff does not in itself seem like anything. Manifestations are what dynamics determine, not what the dynamics are like. Maybe we should rather expect manifestations to have this narrative quasi-linguistic form. We could reasonably expect that narrative form to be appropriate to the dynamic job to be done. However, the dynamic job to be done involves the happy combination of vastly complex mathematical transformations in our knowing machines and if this is the first time evolution has thrown up a system that just about works at this level it may turn out to be full of bugs. Maybe insects have got almost as far with a completely different way of generating manifest signs of space, time, purpose, death and all sorts of other things.

Key point: the form of manifestations that we experience as instances of knowledge is very poorly understood but seems to have some similarities to the form of language syntax, even when non-verbal.

5.57 Knowing and physics

Throughout the development of twentieth century physics questions have been asked about the relationship between knowing and the very existence of fundamental physical events. Schrodinger devised his famous cat thought experiment to try to defuse the suggestion that his own ideas might imply that what is is simply what is known. Quantum theory talks of values being determinate and the word determinate clearly has something to do with determining and knowing. It has been said that modern physics replaces matter with information.

I am sympathetic to the idea that the world we inhabit may be more like a fabric of interwoven stories, build of information, rather than a huge Perspex box full of billiard balls. On the other hand the idea that what is is what is known runs into all sort of problems. Who would know whether or not something was known and known by what? What knows the movements of every atom in my cup of tea? And does knowing once suffice, or does something have to remember it knew? The complexity of the process of knowing, even if we accept that there is such a concept, rather than a conflation of incompatible concepts, makes it fairly certain that we want to allow things to exist without knowing that they are known.

One implication of this is that we should probably take the pragmatic approach of Richard Feynmann and assume that something can be considered to be determinate once it has been measured by a device, irrespective of whether or not the leads to a human observer knowing the result. Wigner’s early suggestion that human consciousness had some unique effect on the way the world really is is as absurd as Schrodinger thought it was.

Taking a panpsychist approach this raises the possibility that we should consider the measuring device, or some part of it, to ‘know’ the value of the variable measured. Certainly we would expect something to be manifest to elements of the device. However, taking into account what has been said above about the multiplicities of pathways that appear to be needed before we can expect knowledge to be framed in anything like the terms we are familiar with, it seems that we can be pretty sure that such knowledge in a machine would be of a very feeble ‘indistinct’ sort. (This, I think, provides reassurance that Leibniz may have been justified in making his claim that his ‘perception’ might come in a common ‘indistinct’ form and a rather special distinct ‘apperceptive’ form for human souls.)

That is not to say that the signals that the measuring device pass on are indistinct. The whole point of quantum measuring devices is that they are set up so as to massively amplify tiny signals from quantum systems to a macroscopic level. The crack of a Geiger counter picking up a single electron or the trail in a cloud chamber is distinct to human viewers. However, any manifestation to parts of the device will be indistinct in the sense that it will not be received in the context of other information that could allow something in the device to infer ‘what the signal was about’ as discussed further below.

A second implication is that it may be worth reviewing, in the light of the involvement of multiple paths in knowing, just how many paths are involved in the concept of a physical value being determinate. A number of experiments seem to indicate that as long as there is a single aspect of the environment of a dynamic packet that allows some signal, however weak, to be passed to the environment in a way that could be passed on to an observer then whatever value this signal could convey is determinate in the sense that it will never be found to be otherwise.

It might seem from the discussion on multiple pathways that this sort of determinacy might need at least two paths. Bohr emphasised that fact that to attribute a value to some quantum system we need to have information relating both to initial and final conditions. In simple terms, we cannot attribute a value on the basis of a black dot on a photographic film unless we know the film had no black dot beforehand. We cannot attribute values in many cases unless we know parameters for our measuring devices. It begins to look as if determinacy really does not, if not a conscious knower, at least a knowing machine.

On the other hand, what quantum theory implies is that a value will never be found to be otherwise if we can be confident that some signal has been passed that could have led to an observation. If we are interested in ‘what is really going on’ then it seems that determinacy does not require the presence of a knowing machine. It is a much simpler phenomenon. It only requires one signal.

Interestingly, what I think this shows is that we may be able to equate the idea of determinacy with the idea of manifestation. This would be manifestation in the sense of both white and red stars in section 3.17. For a human knower it equates to an experience based on an input, which may be knowing if we assume an effective knowing machine, but could equally be a fantastic narrative fed from a virtual reality machine or one of the amorphous immanent horrors that occur in delirium.

This makes manifestation and knowing quite distinct concepts, both based on a causal account consistent with physics. The account for knowing is complex and by definition not circumscribed, because it is an asymptote to an impossible ideal set of dynamics. Nevertheless it is explicit. It restores a clear distinction between ontology and epistemology, if epistemology is about knowledge, but it is also consistent with the idea that the world is composed of information in the sense of manifest signals. This might seem to be in conflict with Leibniz, who seems say that there are no signals sent between things, there is just a progression in harmony. However, in modern physics there are no signals either, if we are talking of mediating events. Nothing carries a message from an electron to a photon. So ‘signal’ is as irreducible and immaterial in modern physics as it is in Leibniz – in a sense it is just a metaphor left over from the world of stuff and things.

Another distinction between manifestation and knowledge is aboutness. Manifestation does not imply any aboutness but knowledge does. Talking about manifestations without slipping in to an aboutness mode is very hard but I think it can be done. It is the motivation behind the concept of qualia. We just have to be careful that it does not constantly get hijacked by implications of what the qualia ‘are about’.

It is also important to note that when a value is determinate in modern physics there is no sense that there is some ‘thing’, such as an electron, that becomes determinate. There are not two different ways for things to be, in the sense of sometimes being fuzzy and sometimes being sharp. The position of a dynamic packet may be determinate without the momentum being determinate.

An important implication for of all this considering the human subject is that in a manifest experience we should expect the input to be composed of determinate features but we should not expect the subject to which the input is manifest to have determinate features in this interaction. This would seem to be in conflict with the idea that a measuring device is always ‘classical’, perhaps interpreted as meaning having determinate features all the time. In fact the goings on that are the measuring device are as indeterminate at the level of quantum grain as any other. What classical means here is that the measuring device is something constructed in such a way that its behaviour as a detector is not significantly affected by goings on at the level of quantum grain. This is usually achieved by setting up highly unstable dynamics such that even the tiniest perturbation leads to a guaranteed change in the device’s dynamics at a level well above grain. This means that the device can be treated as if it were behaving according to classical physics. The same applies to a knowing machine, which must operate well above the grain of individual dynamic packets to preserve signal free of noise. None of this is relevant to a subject as knower that simply has to have a direct immediate input.

Key point: the distinction between something being manifest and something being known may clarify the meaning of the ‘determinacy’ of quantum physics, which may be closer to being manifest than to being known.

5.58 Self-knowledge

It is commonplace to talk about self-knowledge, but it may not be immediately clear what that might mean.

Inasmuch as our bodies exhibit dynamic patterns to the world, that we can observe much as any other, even if with a mirror, there is one rather simple form of self-knowledge we can pass over. To know that one laughs less often than some people is easy.

Neuropsychologists are more concerned with our ability to know our own ‘minds’. What this appears to mean is knowledge of the dynamics of one’s own knowing machine through the manifestations to a knower. It can also include knowledge of our dispositions to feel emotions, desires to act in certain ways and perhaps other things but I am chiefly interested here in knowing about the mode of operation of our knowing machine. We can know, or think we know what sorts of questions we are good at answering, what deductions we are good at making or what we are good at remembering.

What seems to get lost in this analysis is any discussion of what we might know of the dynamics of our knowers. This seems to be at least in part because of a tacit agreement amongst both scientists and philosophers that such subdivisions of the ‘self’ do not exist. There is a pervasive view that the human self is equivalent to the human subject and that can also be equated with the human organism as a whole. This is very strange because at the same time the neuroscientists are dissecting in more and more detail the dynamics of the brain and it is quite clear that a lot of it is involved either in getting things sorted before being ready for the subject and another lot is involved in sorting out how to co-ordinate motor output. It is almost universally agreed that Descartes was wrong to suggest that some internal structure received the final results of the incoming data stream from the world. And yet there is no earthly point in having such a data stream unless something does receive it! As indicated before, knowers may be components of loops in knowing machines but they are also consumers of the products of those machines.

As I shall explore, I think there is in fact a very good reason why we may seem to have little or nothing to say about the dynamics of knowers, but that does not mean that we should not ask questions.

The weakness of the all-or-nothing distinction between whole human being as self and outside world crops up in all sorts of ways. I have mentioned the blind man’s stick as part of his knowing machine. We can examine visually and by touch our limbs and faces just as if they were outside world. We gain knowledge from these using the same multiple pathways we use for tomatoes and the like. As we go further in things get a bit different but it still seems there must be multiple pathways that we use to monitor our internal dynamics. I do not want to analyse these in detail, but just give a few examples. We can check our mental arithmetic. We can amuse ourselves observing the strange dynamics of not quite remembering a name until it suddenly pops up a while later. We know that we have knowing machines that build up concepts of objects from many angles. We know that we have dreaming machines that play us stories that sometimes seem to indicate an insight into our life situation our knowing machines had not yet cottoned on to.

As a general principle it seems that for knowledge of the dynamics of our knowing machines we need a part (A) of the knowing machine to receive signals from a part (B) and also from a part (C) that also supplies (B). If A is told what C is signaling and also how B responds it can infer/know the dynamic dispositions of B. This allows us to think 'it takes 'me' about four seconds to multiply seven by twenty four'. The situation of A getting signals from C directly and also via B seems likely to be a common brain trick - it would help explain how we experience the raw red of the tomato at the same time as the red as the tomato (via some parietal B?). Nevertheless, it also seems reasonable to suggest that the human level of self-awareness reflects some extra path of this sort that allows a further level of meta-analysis.

Key point: self-knowledge has to be divided into knowledge of our knowing machines and knowledge of our knowers. The former is likely to be considerable, but different from our knowledge of the outside world. The latter is a quite different matter.

5.59 Self-knowledge of a knower

When we come to the knower itself things begin to look rather different. If the knower is the true human subject, or self as soul, then I think people may be badly mistaken about any sense of self-knowledge. If we consider what the knower can know about it seems that it must be limited to what information it receives. That information does not come with any indication of distal causal history, which is why we can be fooled about the origin of incoming signs by virtual reality apparatus. This has the implication that any narrative received by a knower that seems to be about the inner dynamics of the mind cannot indicate the role of that knower in the source of that narrative in a way that the knower can verify through some other means. It may receive a narrative that the knowing machine has inferred derives from the very knower being informed, but it has no way of knowing that the knowing machine has made a reliable inference.

What this boils down to in general terms is that any entity qua knower can be informed by signals that derive from its own output but it has no means of establishing that they are indeed from its own output. This leads to the important principle that any knower within a brain cannot know whether it is the only knower present or whether there are many others knowers also present. We have reason to think that our sense that we know how many tomatoes are on the table is likely to be reliable because we can see a means and a clear survival advantage in knowing that. On the other hand it is very unclear what the survival advantage would be of our knowing machines giving our knowers an accurate account of how many there are of them. After all we get no sense of anything much about the internal structure of our brains and it is very unclear how the knowing machine might assess how many knowers there are anyway. It might be able to infer how many different classes of unit there are with certain computational dynamics that we might relate to decision making, but that does not imply any inference about there being manifest experience. That fits with the fact that the relation between conscious experience and decision making is very uncertain. I will return to this issue in detail in the discussion in section II

Another issue about the dynamics of the knower is that of how much we can know about the richness of what is manifest to a knower in experience, as discussed in the last section. How would we be able to know that? I will return to such issue in section II but it seems to me that there must be streams of data that pass through pathways to reach knowers and that these pathways have side branches that feed back to collating mechanisms that generate an idea of having experienced a certain multiplicity of elements. This may seem strangely oblique but remember that there is no possibility of any ‘fly on the wall’ inside the brain being able to take a look at what is being passed from A to B. Only B will get that and respond according to certain rules. If those rules are designed to generate a useful response to what comes from A they are unlikely, in the normal situation, to be designed to indicate how many elements there were. Some strange paradoxes arise here that again will need further discussion in section II.

[image reference is broken]

Fig. X: A highly simplified and stylised representation of a knower unit/domain within a brain. The knower unit exists amongst other units that have been assigned ‘knowing machine’ and ‘motor system’ status. However, since feedback at all levels is possible these have to be considered arbitrary and overlapping categorisations. The knower unit is no different from any other in receiving multiple inputs from other units and having an output that reaches multiple units.

Consider the implications of the relation of knower to the rest of the brain as given in a highly simplified form in figure X. The knower may receive inputs that are in part influenced by its own outputs through feedback. However, it has no means of knowing that the influence originates from itself. The knower has no access to its output directly. The mechanism of output may be associated with some ‘back reaction’ on the knower but this is not access to the output, which is the influence of the knower on other units and that will depend on the receptivity of the other units.

The knower can know about dynamic patterns through having inputs, at least in the sense of receiving signals that generate the corresponding manifest signs, but it cannot know about the input signals themselves, or how diverse they are, because that would require multiple pathways, which implies other inputs and an infinite regress. We have the curious paradox that a richness can be manifest to the knower but the knower cannot has any way of inferring that richness and sending on signals that will form the basis of a reliable report. The report of richness can only be reliable if it is based on inferences made previously in the knowing machine. This might lead one to a mysterian conclusion that in the end we have no way of knowing what a knower is or what its inputs are. We might even conclude, in a positivist vein, that the idea is meaningless. I think this is too pessimistic, but there are some further details I want to explore in section II before I try to justify this view.

This issue about not being able to check the origin or form of an input at the point of input is not, of course, special to what we designate as knower. The knowing machine is effectively faced with this problem at every point of input into its component units. This does not detract from the machine’s ability to draw valid inferences about dynamic patterns but it does emphasise another point, if we abandon the idea of a knower there is nowhere else that will satisfy our intuitive idea of knowledge. It is likely that the most common rebuttal of my analysis would be that in modern times we have given up the idea that there is a receiving knower. Yet this view reduces knowledge to a third person positivist behavioural account which does nothing to explain phenomenal experience, and the sort of people who reject a receiving knower mostly still want an account of that. Without a receiving knower we have an account that is as divorced from the normal causal dynamic account of science as the medieval scholasticism that Descartes rescued us from. Manifestations must be part of a causal moving on, which means a sending and a receiving. There is a lot of double-think going on in the world of metaphysics at present.

If nothing else, some sort of receiving knower or ‘feeler’ seems to be needed to have the dreams concocted by our dreaming machines. One might also ask where the dreaming machine is to be put in the diagram. Does the dreaming machine feed the knower directly or does it feed a subsection of the knowing machine, from which inferences are drawn. In dreams we have an emotional response to the narrative. Is that emotional response fed by the dreaming machine or is it a reaction of the knower, and can one even make that distinction? There is more work to do before we have a clear story.

In summary, I think the idea that self knowledge can be of and by an experiencing subject is a non-starter. Things are more complicated. Some consider awareness of the core self as being the key feature of the human condition and even a prerequisite for a claim to consciousness. I am sceptical. I think we may have more sophisticated knowledge of our knowing machines but I think we may be very badly mistaken about any central self as subject, and indeed, as source of ‘will’. I rather like the possibility that man differs from other animals in that he alone mistakenly thinks he is aware of his true self. As I see it the core subject as knower is peculiarly impotent when it comes to knowing what it really knows.

Key point: a conventional computational dynamic analysis indicates that a knower can have no real knowledge of itself. (The possibility of a dynamic analysis in ‘non-computable’ terms will be returned to in the context of specific models of the human subject.)

5.60 Summary

The more one considers the phenomenon of knowing the more it becomes apparent that it is complex, graded, and never absolute. At best it can tend to an asymptote of true belief. At worst it is perhaps the most overrated of human capacities. We may differ from animals chiefly in the complexity of our misapprehensions.

That human subjects have any reliable account of the dynamics of the external world is due to being connected up to a highly sophisticated computational apparatus or knowing machine. The operation of this machine is to a very large extent invisible to the subject. Thus, just as Spinoza pointed out that the sense of free will is simply a reflection of the fact that we have no idea what causes our actions, one could equally say that the human subject has no idea how it comes to know what it seems to.

The inability of the subject to have any reliable way of ‘auditing’ the reliability of its own knowing machine threatens to make a nonsense of the whole idea of knowing. Nevertheless, if we accept that if we take nothing on trust we may as well forget any attempt to understand what the world is really about, then it seems reasonable to accept that what seems to human subjects as knowing does indeed reflect what is really going. What does seem advisable is a constantly vigilant skepticism about the framework within which those goings on are presented to us.

Thus, although our experience seems to be incorrigibly present to us, since it is a sign dependent on the potential frailty of our knowing machine, we cannot take its meaning to be incorrigible. Even the meaning of ‘being present’ needs to be treated with skepticism. If interpreted in our childhood view of things it may be wildly off the mark. This makes the identification of the clear and distinct ideas that Descartes relied on a difficult task. It is not, however, a reason not to push our capacity for knowledge to the limits. It does seem to be remarkably successful in many respects. To have predicted that if energy levels never normally found on earth were to be achieved in a vast circular tunnel under Geneva we would observe a quantum of the Higgs field that gives all matter mass is pretty impressive.

What is perhaps even more encouraging is that our knowing machines do seem to come with some fairly powerful self-auditing capacities. The physicists are very ready to decide that even if the new quantum is the right size it may not be of the Higgs field. Physics and neuropsychology have managed to show not only that the conclusions of the child’s knowing machine need revision but that more powerfully predictive concepts can be put in their place.

The division of the knowing machine from the knower may seem problematic. Some might ask, what could the knower be, other than part or whole of the knowing machine? Even if the knowing machine makes unconscious inferences, surely it is the knowing subject that evaluates things like truth and beauty? I do not deny that these are very reasonable concerns and I will discuss how they might be assimilated when it comes to the nature of the human subject. However, I think one has to start off with the key premise that a knower, by definition, has little or no way to know anything about its own dynamics or dispositions in general. The human subject may turn out to be very different from what we have tended to assume, even if it does evaluate truth. Unless knowing machine and knower dynamics are considered strictly separately until we know what knowers are we are likely to go round and round in circles for ever.

Section two: The human subject

6 The nature of the human subject

6.61 Subjects and creatures

As Russell points out, all our explanations of what is going on in the world are underpinned by reference to experiences. Experiences are attributed to subjects and subjects are in everyday life considered to be human beings, or perhaps, more generally, creatures. Human beings are the traditional players in all sorts of branches of philosophy, but if we want a precise dynamic account of the world, to which we can link manifestations, it is not clear that they are the units that we should attribute experiences to.

Even within philosophy, the idea of creature, or organism, as subject seems more an arbitrary premise than a justified conclusion. In a lecture on the unity of consciousness to the 2012 Annual Meeting of the Association for the Scientific Study of Consciousness, the Oxford philosopher Tim Bayne began with the assumption that the subject is the organism. However, on questioning he agreed that the choice was arbitrary. The problem with this, as I see it, is that if there is any point in relating the concept of a subject to a physical dynamics entity it must be to link its subjectivity to its dynamics. If the wrong dynamic entity is chosen then all sorts of problems are likely to arise.

In fact, the basis of personal identity, not just of the subject but of the ‘human being’ in other senses, has always been a matter of dispute in philosophy, with Locke’s view that it is just a matter of a continuing thread of memory setting the sceptical agenda. In his 2011 Royal Institute of Philosophy Annual Lecture, Derek Parfitt, who has made personal identity his special interest, concluded that it is not tied to the whole creature, but is associated with somewhere inside the brain.

As indicated earlier, perhaps the current status of the philosophical debate is best illustrated by Andy Clark’s distinction between the ‘extended mind’ and the conscious subject. The extended mind hypothesis highlights the fact that there is no reason to consider the boundaries of the human body as particularly relevant. In terms of the functions I have designated the knowing machine, which is fairly close to what Clark calls the extended mind, the boundary can be thought of as well beyond, out in the shared goings on of the world. Nevertheless, when it comes to what Clark calls the conscious mind, like Parfitt, he sees the boundaries as being well within: not just those of the brain but of a subsection of the brain.

It seems that despite a tendency to lapse into ready-made ideas from Aristotle, philosophy is moving back towards the position Descartes found himself in. There is some small domain within the brain that is informed by a system that can stretch out into the world. Nobody wants to repeat Descartes’s mistake of choosing the pineal so they talk of ‘C-fibre stimulation’ as an example of what might be (but nobody is seriously suggesting is) the basis of sensing pain. At this point the philosophers plead humility and lack of expertise and suggest that it is not for them, but for the scientists, to pin down what in a brain is having an experience. Clark very reasonably adds that it will need to have the right ‘bandwidth’.

For much of the twentieth century the nature of the human subject was considered more or less out of bounds. The great bulk of science makes use of a habit already well developed in everyday life. Accounts of events ‘sanitise’ the troublesome concept of experience by replacing it by surrogate terms based on certain standard patterns of goings on in the world that can be expected to entrain functionally relevant experiences in others, like rulers for the sense of space, clocks for the sense of time and the sky for the sense of blue. This gives us a language of dynamic relations between patterns in the world which allows us to bypass the question of exactly which internal human goings on we should attribute experience to.

Even within neuroscience the majority of scientists even in 2012 may prefer to continue with this sanitisation. They may claim that consciousness is not a useful term, or that they do not do consciousness, even if they work on the physiology of brain cell signaling or the way brain cells form representations. Arguably, it took Francis Crick, turning to consciousness after solving an unrelated biochemical puzzle, to persuade scientists that it might be legitimate to look for the ‘neural correlates of consciousness’ (NCC). The presumption was, again, that manifest experience was specifically associated with certain small domains in brains rather than human beings as a whole.

By and large our experience-sanitising standard objects fall outside the surface of human bodies. There may be two slightly different reasons for this. Firstly, it cancels out any effects of events internal to human bodies that might be irrelevant to a description that is intended to be understood by other people. Secondly, because rulers and clocks are external we have a clear idea of the sorts of multiple pathways described in section I. that we use to know about them. I cannot use a retina the way I use a ruler, not just because someone else cannot use my retina but because I cannot tell that the teapot to my left is giving rise to experience by reflecting light through my pupils to the right sides of my retinae. Because I cannot interrogate my retina in the way I can a ruler I cannot build up a way of describing dynamic relations in the world using my retina quite the way I can for a ruler.

Nevertheless, I have already suggested that these differences are not watertight. In the UK rulers are still 30cm long because in the past the measure was the human foot. Moreover, we have various internal clocks and rulers for limbs and some of our experiences are of goings on inside ourselves. Pains are obvious examples. As indicated earlier, our knowing does seem to draw on multiple internal pathways in a variety of ways, even if it is much more difficult to delineate them.

Kinaesthesia is an internal affair. I can tell if my finger joints bend because of signals generated in joints and tendons in the hand and arm. Quine’s idea that all our information from the world can be described in terms of something crossing the body surface works in general, for the two reasons given above, but it is not by any means universally so.

Moreover, the significance of the body surface is that it is in some situations the limit of the private knowing machine – the apparatus that collates signals from the world so that dispositions can be inferred. There is no reason to equate this boundary with the boundary of the human subject. In fact it does makes no sense to do so, since the human subject cannot perceive a series of bright spots as a line until the information from outside has passed through neurons that collate the retinal signals from dots and identify the line pattern. The knower is by definition downstream of the knowing machine. It might turn out that the same network of nerves serves both purposes – that it passes messages on to itself. However, not only cannot this be assumed but it poses significant problems if we take evidence such as the work of Hubel and Weisel to imply that the primary input pathways in the brain deal with very different tasks than those performed by the frontal regions.

In summary, the philosophical premise that the whole creature is the subject of experience needs to be subjected to a vigorous Cartesian scepticism. To a first approximation there seems to be a good reason for roughly equating the human knowing machine or system with the contents of the body surface but only roughly. The experiencing subject seems likely to be much smaller.

Key point: human subjects cannot be equated with whole creatures.

6.62 Options for an inner subject

Despite the scorn poured on Descartes’s model of the subject the ground rules for proposing a model today have not changed that much and there is no indication that a better model has achieved consensus. We have reason to think that a human experiencing subject is constituted by only a part of a brain and very likely a rather small part. Modern neuroscience very much confirms what was apparent to Descartes: that much of the brain seems to be involved in preparatory collation of sensory signals, co-ordination of motor output and perhaps storage of memories not currently in use.

One can concede that a subject might possibly involve large parts of the cortex and thalamus but it seems more likely than not that it is inappropriate to consider the entire brain as the subject. As previously discussed, all the evidence we have indicates that if we go back in a causal chain, more or less immediately we find no necessary or sufficient relation to the nature of an experience generated by that chain. We have every reason to think that we could pare away not only the peripheral nervous system and the input and output channels through the brain stem but most of the primary sensory and motor cortices and with well placed electrodes generate a very normal seeming experience in what remains. Moreover, there is a strong suggestion that there is far more going on in a brain than would be needed to account for an experience. It is hard to see how many trillions of synaptic connections would be needed for a visual scene that is generated from the data in about a million input fibres.

Thus, so far it seems that we are still looking for a subdomain and perhaps quite a small subdomain of brain that is continually supplied with data from senses and/or memory in a highly collated form: for an apple in terms of outlines, solidity, reflectance disposition and fruity identity in conceptual terms.

Descartes was of the view that a human subject should be one single confined locus. He chose what is definitely the wrong site, the pineal. He may also have chosen the wrong type of topology. If the human subject is extended at all (rather than a point locus, which Leibniz pointed out might make things much easier) and biological processes would seem to require that it is, then there may be no objection to it being a delicate spidery net extending quite widely in the brain matter. The subject as neural network is currently the most popular view in scientific circles, in as much as there is a view. Giulio Tononi has built a model around this idea with the richness of experience being related to the information carrying capacity of a particular network subdomain or ‘complex’. Bernard Baars’s conception of a global workspace throughout which the contents of experience are broadcast, would seem to fit with a distributed network structure. Victor Lamme’s concept of activity in a connected network would do the same. I shall consider the merits of these models below.

A single confined locus of the sort Descartes envisaged is, as Descartes realised, difficult to find a place for, if only because most brain structures are paired left and right. However, if one forgets the duplication for a moment the cellular aggregates of the basal ganglia, and in particular the thalamus, might provide a local home for a subject. There are a few vocal advocates for a thalamic site of experience. Christof Koch has raised the suggestion that the claustrum, which is even more localised, if a rather flat shape, might be well suited to multimodal NCC. Even the brainstem nuclei have been suggested as loci for experience, by Jaak Panksepp in particular. I will consider these options too.

A third possibility is that there is no one locus of experience: that there are many such loci supporting experience all the time, or perhaps in turn. A well-known proponent of this idea is Daniel Dennett, with his ‘multiple drafts’ theory. Many people find this model too counterintuitive to handle, but I shall have considerably more to say about it in later subsections.

Finally there is the view that there is no defined locus for the subject, but that subjectivity follows patterns of activity over different areas of neural network as these change. This seems to have the advantage that it overcomes all the observations that indicate that no one particular place is where subjectivity resides. However, it hits a serious drawback when considering how patterns of signals might encode manifestations in terms of the difference between positive and null, or negative, signals if the subject itself is defined in terms of a pattern of signal traffic.

Any choice of domain for an experiencing subject depends on a range of perceived needs for a viable model. One perceived need is that that there should only be one such subject per person. Although widely held, I shall argue that the basis for it is unsound in principle. Another perceived need is for endurance through time of the subject. This is probably less universally required by modelers, with a number, going back to Locke at least, recognising the fact that human subjects could be evanescent, with continuing memory of identity stored outside experience. I tend to agree with Locke that endurance should not be a prerequisite. There is also the perception that the subject should lie somewhere along a causal chain that links sensory input from the environment (plus perhaps memories etc.) to motor output, whether or not experience itself is seen as having a place in that causal chain. I think this requirement is perhaps the one we have to take most seriously. That is not to say, however, that subjects might not find themselves in feedback circuits that give them a much more indirect and less ‘controlling’ role in output than many would like to assume. Finally, there is the apparent need for the ‘content’ of the experience to be delineated or bounded in some way that makes the experience a rich ‘unity’ of ‘compresent’ elements in contrast to elements outside the boundary that are either not experienced at all or, if experienced, only by other subjects. This seems to me a highly desirable feature but, as indicated in the section on knowing, it may not be as simple as one might like to think

I will consider various forms of each of the model types in the context of these perceived needs in turn. Before doing so I want to return to the central issue of how we might fit the manifestations of experience into a causal account of what is really going on in the world. If we are going to try to tie subjects to the concepts we deal with in biology we need to stick to the rules of causal description that are basic to both science and, in fact, everyday life. Strangely, some of the accounts that look the most respectably biological seem to be least good at this.

Key point: like Descartes, we need to consider all the possible options for what a human subject might be and consider which option, if any, can satisfy a subject role.

6.63 Manifestations to: the role of input.

I return to the central issue of relations between dynamics and manifestations here because I believe it is fundamental to the identification of a coherent account for the human subject. In common parlance a human subject is something that has experience of the world, or at least ideas of a world. I think one can reasonably equate that with saying that a subject is something to which the world is manifest in the form of ‘appearances’ that seem to be ‘of’ or at least ‘about’ goings on in the world. In causal terms that seems to be something with an input from the world, defined functionally in terms of having that input. There would seem to be two alternatives to this view. One is that there are no human subjects, perhaps because phenomenal experiences just happen, without needing a subject. The other is that there are subjects with experiences but that the experience is not a manifestation of input but perhaps of some other process, such as output. As I shall explain, my considered judgment is that neither of these alternatives can work.

The task is to identify for a typical phenomenal experience the dynamic goings on that are necessary and sufficient for that (token) occurrence of that experience and preferably immediately associated with it (with no intervening mediating processes). This is the spirit of the concept of a neural correlate of consciousness. It is more or less universally assumed that these goings on are a subset of the goings on that constitute a group of brain cells (neurons or possibly astrocytes). It is also typically assumed that they involve electrical or chemical perturbations that mediate cell communication. If we accept the term subject this pattern of goings on will be it. Some might like to continue to think of a subject as a ‘material thing’, in which case we can probably just say it is the group of cells without too many problems at this stage. Others might feel that the approach is getting too ‘reductionistic’ or ‘tied to objects in space’ and that it should be more ‘functional’. My response to this latter objection is that if functional means causal dynamics, which to me is the only unprejudiced meaning, then that is exactly what I am wanting to propose – to link experience to goings on. If it means something that presumes other distant (and particularly future) goings on or teleology I think the objection is misguided for reasons given already.

The question is then, which goings on in this group of cells will be necessarily and sufficiently associated with the experience. As I see it we should expect the pattern of the experience to be determined by the input to these cells. Such an input could be said to inform those cells, or be manifest to those cells, or indeed to be experienced by those cells, using meanings that are standard to our language of causation. The cells would be a subject. We retain the idea of a subject much as in normal parlance but, as indicated in the opening section, recognise that the boundary needs to be moved in from the body, or even brain, surface. Despite the natural consistency with the sort of causal account we use in both everyday life and science this option seems to be out of favour, but for reasons that seem confused, as I shall discuss shortly.

The next option would seem to be that the experience is most immediately associated with the output from these cells. This does not seem to fit with our ordinary ideas about subjects and experiences. The output of an entity is the way it influences or informs other entities and so it seems appropriate to consider the subjects of such an output to be the next cells in line that we would ordinarily see as experiencing the output. There is another problem. The significance of the outputs in informational terms will be dependent on where the outputs go and the context in which they are received. In theory one could suddenly block the outputs by putting voltage clamps on the outgoings axons. We then have the idea of an experience being associated with some goings on that would have been outputs but in fact are just local perturbations with no further functional significance. A further issue is that we do not seem to have any explanation for the pattern of outputs to be unified in a single experience. Defined just as outputs there is nothing to indicate where they are going or whether they are integrated anywhere further along the line. They may well be integrated, but then one would expect the associated integrated experience to be further down the line as well. In other words, it is easy to think of inputs being integrated but I am not quite sure what it would mean for outputs to be integrated.

In fact, I am doubtful that any specific theoretical models are explicitly based on the idea of experience being associated with outputs. As mentioned before, this might seem to be the case when there is talk of ‘C-fibre firing’ rather than ‘C-fibre stimulation’, but inasmuch as this is thought through I suspect the idea is that firing is not so much output as just ‘being active’.

This brings us to the third option, which is that experience is associated with a pattern of goings on within our group of cells that is not defined either in input or output terms but what might just call ‘activity’ or ‘traffic’. The immediate heuristic problem here is that we now have some goings on dissociated from any role in an ongoing causal chain. Our group of cells has become more of a black box that, through being ‘busy inside’, ‘has an experience’. This seems a bit like a Wimshurst machine that, when its wheels spin fast enough, flashes sparks between its electrical terminals or a cellar party where, when the music hits the right rhythm, everyone suddenly starts jigging. The idea that either in all cases (for panexperientialism) or only in some special cases in brains some goings on lead to those goings on experiencing themselves does not seem to make any contribution to a causal description of the world. Do we want to suggest that a cell undergoing depolarisation is experiencing its own depolarisation? Is it not more appropriate to talk of the next cell along experiencing the effects of the first cell’s depolarisation? When we use the term ‘experience’ in a general inanimate context we think of a house experiencing the force of the wind just as a man might experience the force of the wind. We do not talk of a wind experiencing its own force. At the most basic physical level, if manifest experience corresponds to the aspect of physical interaction that we call something having a determinate value then we have a causal story. If experience does not have a place like this in a causal chain it is difficult to know how to make sense of it.

I prefer to take a panexperientialist approach and will in due course give some details of how I think that can be made plausible (for those in doubt) so I do not see any particular need to ask for some ‘complex activity threshold’ that has to be reached for experience to pop up. For those who feel there should be a threshold I think there is another problem. Brains do not work quite like computers but there are some common principles. One of these is that patterns of information consist of both positive and null, or even, in brains, negative, signals. For computers 0s are as important as 1s. There is ever-increasing evidence for a lot of neurological processes making heavy use of active inhibition. We ought to expect inactivity to be manifest just as much as activity. A ‘determinate value for a position’ should perhaps be couched more in terms of not here, not here, not here, here, not here. Moreover, recent evidence suggests that with propofol-induced anaesthesia there is no general reduction in activity in the cerebral cortex; if anything it is the opposite.

The advocates of models that associate experience with local cellular activity would, no doubt, say that it is not that simple. It is not suggested that the sort of activity one sees in a forest full of fireflies all just flashing on and off is what matters. It is often suggested that what matters is the traffic of messages amongst the cell in the relevant group. This is very much the line taken by Victor Lamme, it seems. A common elaboration of this is the idea that what matters is the presence of repeated feedback or re-entrant activity.

One problem with this is that it begins to break the activity down into inputs and outputs again, if it brings in any causal roles at all. If it is inputs that matter, let us just go with that. Another one is that one tends to think of feedback loops as involved in regulatory or collative activity and our experience of the world seems to be much more a matter of being presented with an inference about what is going on, not a mass of housekeeping signals. Re-entry would seem to be very relevant to keeping certain patterns ‘in frame’ for ongoing attention but what we experience is the frame, not the keeping of it in frame.

Perhaps an even deeper problem for the ‘experience as traffic’ model relates to a general problem that all models face that I will raise again later in the context of an input model. If a group of cells forms a subject, or is the neural substrate of an experience, we want to have some way of understanding what determines the functional boundary for this group. What makes this group associated with experience E and not the next cells along? The problem with the traffic model comes back to the idea that null or negative signals are just as important as positive signals. We cannot expect our group to be just firing cells. To generate an experience with a specific pattern we would expect some cells to be firing and others not. Yet if we allow cells not firing to be in the group, what is different about the cells not in the group? I think it is very problematic to define a locus of experience in terms of traffic. It makes no informational sense. And I think the problem is just as bad for re-entrant signals. If we do not allow any possibility of either firing or not firing we seem to be left with endless cycles unresponsive to changing input. And what about a feedback loop that works on the basis of repeated inhibition of inhibition, with resulting continued firing of one element? As far as I can see, as soon as one tries to apply a traffic based model to real neurobiology one finds it unworkable.

To conclude, as far as I can judge the only workable theoretical models for experience will be those that associate it with input. Other models hit paradoxes and do not do the right causal theoretical work. Why is an input model so rarely invoked? One major issue is the perceived bogeyman of the Cartesian theatre or homunculus and I will address this next.

Key point: just as the whole human being is traditionally seen as a subject by dint of receiving sensations from the world, a model of an inner subject needs to be based on a relationship of receiving signals.

6.64 Homuncular problems – real and false

The term homunculus is used in at least two different contexts. The motor cortex just in front of the sylvian fissure shows a division of labour such that it effectively forms a little map of the body called the homunculus with the leg and tail regions at one end and the mouth at the other end. Nobody has any argument with this. The usage relevant here relates to the idea of an inner subject (or ‘little man’) within the brain that in some sense ‘sees’ the collated signals generated by primary sensory areas in a way analogous to the ‘seeing’ of the ‘outer man’. This idea has been said to lead to an infinite regress. However, as Daniel Dennett has pointed out, it only leads to an infinite regress if you set it up in such a way that it must do. It seems unlikely that anyone has ever wanted to set the idea up in quite this way so there is a strong argument for saying that the infinitely regressive homunculus is a straw man. Models of inner subject that have no tendency to infinite regress might be said not to be homuncular but there is no doubt that there is serious confusion in the literature in this area.

To begin with, as already discussed, the idea of ‘seeing’ even for the outer man is open to suspicion. What we now know is that in a typical situation light is reflected in the outside world towards our eyes, the light is focused on to two retinae and then signals are sent down nerves to a complex collating apparatus involving both visual and parieto-temporal regions and the corpus callosum (to give solidity from binocular input) at least and at the end of that a pattern of signals is generated which signifies the dynamic patterns inferred to operate the outside world. The eyes do not actually ‘see’ anything as far as we know. There is no ‘image’ on the retina, despite the fact that so often this is claimed. (There is only a pattern of light arriving that, if the retinae were a white piece of paper, which it is not, then it would appear to have a picture on it for another human subject viewing it via some clever periscope that did not get in the way of the pupil!) So if there is no outer seer after all then presumably we need an inner seer. So there is no suggestion that models with inner receiving subjects need ‘repeat’ any activity.

Dennett produced a rather poetic, if slightly strangely constructed, phrase in relation to this. ‘Homunculi are only bogey men if they repeat entirely the talents they are rung in to explain. The point made is that if you want to explain an ability like seeing then if you do so by postulating some identical form of ‘inner’ seeing then you are then obliged to give a further explanation. Clearly, in such a case you have not yet solved your problem. However, it is very unclear why this should lead to an infinite regress, unless you suggest that we postulate further inner versions, which nobody does. The fact that the job has not been done first time around in no way stops one from doing it the second time around. If I say that Jack got an poem from Jill because Jill got it from Tim that does not stop me from going on to say that Tim composed it on his iPhone.

I am not sure that we even need to invoke any repeating in a brain but there is no problem with postulating a whole string of transductions if we suggest that each one has a different role in the total collation and inference process. Strangely, Dennett himself denies any ‘second transduction’ but if transduction is converting signals in one form into signals in another then there seems no doubt at all that most sensory information goes through a dozen or so transductions, from electrical to chemical and back and stepwise from uncollated to fully collated. That leaves us with absolutely no problem in saying that once the dozen transductions are finished we are at the subject.

Descartes understood the regress argument well and made it clear that the soul he postulated would not itself collate incoming information in the way that that body had already done. He understood that the question of how dynamic signals coming in from the world, for him through mechanical means, could be manifest to the soul in the form of a narrative about stuff and things could only be answered one way. The ‘interpretation’ of the signals as a story of stuff and things must be an integral part of the immediate interaction of the signals with the soul. No mechanism would account for it. Put differently, the regress buck stops here. We have no understanding how dynamic signals give rise to incommensurable narrative manifestations and no mechanical dynamic (or indeed modern neurophysiological account) would do the job so we just have to accept that this is a brute fact about the relation of dynamics to manifestations, at least for human subjects/souls.

One of the odd things about discussion of the homunculus regress problem is that those who assume its validity are often those whose models most fall foul of it. Those who shun the idea of in inner receiving unit tend to prefer the idea of some widely distributed pattern of cellular activity that somehow forms an ‘inner map’ of what an external dynamic pattern ‘represented’ in experience. However, the is no point in having a distributed map of this sort unless there is a map reader. There is only any point in having a map if you have a collating device of the sort we have in our eyeballs and occipital cortices. That job has already been done. The signs of external dynamic patterns resulting, and manifest in experience, would not be expected to be arranged in anything like a map form. They would not be expected to be distributed over a wide area unless there is some subject whose domain includes that wide area for which the signs form an input.

Key point: the widespread belief that a model of a subject based on an inner receiving structure or homunculus leads to an infinite regress is groundless. Ironically, the real problems of a homuncular model tend to be most serious in those models proposed by those who disown homunculi.

6.65 Everything coming together and real-time experience

Another objection to an inner receiving subject is the popular adage that there is no one place in the brain where everything comes together, so you cannot expect there to be a receiving subject for which everything comes together. This idea reflects Descartes’s original difficulty with finding a unique place for a soul in a brain that seems to have two of everything. The brain stem has some midline structures, but they rather give the impression of being part of an entry and exit portal. The only solitary midline structure otherwise (more or less) is the pineal. The pineal is now regarded as having rather little to do with the mind since it can be removed, if diseased, without any obvious effect on behaviour.

What this objection to a receiving subject conceals, however, is that there is no reason to doubt that there may not be many places where enough signals come together to account for the patterns manifest in our experiences. If we insist on finding just one place then we may have difficulty, but if we allow there to be several the difficulty disappears. Moreover, it has long been recognised that as the brain develops in life tasks may be taken over by one side or other of the brain, despite the fact that both sides probably have the potential to fulfill the task. Thus, in most people who lose the use of their left hemisphere the power of speech is more or less completely lost. However, if the left hemisphere is damaged early in life speech can be acquired. Michael Gazzaniga has suggested that in most people, in addition to controlling speech, the left hemisphere normally has a general controlling function on what is perceived through a functional area he calls the ‘interpreter’. On a shorter timescale we also have good evidence from binocular rivalry that one set of sensory pathways can suppress inputs in other pathways. So if we want to stick to the idea of a single human subject in a brain then that is quite compatible with there being several equally good candidates, with some sort of short or long term competition between them. If we are prepared to allow more than one subject per brain then the whole problem disappears.

If we consider the question from the other side and ask if we need to postulate somewhere where everything comes together then I think we have to say yes. If we are to respond to a complex pattern in a way that depends on all the relationships between all the elements of that pattern, as it does for a face or a fruit that is suitable to eat, then somewhere all the elements have to be integrated. That integration need not be in one place in real time, however. In computers this sort of integration is achieved piecemeal by integrating two signals at a time and putting the results into some sort of store to be retrieved according to a complex set of rules which allow an output to be generated that is specific for the total pattern, with all its implicit internal relations. This procedure does not require ‘everything to come together’ in one place at one time. It does nevertheless require that everything be brought together with everything else in a computational sense.

This raises the question of whether or not we think that the apparent unification of all the elements that seem to be combined in an instant of experience must reflect some real-time dynamics or whether we could conceive of an experience arising out of a chain of piecemeal integrations. I doubt that this is something we can provide definitive and useful arguments about. Nevertheless, it is interesting to see that, as far as I can tell, there are no proposed models for experience that invoke the phenomenal unification of a series of piecemeal integrations occurring over time. We hit the intuitive objection epitomised by thought experiments such as those known as the Chinese Nation or the Chinese Room. The Chinese Nation thought experiment asks whether anything has an experience if individual members of the Chinese Nation perform the same tasks in functional terms as those performed by cells in the brain through making telephone calls to each other. In theory the Chinese Nation could be very laid back about this and take twenty years to simulate the single experience of an orange. DO we think that experience can arise out of events strung out in time, even over a period of milliseconds?

Maybe we can, but the computer situation seems to be worse still in that the same pathways are going to be used for a sequence of serial integrations. Can we get an experience of an orange associated with a processor that handles a thousand different elements of the orange one by one, with each last element having been ‘overwritten’ by the next. This seems hard to believe in.

This is probably why most models that do not seem to implicate any specific receiving element for an experience tend not to suggest that the information to be experienced is shunted serially through one site. In contrast, they tend to suggest that the information is spread out over a network of active elements. The problem I see with this is that it does not actually give us a situation where all the elements are brought together for computational purposes at all. It does not do it either way. It could be argued that serial integration is going on somewhere else but then why should this spread out pattern of activity be associated with the phenomenal integration in an experience that seems to need some sort of dynamic basis.

This difficulty is frequently recognised as posing a ‘phenomenal binding problem’. Yet it often does not seem to be taken as seriously as it needs to be. A very popular way of trying to deal with the problem is to suggest that if activity is synchronised it will in some way be unified. Apparently this idea started as a speculation by Sherrington. It has been repeatedly resuscitated to try to give functional relevance to electroencephalographic findings, despite the fact that it does not seem capable of explaining quite what some people want it to explain. It is easy to see why synchronisation might improve the efficiency of integration of signals when they do come together but to my mind the synchronisation of signals at points where they are not actually being integrated is of no relevance to unification or binding whatever. If we want to meet a friend at a precise time it is irrelevant whether or not we synchronise the times we leave our houses. The time taken to get to the meeting might turn out to be identical, but there is no need even to consider it because the only thing that matters is that our meeting is synchronised. That, and only that, would provide the sort of binding that you might get from asking a passer by to photograph both parties together.

As indicated in the last section this sort of model, designed to avoid the illusory homunculus regress fallacy, seems to fall foul of the true homunculus problem: that of postulating a distributed pattern of signals that really does need another inner man to bring together. As for the mistaken idea that there is an image in a retina, it is a mistake to think that just because a third party can think of a distributed pattern of signals in a brain as being like a map or image, that is only because they are considering it is ‘viewed’ by a fly on the wall third party human knowing machine plus knower.

It might be argued that the mistake is to try to tie experience to a single point in time and that in a network of re-entrant circuits the experience emerges out of repeated cycles of activity over perhaps half a second. As I shall come to later, I think it is entirely reasonable, and probably necessary, to think of experience as arising over a finite time span, rather than a time point. However, this does not solve the problems raised above. The relationship of the experience to the goings on remains one divorced from any causal dynamic explanation. It is in a sense totally ‘unphysicalist’ in that it has no point of contact with normal scientific theory.

Key point: all the elements of incoming patterns on which we base decisions must ‘come together’ either serially or in real time for purely computational reasons. There are strong reasons for thinking that for the human subject this coming together is in real time.

6.66 ‘Orthodox’ views of the human subject

The man in the street probably tends to assume that science has an orthodox position on any major question about what is going on in the world, even if it is one of ignorance. The orthodox position on matter is that something like 70-95% of it is totally unaccounted for by present theories. That is still an orthodox position. With regard to the human subject it is doubtful that one can even identify an orthodox position. Generally speaking neuroscience textbooks say nothng about it at all. Even books written by neuroscientists that deal specifically with the nature of human experience tend to skirt around the issue, and in any case, these are probably considered on the fringes of orthodoxy by most of the neuroscience community.

The lack of an orthodox view in an area of active debate and research may not be a bad thing. It forces one to consider theories on their merits rather than on their reputations. Nevertheless, it makes it more difficult to know where to start and in what order to discuss current ideas. I shall take the approach of discussing the broad anatomical issues about where subjects might reside in this section and then turn to some popular individual theories one by one.

It is probably easiest to start by describing those parts of the brain that seem unlikely to be locations for subjects. Certain areas, including the cerebellum, and the outflow path from primary motor cortex to brainstem look to be more or less exclusively involved in output rather than experience. It is usually also assumed that the major inflow pathways through brainstem, including structures like the geniculate bodies and colliculi, are not going to be sites for subjects. They appear to handle data from only one sensory modality each and in a form that has not yet been collated to produce the basic building blocks of experience like shapes and phonemes.

The major input and output tracts take up much of the brainstem but there are also groups of cells, or nuclei, that are involved in the change from sleeping to waking and general arousal. The most widespread view is that these areas do not themselves support the experiences of waking life, but merely provide on-off switches that regulate the experience of other parts of the brain. This is not a universal view. Panksepp has argued for the idea that basic emotional experience at least may reside in the brainstem. The argument is that brainstem structures are very old in evolutionary terms and if they were the only structures available for more primitive animals to experience with then may be they are still used for experiencing in us. The main argument against the idea of brainstem based experienced seems to be that much more complex activities relating to sensory modalities are going on in the cortex. I am not sure that this is a relevant argument, because I am not sure that complexity of activity is what we are looking for. Few people would support the idea that the brain stem is the major locus of experience, but the fact that one or two might indicates just how open the issue is.

Moving ‘in’ from the brainstem there are a number of areas rich in cell bodies (grey matter), paired left and right, that come under the umbrella title of basal ganglia. These structures are relatively infrequently considered as sites for supporting experience but there are some quite strong arguments for considering them. Whereas almost any amount of cerebral cortex can be damaged without loss of behaviour suggesting the continued presence of some sort of subject, damage to the basal ganglia, and in particular the thalamus, tends to leave an individual in a vegetative state. The same is true of the brainstem but the thalamus is a relatively large structure with a complex network of cells that connects directly with most other brain areas, including the cerebral cortex. Whatever information might be available to cortical areas would seem to be readily relayed to thalamus. When activity in widely separated parts of the cortex is synchronised there is evidence for this being due to control by activity in the thalamus acting as a ‘third driver’.

There are probably two main reasons why the thalamus is not the most popular choice as a locus of experience. The much greater complexity and more recent evolutionary development of the cortex has made it attractive as a site that might support a form of experience special to humans. This has been a significant motivation in the past but may be less so now, since people are now more generally ready to accept that many vertebrate classes may have experiences similar to ours, if less sophisticated. The other problem with the thalamus is that it comes in two rather poorly interconnected halves. Although the cortex is also divided down the middle it looks rather more like a single massively interconnected network than the thalamus does.

There are other specialised grey matter structures in the central parts of the brain that might also be candidates for hosting experience but they tend not to be given special consideration. The mammillary bodies and amygdala are involved in emotion but are rarely if ever posited as experience loci. A number of other cell rich areas exists close to the thalamus, of which the claustrum may be worth special mention. Koch has pointed out that the claustrum looks to be well placed to be a site of integration of multimodal sensory information. In other words it might support the sort of unification or binding that we seem to need to explain experiences that apparently fuse information from different senses. As for other areas the question is whether the claustrum is merely a site that makes it possible for such experiences to be had somewhere else or whether it is itself a locus of experience.

The cerebral cortex is the largest mass of grey matter in the human brain and there is often a tacit assumption that it is the locus of experience, for the reasons given earlier. Very often it is not clear whether such an assumption treats the whole cortex as subject or is more restrictive. Certain areas seem less likely to support multimodal experiences. I have already mentioned the primary motor cortex, which is essentially an exit route. The primary areas of visual cortex, such as V1, are also often considered not to support experience, for reasons relating to specific experimental data. On its own V1 does not seem to achieve sufficient collation of incoming signals to support experience of objects. If V1 is intact but other visual cortical areas are not then visual experience is generally absent. The arguments here are always subject to caveats but there does seem to be quite a good case for considering V1 not to have access to patterns sufficiently well collated to support the sort of experience we discuss.

A more general point may make this discussion redundant, and that is that all of the visual cortex is limited to processing visual input and therefore is unlikely to support multimodal experience. This argument would apply to other primary sensory cortices. However, as I shall return to later, things may not be that simple. We know that the visual cortex infers the presence of objects partly in the context of information gathered by other sensory modalities. It is not the primary input route for auditory information but auditory information is sent to the visual cortex it seems. This gives us a reason for not excluding sensory cortices entirely from the locus of experience. Moreover, some theoretical models allow for the existence of non-multimodal experience as well as multimodal experience and the former might well occur in sensory cortices.

A fairly popular theme is that experience does not really get going until inputs to sensory cortices have been collated and the resulting signals have been sent on, via parietal and temporal areas that add further context, to the prefrontal region. This is in part motivated by the fact that when people are self-consciously thinking about themselves and their relation to the world there tend to be specific patterns of activity in the prefrontal regions. Since the prefrontal regions do not seem to be involved in primary sensory or motor functions, or indeed memory storage or shape perception (temporal and parietal functions) it seems to make sense that they should be involved in some ‘higher level’ of integration associated with a sense of self, and indeed a sense of the world in general.

There are, nevertheless, reasons to be cautious about the simple idea that the prefrontal region is the experiencing subject. Removal of prefrontal brain tissue does not seem to remove a sense of experience. Rather, it seems just to make people change their personality in rather subtle ways, such as developing a flat affect or have difficulty in making decisions that have emotive connotations. This might be taken as an indication that the prefrontal regions are involved in unconscious integration of sensory and emotional material so that the true human subject, maybe in the thalamus or further back in the cortex, can feel all aspects of their world. There are reasons to think that medial and lateral prefrontal areas need to be considered differently in this context. The medial areas may be more involved in more abstract aspects of a concept of self in the world whereas lateral areas might be better candidates for a multimodal receiving area that gets the final pattern of inferences from everywhere else. A few people have developed relatively specific theories about these functions but in the end we come back to the fact that everything is up for grabs because nobody is really clear what criteria should be used for identifying a subject domain, or even whether that is a valid concept.

On this background I shall consider some popular theoretical approaches to see how they fare in the context of previous discussion.

Key orthodox accounts of inner human subjects, in as much as they are explicitly advocated, may have various anatomical and functional merits but in general are not considered in a sufficiently rigorous causal framework.

6.67 Global workspace theory

Global workspace theory (GWT) has been developed by Bernard Baars and provides a good illustration of how some of the issues about the basis of experience have been handled in recent times.

The central idea of GWT is that at some point signals between neurons have been sufficiently collated and weeded out on the basis of relevance, salience or whatever, to be suited to enter a set of pathways that carry the information in the field of attention in experience. There is room for debate about the relation between a field of attention and what is phenomenally experienced, but the assumption here is that they are equivalent. The set of pathways carrying the information selected for current experience is considered as forming a wide ranging ‘broadcasting ‘ system that can feed the relevant information to many areas of cortex for parallel processing. This broadcasting system forms the ‘global workspace’ and signals broadcast in the workspace are those that are ‘conscious’, in a way that signals in other pathways are not.

Baars has used an analogy that seems initially a bit like a Cartesian theatre but has an important difference in interpretation. Baars likens the signals in the workspace to actors in a spotlight on a stage, whose actions are broadcast to an audience of a large number of computational units in relative darkness.

An attractive aspect of this theory is that it makes explicit a distinction between signals involved in preliminary collation of inputs and signals that are ready for involvement in the sort of computations that we associate with conscious thought. It seems reasonable to suggest that the brain is designed to make this distinction and only ‘broadcast’ material to such computational apparatus once it has been fully collated and selected as relevant. It also has the plausible feature that once information has reached this stage it is sent not just to one processing unit but to many processing units to allow for rapid parallel processing.

Baars’s theory, at least in its original form, does not stipulate which anatomical regions are involved in the global workspace. The implication is that it is cortical and probably involving areas other than primary sensory cortex, but this is left open.

The difficulty I see with Baars’s approach is in how one interprets the suggestion that signals in the global workspace are conscious. One way of looking at it is that Baars is not giving us an account of experience so much as an account of those neural signals that encode the material we report as being in our field of attention. This can be given an entirely third party behaviouristic interpretation. However, if this is all the theory is doing it seems plausible but not very surprising or powerful in terms of predictive power. It does, however, have the interesting implication that ‘broadcasting’ in the workspace means that the same pattern of information is being sent to many sites. If some of the pattern was sent to one site and some more to another this would not be ‘broadcasting’. It certainly would not fit the actors on the stage analogy. This sending out of ‘multiple copies’ of a pattern is attractive in the neurobiological context since we know that most signals sent out from cells in the brain go to thousands of sites.

Yet, an interpretation of GWT that simply makes it a behaviouristic definition of the information we can report does not seem to get much of a grip on experience itself. Is Baars suggesting that only the signals sent out in global workspace are associated with phenomenal experience and if so why should they be and is it something we really need to postulate? If computations at the level of language are tied in to the global workspace signals then it is easy enough to suggest that we only report global workspace signals as being in our experience because these are the ones our language faculty handle. This might not explain why our non-linguistic behaviour seems to tie in to experience as reported but it is easy enough to cast a rather wider link between the workspace and the sort of behaviour we interpret as linked to experience (i.e. not just blinking or not falling over). Moreover, even if we take a behaviouristic view we need to have some idea what the neurological basis of a reported experience is, in terms of whether it is an input to some structure, an output, or just some ‘activity’.

The analogy of the actor on the stage seems to suggest that Baars is wanting to imply rather more, and indeed to give an explanation for phenomenal experience. The suggestion is that the signals in global workspace, like the actions of the actor are somehow ‘manifest’ in a way that others are not. Here we run into the issues raised in the previous discussions of manifestations being to something. Clearly we do not want to suggest that the signals in workspace are manifest to some third party. Nor does it seem to make much sense for them to be manifest to themselves. The actor may be shouting and making violent gestures in the limelight but with his own eyes closed. If he is in a spotlight he may well not be able to see much even with eyes open. Surely the manifestations of these actor signals will be the audience, or the extended domain of cortical areas that receive the signals. They may be in the dark in the sense of not being able to read their programmes but the assumption is that they can see the actor very clearly.

In summary, the interpretation that seems to me logical for GWT is that there is a particular pattern of signals in a brain at any one time that is sent out along certain pathways that branch to many sites within the cortex, such that each site receives a copy of every element of the pattern and this pattern is the one we report as being in ‘our’ experience. The implication is that what is being reported is a pattern that is not manifest to any one receiving unit but rather to many receiving units in parallel. Thus although we report this pattern as if it were received by a single subject the model suggests that it is being received by a large number of ‘subjects’. This conclusion may seem highly problematic and even absurd, but as I shall discuss later, it seems to me highly plausible and probably unavoidable.

The only difficulty with this interpretation is that Baars himself has indicated that he is not happy with it. He seems to want to hold to the idea that it is the signals being sent that have a property of ‘being conscious’ rather than units receiving the signals being conscious of them. I am not clear what ‘being conscious’ means here, other than in the behaviouristic sense of correlating with reports of being experienced.

Key point: global workspace theory provides a useful general functional model for a limited domain of the brain dealing with signals that are associated with manifest experience. However, in a sense the actors on a stage and audience metaphor seems to be the wrong way around.

6.68 Integrated information theory

Giulio Tononi, together with David Balduzzi, has proposed an approach to the domain of experience in the human brain based on the concept of integration of information. The central tenet in this idea is that experience arises out of networks and that the richness of experience can be related quantitatively to the amount of information being integrated in a limiting subnetwork or complex.

The theory appears to be designed at least in part as a way of contrasting the computational processes in brains with those in man made digital computers. Whereas computers are engaged in a long series of simple serial manipulations based on very limited integration brains engage in very complex integrational events in real time. In a computer the output of each computational event can be seen as dependent on the relation of one signal to one other signal. In the brain it may be dependent on the relation of a signal to a large number of other signals and the relations between those signals – or just to the complex interrelations of a large number of signals. This has a combinatorial effect on the complexity or richness of the information being integrated. In a computer two signals that can each be 0 or 1 can exist in four permutations 00,01,10,11. However, if we have ten binary signals the permutations jump to a thousand.

Tononi suggests that the richness of our conscious experience reflects the number of signals being co-integrated in what is effectively a single computational step. The motivation seems to be in some ways similar to that of Baars, to define a relatively limited number of signals that form a pattern that is the field of attention or experience, amongst a whole lot of other signals used for other purposes. We do not want to consider the network that hosts experience as being the entire activity of the brain. Apart from anything, the activity at several trillion synapses looks far too much. We want to find some sort of ‘informational bottleneck’ that functions as a single integrational step that would carry the pattern we are interested in.

Again, I see a lot of sense in this general framework, even if it is somewhat abstract and divorced from specific neuroanatomical considerations. The problem that I see for Tononi is in justifying what is to be considered a single integrational event. If the brain as a whole is a network, what defines subnetworks as complexes such that the signals being integrated within them are isolated in some way from everything else? As far as I can see, in terms of information theory there is no way to do this. The size of subnetwork taken under consideration is arbitrary. The only solution I can see to this problem is to consider the ‘complex’ as a single biological integration unit, which is a neuron. There are potentially ways of making the unit a larger unit of cells that functions as if it had a single shared pattern of input, and suggestions along these lines have been made. However, they go beyond what is generally accepted in the field of neuroscience that deals specifically with electrical signal integration. Moreover, it is not clear what their advantages would really be.

The problem, as I see it, comes back to the issue of richness in real-time. In a computer the real-time integration processes are trivial. If complex patterns are being subjected to computational analysis then these patterns are always to be found stored ‘off-line’ in the equivalent of a Turing machine tape. Most people find it implausible that such a situation should have an experience of the total pattern at any point in time. Tononi is proposing a subnetwork or complex that is carrying a pattern of information in real-time that is rich enough to encode a human experience. The difficulty is that a pattern of signals only functions as information in the context of being an input to some structure that is ‘informed’. The activities of cells in a network are not related in informational terms until the signals produced have actually been integrated; that is to say taken part in a computational process the output of which is dependent on all elements of the pattern. Some of the active cells may generate activatory neurotransmitters at their axonal synapses but others may generate inhibitory transmitters. Branches from some axons may go to different banks of receiving cells from others, such that the whole pattern of activity influences some parts of the brain and only part of the pattern affects other parts.

Thus in the network there is still no real-time informational pattern to support an experience. The activity of an element of the net may contribute to a total integrational computation when the signal has passed on to some other point in the net, but by this time the original element may have quite a different activity status. The real-time problem has not been addressed.

Put differently, the extra combinatorial richness that Tononi wants to achieve through the integration of many elements is just as much available to the plodding computer or Turing machine. After a sufficient number of serial events an outcome can be obtained which is just as much dependent on a combinatorial richness. The network does better in real-time only to the extent that individual neurons have rich input patterns that are integrated simultaneously.

Key point: integrated information theory also provides a useful analysis of the computational function of a subject, but again appears to attribute subject status at the wrong level.

6.69 Neuroanatomical workspace models

The models mentioned so far are based on abstract concepts with little or no attempt to relate these to instantiation in a particular neuroanatomical format. Changeux and Dehaene have taken the different approach of trying to build a workspace style model with specific reference to brain structure.

An interesting feature of this model is the suggestion that what I have called the collational process in the sensory pathways exists in a hierarchical form. This is highly plausible, not only for general theoretical reasons but also in relation to the work of Hubel and Weisel and others which suggests a sequence of collative steps converting single input signals first to signals for lines or directions of movement and then to signals for shapes and then signals for objects of specific kinds or related to specific episodes. At the top of the hierarchy might be signals for narrative concepts in the context of a concept of a self in a world.

Such a hierarchy is generally assumed to exist and the neuroanatomy of the various levels is quite well established. Identification of simple lines occurs in primary visual cortex, with more complex shapes being identified in other visual areas. Objects appear to be indentified in the parietal cortex and relationships to time and memory in the temporal cortex. Concepts dealing with the self and emotional responses are thought to occur in the prefrontal areas.

The simplest model to propose within this anatomical framework might be one in which cells feed signals forward in a simple sequence through the hierarchies, with first hierarchy cells in primary sensory cortices, second hierarchy cells one step further forward and so on until maybe sixth hierarchy cells are found in the prefrontal regions. However, there are good reasons for thinking that this is too simple. We know that even the most basic visual sensations are dependent on feedback signals from areas involved in recognition of specific objects. We can be fooled into feeling ‘raw’ sensations that seem to be incorrigibly ours yet located in an artificial rubber hand. Moreover, there is a lot of historical and recent evidence suggesting that different regions of cortex can perform the same functions, particularly if the area that normally seems to have a dominant role in such functions is damaged. I do not intend to review this in detail, but simply to indicate that there are good reasons for thinking that any hierarchical functions of cells are not tightly confined to specific cortical areas but are more spread out.

Changeux and Dehaene make what seems a very plausible suggestion, in which cells of different hierarchies are present more or less throughout the cortex but that the posterior primary sensory regions contain a much higher proportion of cells of low hierarchies and the frontal areas contain a higher proportion of cells of higher hierarchies. Thus the separation of hierarchies would be statistical rather than categorical. It may also be suggested that links between hierarchies follow statistical rules in which most connections are between adjacent hierarchies but an increasingly smaller number connect more widely separated hierarchies.

This is another type of model that I think has some major attractions. It could help to explain why our experiences seem to contain both high level and low-level patterns of data. It provides a way of explaining why, if prefrontal areas are principally involved in ideas about a self in a world those ideas do not disappear completely when prefrontal areas sustain major trauma or surgical procedures. The implied statistical rules for connectivity would fit with a system that uses selection rather than instruction for recognition purposes, like the immune system. If connections are programmed in part at random then you can get a very wide range of options from a fairly simple genetic programme. Biological structures frequently seem to work on the basis of broad rules of this sort. However, they also often show exceptions to the rules as well. Thus it would not be surprising if it turned out that first hierarchy cells only connected to second hierarchy, with the statistical rules only applying to higher levels. Such a situation might explain why the signals generated in visual cortex V1 do genuinely seem to be outside the range of manifest experience.

A statistically distributed hierarchical network model of this sort does, therefore, have several advantages. However, it is, as before, unclear exactly where experience is to be located. An interesting potential of this model is that it could provide a basis for what Steven Goldberg has called ‘Jonah Minds’. A Jonah Mind is his term for a subject of experience within a brain that is not the subject of what we customarily talk of as ‘our experience’, or what he calls ‘customary consciousness’. Goldberg raises the possibility that there are subjects of lower level experience (which might correspond to lower hierarchies) that we do not talk about because they are not connected up to our linguistic centres in the way that ‘customary’ subjects are.

This is not, however, something that I think Changeux and Dehaene had in mind. In their model customary experience arises out of the top hierarchy of fully multimodally integrated and self-related signals. No other subjects are envisaged. Their model is similar to that Baars and Tononi at least in that it attempts to identify a subnetwork that is associated with experience. It goes some way towards explaining some of the anatomical paradoxes about the locus of the subject. It makes the subject widely distributed, while still being restricted.

There are still, nevertheless, puzzles relating to models of this sort. It is often suggested that experience arises out of the complexity of the dynamics of a network of computational units. This is often considered in terms of an ‘emergent’ property of a complex system. What is worrying here is that it seems that removal of almost any, and quite large parts of the anterior cortex produces no catastrophic failure of such a system. In man made models of complex systems relatively minor changes often have catastrophic effects on behaviour. In fact the idea of a complex system to some extent depends on the premise that unless you have the whole system you will not get the special functional property that is said to emerge. If large parts can be removed then it is hard to justify the idea that there is a well-defined system at all. It seems that half a system is just as much a system. If the components of the system have different roles which are tightly interdependent in a way that is necessary for the special emergent effect then we would expect removal of part of the system to lead to a severe failure.

A way around this is to suggest that the ‘system’ has ‘holographic’ structure in the sense that every part of the system carries a copy, even if a degraded one, of all the information in the whole system. If you remove parts then you do not lose the content to be experienced, you just lose precision. That may be a reasonable way to look at it, but one could also argue that the concept of a defined system has become redundant. A holographic structure does not need any defined boundaries. It can equally be thought of as a large set of smaller systems. An analogy is a Fresnel lens. A Fresnel lens works the same way as a normal lens but has been cut into a set of concentric rings. This allows the same optical work to be done with a set of rings all of a small similar depth rather than needing the whole bulk of a lens with a single continuous curved surface. (Fresnel lenses are useful for very large lenses such as those in lighthouses, which if normal lenses would be hugely heavy.) The Fresnel lens can be considered a single system or lots of concentric systems and you can take any number of the rings out (except the last one) and still get the desired effect. That suggests to me that there is no emergent property of the whole lens as system. The optical property of interest exists in the component parts as well.

This analysis leads us back to a consideration that arose with Baars’s workspace theory. It seems not so much that the brain has one site of integration suited to being the subject of experience but rather that there are likely to be many such sites receiving the same informational content. That is of course the point made earlier in response to the dogma that there is no one place where everything comes together, we still have the option of having many places where everything comes together.

Key point: neuroanatomical workspace theory begins to suggest a sophisticated way of dealing with the relationship between localisation and distribution of function within the cortex. Interpretation in terms of a single subject remains problematic.

6.70 What would the output role of customary subjects be?

I have so far focused on trying to find an account of some biological structure, or domain of goings on, that might be something to which experience is manifest. I have argued that what is manifest to that ‘subject’ ought to be its dynamic input, since something being manifest to itself seems to have no causal dynamic role and we would not expect an output to be manifest to the outputting structure.

What I have not touched on is what we think the role of the output of such a structure might be. This is one reason why I have tended to keep away from discussion of ‘consciousness’ here, since for some people consciousness implies something about output as well as experiential input. In medical terms, consciousness is taken just to mean having experience. The medical and legal professions recognise the idea of someone being fully conscious but showing no outward signs of being so (for instances in cases of inadequate dosages of anaesthesia with adequate dosages of muscle paralysing agent.)

The naïve presumption seems to be that whatever is the experiencing subject is whatever has human will, or whatever ‘controls’ behaviour. I agree that there is a strong case for thinking that the subject lies somewhere along a causal pathway that leads to overt behaviour in normal people, but exactly where on that path it lies is far from obvious. Bernard Libet’s experiments on people judging when they have made a decision to act show that there is evidence of decision-making before the subject is aware of this. This came as a surprise to many but it probably should not have. If a decision is an output from an input then it seems necessary that whatever generated that output-qua-decision could only know of it later, through feedback.

The other problem about postulating that a single human subject is the sole determinant of ‘willed behaviour’ is that it is quite difficult to see how a unit with a single integrated input pattern could determine all of the aspects of behaviour with signals in real-time. It is easier to see such a unit simply signaling ‘yes’ or ‘no’ but decisions are more than just yes and no, they include the choice of whatever it is these are applied to. Most behaviour looks like a choice amongst a vast number of options. A decision to have a beer is often a decision not to have twenty other drinks on offer. We think of our decision making or will as more than just saying ‘yes’ to ‘will it be a beer then?’. We also have the problem already discussed that we think it may be unlikely that there is any one place where everything comes together. And if all the relevant information that determines a choice like ‘that looks the best pear’ is not integrated either in real time or through a serial mechanism then no rational choice can be made.

If a human subject is not the single determinant of ‘willed behaviour’ then what more limited role might it play? One possibility that we have to consider is that any given subject is one of many that together generate a pattern of outputs that determine behaviour. This may seem odd, yet we readily accept that the brain contains vast numbers of computational units which co-determine behaviour in a highly coordinated fashion. I will have more to say about this in due course.

The second possibility, of a rather different sort, is that a subject may play a more delayed role in behaviour than our intuition makes us think. If a subject generates an output and in the next cycle of processing receives information about that output having been made, and as a further result generates another output that again comes back as part of the next cycle of input we have a mechanism in which the subject is an integral part of the causal mechanism, despite the fact that its awareness always relates to a prior decision. There are reasons, given by Neil Levy, for thinking that this is the only plausible way in which a subject can play a causal role. It does not deny the existence of a conscious subject determining behaviour, it just corrects an error in an intuitive pseudodynamic idea about how that could happen.

Such a delayed role still suffers from the problems of attributing all causality to a single subject and the caveats about needing several places where things come together still applies. However, if we go back to Martin’s view that causality is as complex as the weather we may want to suggest that any decision actually arises through the combined effects of a very large number of previous brain events and that any given subject may play a unique role in any one of these.

An example of such a situation would be if the human subjects that experience in the way we discuss as ‘our experience’ (Goldberg’s customary consciousness) are in fact situated in pathways that are involved in laying down memory and perhaps episodic memory in particular. In this situation it is not that a subject clinches a decision with a yes or no, but perhaps rather that it makes a yes or no decision about laying down a new episode in memory by attaching a ‘tag’ through the sort of loop mechanism discussed under the section on meaning. This subject will then have a causal role in any future decisions that draw on the memory of that episode. This seems to relegate subjects to a sort of librarian role that is at odds with the intuitive idea of a human subject as having the decisive power of Alexander the Great or Winston Churchill. Many people find that sort of role in conflict with their intuitions about themselves but that should not divert us from trying to find out what is really going on and what role human subjects do really have in that.

One aspect of experience that might seem to favour such a role in episodic memory is its apparent relative paucity (at least according to some) in relation to our intuitions. Colin Blakemore has suggested that experience may be rich but it is quite a limited richness, with perhaps fifty bits of information. I think this is too small but it is a guesstimate of the rather small number of aspects of a scene remembered as having been seen in a single episode. Individuals shown pictures briefly tend only to be able to recall a few aspects of what they depict. This paucity is challenged by Lamme but I think it is reasonable to suggest that manifest experience contains fewer details than the whole nervous system is being provided with at any one time. Why do we want these fewer details to be manifest? One answer is that they are the relevant ones for current decision making and that option still stands. Another option is that they are those details that will be tagged as belonging to an episode for memory purposes.

I will leave any further discussion of the role of output from ‘customary subjects’ at this stage but will return to it after having developed some rather different arguments about the extent of such subjects.

Key point: the intuitive presumption that the human subject of experience has as its output ‘human will’ needs to be treated with scepticism. What we customarily call experience may have a much more indirect role in behaviour.

6.71 The arguments for multiple subjects

I have raised several arguments in relation to the possibility of there being multiple subjects in a brain. At this point it may be useful to bring these together to show just how substantial they seem to be.

Firstly, there is the point that however much we may be convinced of the unity of phenomenal experience in terms of all the elements in a single experience being somehow bound together or compresent this is a quite separate question from that of the number of experiences being manifest at one time in a brain. Moreover, multiplicity of experience does not imply any possible conflict if the content of all the experiences is derived from the same input and therefore consonant.

Secondly, there is the point that if there are many simultaneous experiences in a brain, each had by a different subject, then this would not give rise to any sense of multiplicity, since the content of the experience need not include any inferences about there being such multiplicity. Indeed it is difficult to see how it could, except at the sort of level of metaphysical argument given here.

Thirdly there is the general agreement that there does not seem to be any single privileged site where integration of input would occur in some special way different from that for many other sites. This does not of course demand that there be no such privileged site. There might be some competitive or collateral inhibition process that ensured that at any one time only one input was integrated in the necessary special way. Nevertheless, the apparent multiplicity of similar sites, if only in terms of the duplication of structures right and left, does mean that some special mechanism would need to be invoked, and it is unclear what role this would play in causal terms.

Fourthly, there are reasons for thinking that the neural domain of a subject should be quite small. Blakemore suggests that the manifest input should be only fifty bits. Even if we increase that to five thousand, if we consider each bit to require a synapse this is still minute in comparison to the trillions of synaptic connections in the entire brain. It might be argued that each element or bit is input through thousands of synapses but this does not seem to fit with what we know of neural connections. We do not expect a thousand synapses all mediating the same piece of information from the same origin contributing to a single integrational event. We expect integration to involve signals from diverse sources as in signals from a series of retinal receptors being integrated by a cell that responds to the presence of a line. We know that most signals sent down axons go via branches to thousands of sites, but each of these sites is a separate locus of integration. It is not as if all these branches end up adding to the same integrational event. Multiplicity is only of value if it provides some sort of independent corroboration, not if it simply repeats the same signal.

If the anatomical domain of a subject is small then it will be vulnerable if single. Everything we know about brain damage is that it tends not to ablate reportable experience until it is major or affects deep seated structures that appear to be involved in arousal rather than complex pattern encoding. This very much suggests that there are many domains supporting subjects and that loss of even a significant proportion of these need not ablate reportable experience.

Fifthly, if, as indicated, most signals in the brain are sent for integration to at least a thousand places then the prima facie case would seem to be that subjects are likely to occur in banks of thousands, or at least hundreds. To those who have worked within a community committed to a single subject this idea may seem bizarre. However, if one takes away that simple presupposition the case seems to be pretty clear that this is the default hypothesis that needs to be disproven before being rejected.

Sixthly, there is the problem that if there is only one subject receiving an input that includes the entire pattern (e.g. a face) on which an output depends the total subject output can, in the context of neural computation with outputs being either firing or not firing ,only be yes or no (or perhaps a quantitative response along one degree of freedom indicated by firing rate). That is to say it cannot be ‘that’s Jane’ but only ‘yes’ in response to ‘is that Jane”. That seems unsatisfactory since we would expect a computational unit that answers to ‘is that Jane’ to be one of a large bank of parallel units that respond to ‘is that Sue, Nora, Ellen, etc. ’

The problem here is not discussed in the literature as far as I can see, but it seems to be a very real one. It might be argued that a subject that receives all the information needed for Jane’s face could produce a complex, and thus very specific out put that meant ‘that’s Jane’. If the output derived from twenty cells in a subject network it could take one of a million forms. However, there is a difficulty here, which can perhaps be illustrated by Tononi’s concept of integrated information content or f. For each of the twenty outputs from our net there must be somewhere in the causal history a causal nexus in which the pattern of elements f is integrated either in real-time or through serial interactions linked by a ‘Turing tape’. In a computational system like a brain or a computer the output from a causal nexus only has one degree of freedom. That seems to be both a requirement of the physics of the system and of an efficient computational device. If we want a net to have an output with twenty degrees of freedom (signal or no signal from twenty cells) then we need to have the elements f being integrated at twenty different junctures. These junctures might, in Turing machine terms, be only one move of the tape apart but they are separate junctures. The integration has to occur twenty times in this case because each time will effectively compare the pattern of f with a different pattern in memory – one for each person to recognise.

The point of this argument is that if we think in terms of Tononi’s bottleneck of integration that gives the richness of experience we really have to propose that there are twenty such bottlenecks. There is no single bottleneck or complex that does the work. There may be an all encompassing network within which all the relevant events go on but that does not mean we can ascribe subject status or experience to it, any more than we can ascribe experience to a choir singing a specific piece of music. In the choir every singer has a copy of the score. What they sing may be different because they have been assigned a part. The effect will be a specific harmony but that does not make the choir a subject.

Seventhly, we have the historical evidence indicating that large areas of cortex can be removed without apparent ablation of the subject and also that division of the cortex by corpus callosum section appears to yield two semiautonomous units that under controlled conditions behave very much like two subjects. The idea of a single local subject is hard to sustain in the face of such evidence. The standard response has been to suggest a widely distributed subject but there is the perfectly viable alternative that there is massive redundancy of subjecthood in brains. When areas are damaged there are still many functioning subjects. When the brain is divided large numbers of subjects on one side contribute to one pattern of behaviour and similar numbers on the other side contribute to another pattern. This explanation does not run into the problem of why major damage to a distributed system with some special emergent property does not lead to catastrophic failure of that system.

Eighthly, I would add that proposing multiple subjects dissolves the paradox raised by David Chalmers referred to as ‘fading qualia’. The puzzle for a single network based subject is that it is hard to see what would happen to experience if increasing numbers of subunits were replaced by silicon chips that produced the same output from a given input. If there are multiple subjects then replacement of part or all of one will still leave experience unchanged in others.

Key point: there are a number of reasons why it makes more sense to think that there are multiple subjects in a human brain than just one.

6.72 The subject as integrating unit

Despite being organised in a different way, many, if not all, of the arguments presented so far are to be found in a wide range of texts going back to the nineteenth century and in some cases much earlier. One can flesh out the preamble as much as one likes but at some point one is faced with the central puzzle that faced René Descartes, Gottfried Leibniz and William James – what is a soul? Or in the terms given here, what could be a human subject to which aspects of the world are in some way manifest in experience?

The key point made by all three historic figures is that it seems that the subject must be some causal juncture or nexus that has a pattern of immediate input that is manifest to it. That is to say we cannot expect something that has such a pattern available to it only indirectly through the mediation of mechanistic steps to have that pattern manifest to it. If we think manifestation is an integral part of a causal path, in order to retain it within a causal account of the world, which is the only sort we know how to make sense of, then we have to take note of the fact that in causal paths only the immediate interaction determines the subsequent causal chain. All previous steps are optional because the number of possible permutations of preceding events for each step fans out in a combinatorial fashion. Put simply, the central processor in a computer receiving a 0 or a 1 signal cannot be expected to have manifest to it the millions of previous shuntings of 0s and 1s that together determined the arrival of this particular 0 or 1.

What this implies is that a network of communicating units is simply the wrong sort of thing to be a subject because nothing in the network has access to more than its immediate input from other units. The very fact that units in a network are related by causal connection disqualifies them from being a single subject with a single input. One element A may have input X and may then pass a message on to element B, but this will not be X but rather some Y generated by integrating X not only with other inputs but also in the context of the integrating dynamics of B. I suspect that when people consider information in terms of a number of ‘bits’, rather as Tononi does, there tends to be an assumption that all that matter are the inputs or trafficking signals. However, the informational significance of any signals also depends on the integrating dynamics at each point. In a computer, integrating dynamics at each juncture are simple and stereotyped, but in a brain this is very far from the case. Thus for a network of units A, B, C and D the combined inputs to all of these do not constitute an input to anything and so cannot be manifest to anything.

This leaves us with the idea that the subject is a single integrating unit. As William James points out, everything we know about brains suggests that this is an individual neuron. Each neuron has a rich input, often involving thousands or tens of thousands of synapses. It has a single output pathway that passes along an axon that branches to thousands or tens of thousands of other cells. If we were to draw a flow diagram for signal pathways in the brain the individual neuron would clearly be the single integrating unit we seem to be looking for.

I think there are good reasons for taking this conclusion seriously and suggesting that to a first approximation we should consider a human subject of experience to be an individual neuron, one of many other neurons simultaneously acting as experiencing subjects. There are a number of qualifications to this conclusion that I will need to explore, but I am doubtful that they lead to a better alternative.

James discusses the idea, apparently well-recognised in the nineteenth century, that individual brain cells are experiencing subjects. He suggests that it is the only hypothesis that is not self-contradictory. Yet he has two reasons for rejecting it. Firstly, he is not happy about the implication of multiple subjects, despite the fact that a page earlier he has pointed out that such subjects would be oblivious of each other’s existence. Secondly, he makes the cogent point that individual cells are, at least in one analysis, no more individual integrating units than networks. After all, the input pathways to the cell body, within the boundaries of the cell itself, are themselves a sort of network of dendrites. The dendrites have multiple branches and sub-branches so one can argue that the cell body cannot have an immediate input from the synapses, but only an indirect input from these via a series of integration steps at each dendritic branch. A notable caveat here is that the dendrites do not form a network in the sense of having diverging paths as well as converging paths and I will return to this because I think it provides important support for the idea that the cell might be a single integrating unit after all.

James takes the argument further, essentially pointing out that in physical terms we do not get to the necessary immediate input interaction that we want to be manifest until we get down to the level of atoms, since everything else is just a chain of intermediate innteractions between other atoms on the way. This argument seems to suggest that the only thing that could be a human subject, taking into account the discussion I have already covered, is an atom, or even perhaps a single subatomic particle. Something seems to have gone badly wrong here. We should not blanch at leaving behind comfortable intuitions about human subjects as ‘whole beings’ but if we get down to quarks we find ourselves amongst interactions that cannot possibly be of biological significance. A subatomic particle is not going to have an input that correlates usefully with some inference about a dynamic pattern in the world. Below the level of the cell and its dendritic tree this information will be completely lost and replaced by random fluctuations in atomic level chemistry.

James concludes that looking for a single integrating unit to have an experience hits a brick wall. To put it bluntly he bottles out, and in fact he bottles out a second time about four hundred pages later in such a way as to leave us with no cogent conclusion from his detailed and very well expressed arguments.

The straw to clutch at in this situation seems to me to be the fact that in terms of diagrams of information processing the individual cell does at least represent the level of an individual integrating unit in the sense that there are no outputs in between the synaptic inputs and the output of the cell body, which means that all the inputs are only inputs to this one output and there are none of the difficulties described for a true network with diverging as well as converging paths. This does not seem to solve James’s problem with atoms but it does solve the network issue. Moreover, taking into account some of the issues discussed about the limits on what entities can know about their inputs or outputs we may come to conclude that this is all that matters. It may be impossible to decide, or even conceivably meaningless to ask what subdivision of such an integrating unit would be the true subject.

That said, to abandon the possibility of there being some physical dynamic unit that is a subject by dint of its experience of an immediate manifest input would be to abandon the essential project I am undertaking. So, for that project to survive I need to explore the possibility that James’s atomistic account is mistaken and that somehow a cell, or a dendritic tree, that might perhaps be rather less, or perhaps rather more than that of one cell, can indeed have an immediate input of a pattern of signals that signifies a dynamic pattern in the world.

A key advantage of this cellular approach is that it immediately resolves the issue of what sort of biophysical pattern is going to be that which is directly associated with manifestation in experience. As indicated in figure X it also makes clear the distinction between what one might call a representation qua output and a representation qua input (with the caveat that this is not a map like representation but a manifest sign for a dynamic pattern). In the outside world of daily life we take representation to mean some source of signals, like a photograph or page of text, for us to obtain an input from. In terms of experience we need to think of a representation as a ‘presentation’ as an input to what has Ruth Millikan has called a ‘consumer’ – in other words a subject. If that subject is one of a bank of cells receiving inputs from a previous bank in the manner shown we can see clearly how such representations qua output and representations qua input could relate.

Key point: the central difficulty in identifying a human subject is that of finding a single integrating unit that has an input rich enough to be manifest as our customary experience.

6.73 The individual cell as subject: historical ideas

The idea that an individual cell might be a human subject is not new by any means. The main difference between the line of argument I have taken and what has gone before is the emphasis on the possibility of there being many subjects of full, or customary, human experience present in a brain at one time. That is to say that I am not suggesting that individual cells have ‘proto-experiences’, or multiple drafts, that somehow add up to a full experience, nor am I suggesting that there is a single, unique, ‘me’ cell at any one time.

The suggestion in the form I am proposing tends to give rise to a number of reactions. Many lay people simply do not know how such a situation could be conceivable. I hope I have indicated that the alternatives are even less conceivable. Others like the idea just because it is rather unconventional and might be seen as having some of the ‘democratic’ features seen in Eastern religions that downplay the special status of human beings. Many such people, however, have difficulty in absorbing the multiple subject aspect, with its abandonment of a single ‘me’. Philosophers tend to be uneasy about anything that invokes detailed neurobiology, even if, like Andy Clark or Derek Parfitt, they see the subject as being within the brain. They tend to have problems with the idea of multiple ‘me’s too. Neuroscientists in general reject the idea of the single cell as subject for reasons that seem to date back to a debate in the early 1970s, which I will come to shortly. Ironically, they seem to be wedded to network-based subjects almost to a man, despite the apparent incompatibility of such models with a causal (i.e. scientific) description.

Cells were first identified in biological tissues with the development of the microscope in the seventeenth century. It seems that more or less immediately cells were attributed some form of sentience. This may not be surprising since it was probably more common to attribute sentience more widely in the universe than today. Leibniz went through a variety of positions in relation to the existence of microscopic souls. At one stage he suggested that all large creatures with souls were made up of smaller creatures with souls and that this went on ad infinitum. He considered cells to be one of these lower creature levels. In the Monadology he does not discuss cells but indicates that a human body is made up of a large number of smaller units, each itself a perceiving monad. For most of these monads perception would be indistinct, as might be expected in the context of my previous discussion if their inputs were not collated by a sophisticated knowing machine. Only one dominant monad would have the apperception that equates to full or customary human experience.

Leibniz is clearly not wanting to have multiple ’me’s and so postulates that one monad somehow comes to dominate all the others and take on an apperception suited to a single human soul. Perhaps the weakest part of his theory is the absence of any indication how one monad comes to dominate others and why one should do this more than others. However, there is an extent to which modern biology would parallel Leibniz’s idea at least in that it seems that brain cells have specialised roles in both perception and determination of behaviour in a way that could reasonably be called ‘dominance’.

It may be useful to mention here some of Leibniz’s thoughts about James’s problem of what could have a rich immediate input from the world. Leibniz was concerned about this and suggested that maybe a single point could have a rich input in the way that a single point can have light arriving at it from many directions. The idea of a point having sentience is something that seems to occur to many people puzzled by sentience and wanting a truly fundamental account. Those attracted to panpsychism often try out the idea of every point in space or every fundamental particle having its own experience.

However, Leibniz realised that a single point would not do because it would have nothing to distinguish it from other points. He realised that the monad must itself have some sort of richness or complexity, but without mechanical parts. In the Monadology he does not relate this to cells, perhaps because, like James, he sees looming the problem of cells having just the sort of mechanical parts we do not want: they would make interactions indirect rather than immediate.

Another issue with the idea of every point in the universe being sentient, or indeed with every fundamental particle being sentient, is that we need to explain why human reports of sentience relate to rather special sorts of experience, presumably associated with rather special points or particles in brains. The dynamic basis of the manifestation needs to be at an appropriate level and in terms of standard neurobiological teaching this needs to be at least at the level of the integrating architecture of dendrites.

This raises two further issues that I will need to return to. The first is what could be meant by a dynamic level in terms real enough to support a real manifestation.

The second is how reports can be about such manifestations. It seems that we are going to need some complex dynamics between whatever the experience is manifest to and the apparatus of speech. These dynamics may need to have something of the complexity discussed under the idea of a knowing machine but rather than being collative or inferential they would seem to be something with a reverse function. Thus when we talk of a report being about an experience it might seem that we would be referring to an aboutness that is somehow the reverse of the aboutness of ‘intentionality’. Curiously, I think it may turn out to be rather similar, but more of that in due course.

Very little attention has been paid in the last fifty years to the discussions that occurred in the eighteenth and nineteenth centuries about life units and what made them living rather than inanimate matter. Concepts of ‘life force’ are derided, but I suspect that the level of debate at the time was considerably more sophisticated than is often portrayed. William James refers to the fact that ‘polyzooism’ or the idea of multiple cellular sentient beings within a human being was well recognised prior to 1890 and he mentions in this context the names of Lotz and Herbart. It has been said that Darwin was of the view that protozoa had some simple form of sentience.

The status of theories about sentient cells in the first half of the twentieth century is not clear to me. As yet I have not seen any detailed account. Sherrington discussed the idea of convergence of pathways in perception on to individual cells but by this time the agenda seems to have moved from a general one about sentience to specific issues about mechanisms of object recognition. As I shall discuss later, it is easy to conflate sites of recognition with sites of experience but the relationship between the two may be subtle. By the 1970s, when I studied neuroscience as an undergraduate the issue of the nature of the human subject was never mentioned. All interest was focused on mechanisms rather than experience.

The loss of interest in the nature of the human subject in the first half of the twentieth century presumably reflects a trend also indicated by the tendency to behaviouristic analysis. Discussion of experience was said by people like Skinner, to be of no interest to the understanding of the human mind. In philosophy Gilbert Ryle was also taking a behaviouristic approach in the sense that he denied that there were any mechanisms within the human mind of the sort we encounter outside. Strangely, although the term behaviourism is applied both to psychology and philosophy at this time there are ways in which Skinner and Ryle could not be more different. In some sense Ryle is proposing that mind has to be treated quite differently from matter, much in the way that the man he ridicules, Descartes, did. Moreover, a salient feature of academic life at this time was the polarisation of the ‘two cultures’ of CP Snow with philosophy aligning itself with the humanities and taking pride in being scientifically illiterate.

Interest in sentience seems to have returned in the late 1980s and early 1990s, chiefly in philosophy. By this time neuroscience had been through a phase in the 1970s when the idea that individual cells had critical computational functions was out of favour, with a new emphasis on functions being carried out by nets of cells. In 1972 Barlow had emphasised the potential importance of the contribution of individual cells but by 1980 all the emphasis was on cells operating in banks or networks. The neuroscientists were not discussing sentience but when the philosophers returned to this topic the level of interest had changed.

In the last twenty years there has actually been a shift back to emphasis on the role of individual cells, with Christof Koch suggesting that much of the apparently random variability seen in the behaviour of individual cells in experimental systems was likely to be due to our ignorance of regulatory factors that we could not control in the normal way. Koch has not, as far as I know, raised the issue of sentience for individual cells but he has reversed much of the received wisdom about cells only functioning in groups.

The idea that individual cells might be sentient has probably been raised from time to time throughout the last century. However, the most apparent example in the last years of the century was from a Russian information scientist called EA Liberman. Since the millenium a number of people have raised the issue, including Erhard Bieberich, Steven Sevush and myself.

Key point: the idea that an individual cell might be a subject of experience goes back almost as far as the discovery of the cellular basis of life.

6.74 Which cells might be customary subjects?

If we propose that individual neurons are human subjects questions immediately arise as to whether all brain neurons are subjects and if not why are some neurons human subjects and not others.

If we start from a panexperientialist base and allow neurons to be the sort of entity that has experience then in a sense we can say by definition that all neurons will be subjects. However, we are interested in the sort of subject that is implicated in descriptions of experience of the customary sort. There is a strong suggestion that we want to limit the number of neurons getting this sort of experience maybe to those all receiving the same pattern of signals ‘broadcast’ in a workspace such as that suggested by Baars. Perhaps these cells are in the prefrontal region and confined to a particular cortical layer.

This simple model may need to be modified in view of the apparent resilience of experience to prefrontal damage. We may want to invoke a wider distribution of cells, perhaps as suggested in the context of the Changeux-Dehaene model. Perhaps more importantly, we may want to admit that cells operating at lower levels of hierarchy might also have a form of experience, consistent with the ‘customary’ account, but perhaps lacking the highest ‘meta’ aspects.

This raises the important possibility that there may be no definitive answer to the initial question in the title. If I write this text and it is read by a reader then it may give rise to experiences at different levels of abstraction or sophistication which nevertheless all contribute to a report of a customary experience. If field workers are sent out to study and then report back to supervisors who then report back to principle scientists who then report to the world in the form of a polished scientific paper then it is still reasonable to say that this report reflects and is caused by all the experiences of all those involved.

Nevertheless, we do seem to be able to make a distinction between those experiences that are associated with dynamics that in some way help to determine reports and experiences of the type that seems to have a narrative structure of the sort that lends itself to a report as a narrative. It seems right to think that if there are multiple subjects at the cellular level there is a specific subset that has this narrative type of experience by dint of being fed signals collated in a particularly sophisticated way as in a global workspace of the Baars type.

There might, however, be a caveat here. Different people have different views on the relation between language and ideas. Some hold that without words we cannot even have conceptual ideas. Probably rather more people would say that language helps us to develop ideas but that the ideas themselves are not immediately dependent on a form of words. My own view is that the sorts of experiences that seem to be ‘mine’ and what ‘I am writing about’ are non-verbal. I can conceive of there being other experiences in my head that are more directly dependent on words. I actually see these verbal ideas as less sophisticated, but they may constitute a ‘higher hierarchy’ of collation. We begin to be faced with a situation in which highest level of hierarchy cannot be assumed to be mostly closely associated with customary subjecthood.

I do not see a way to adjudicate clearly on the answer to the question in the section title. I think that it may be based on another pseudoconcept. Or perhaps, less severely, I would suggest that it may need to be accepted that the answer is a complex one that breaks down into rather different sub-questions in different contexts.

What I think would be reasonable to suggest is that many cells will have inputs of signals, which may be manifest to them but which will not be interpreted as a sense of a pattern of world goings on occurring in a context of episode and in relation to a concept of an observing self. Some cells will simply receive signals that encode elements of a visual pattern at a level before assignment to any specific object. Others may deal with signs for objects but not relevance in terms of ownership or emotional significance. Still others may deal with objects in a full conceptual and evaluative context and these may be the ones we are most interested in.

Just as we would expect customary subjects to be limited to those with the relevant inputs, it may be that they are further limited to a subset that deals with those inputs in a particular way. It would be reasonable to suggest that of the cells receiving signals within a global workspace, some will be adapted to dealing with previously identified and tagged ideas and some will be adapted to dealing with novel patterns, that may or may not deserve tagging for future identification. The sort of integrative dynamics appropriate to these two types of cell may be very different. Cells identifying familiar tagged patterns need only entrain signals in established associative pathways. Cells dealing with new patterns will need to be involved in comparative and collative processes that allow these new patterns, if to be tagged, to be linked to appropriate associations. Put simply the task of dealing with a known type of dog, like an Alsatian or corgi, is different from dealing with an unfamiliar hound that must be categorised as ‘definitely some small sort of dog and rather more like a dachshund than a Jack Russell’.

Since we associate our conscious experience particularly with deliberative and conceptual thinking rather than automatic responses it may be that customary experience relates very specifically to cells that are involved in making responses to novel patterns. We might think of these as being ‘Solomon cells’ in the sense of having the function of making a novel response to a novel scenario.

This possibility leads to a rather curious idea that I think is worth raising, not so much because I think it likely to be the case but because it illustrates the uncertainty of the ground I have to tread. It is conceivable that cells hosting customary experience are those which have not yet been bound into a tagging mordant loop. They are cells ready to form memories but not yet used. Once used they may lose their Solomonic wisdom and, perhaps by bypassing complex integrative dynamics and replacing it with a rigid alternative, become automata. This has the strange implication that each cellular subject only has its one go at hosting customary experience. This is probably unnecessarily fanciful but I raise it because Marr’s model of cerebellar Purkinje cell adaptive learning seems to follow a mechanism rather like this. The cell initially has the capacity to respond to very complex patterns of input but when allocated to a routine acquires a responsiveness to a much simpler input.

Key point: it is likely that if cells are customary human subjects then they belong to a specialised category at the apex of the knowing machine’s collational activities.

6.75 What sort of entity would provide a cellular subject?

If we want to develop the idea that an individual cell is a human subject then we need to identify a specific dynamic entity that might have the right sort of input and which overcomes William James’s argument that only an atom (or subatomic particle) has immediate direct inputs from the world. We need to find something at a larger biological level that genuinely has a rich immediate input relevant to neural signaling and which in some way has an intrinsically defined boundary. Post-1980 physics provides some interesting options.

The most promising starting point that I can see relates to comments made in section 2 on the existence of distinct macroscopic dynamic entities accompanying aggregates of atomic or molecular entities in structures such as vases or tables. These distinct dynamic entities often have domains that correspond to everyday ‘objects’ but are not collections of parts, but rather additional indivisible acoustic units. Note that these entities are modes, not fields. Fields are arrays of values, like electrical potential, in an arbitrarily chosen domain, that need not arise from any one particular entity. Strictly speaking, for any particular type of value there is only one field – the entire universe. There is often confusion because energy-bearing modes come under what is called field theory. For modes based on spatial asymmetries – the acoustic modes, spatial domain is non-arbitrary. Unlike other modes it is stable and time independent.

Modern field theory indicates that all energy-bearing entities are associated with asymmetries in the universe. The types of asymmetry associated with energy packets like photons relate to aspects of group theory unfamiliar to most people. The asymmetries that give us ‘separate solid objects’ (as opposed to some homogeneous structureless fluid) are associated with modes of rotation or vibration. Rotation does not look to be relevant to neurodynamics. Vibration might also seem alien to neurodynamics but at least we know that certain types of cells, including nerve cells are capable of supporting vibrational, or acoustic, modes that are in some cases functionally important. The most obvious example is in the outer hair cells of the sensory apparatus of the inner ear.

The general principle is illustrated by a Caribbean steel drum made from the end of an oil drum. The drum end starts as a homogeneous sheet of metal, symmetrical in all directions, which has only one mode of vibration – a deep booming sound. To form the musical instrument the sheet is beaten out into several areas of thinner metal, each with a sharply defined outline. The new asymmetries in the sheet are associated with new vibrational modes, one occupying each of the thinner beaten areas and each producing a different musical note. In everyday terms such asymmetries might be more familiar as discontinuities. Every discontinuity creates a new way to vibrate.

It may seem peculiar but I think there are reasons for thinking that inasmuch as familiar objects include legitimate individual entities, those entities are the associated vibrational modes. A chair really is a separate chair because it can vibrate separately from other things. So if we want a cell, or something occupying a domain at more or less the cellular level, to be a separate receiving entity that can have a manifest input then the prime candidate for such an entity is an associated vibrational mode.

Intuitive resistance to the idea of a human subject being ‘merely’ an acoustic mode or ‘buzz’ in a cell may be strong, on the grounds that at least ‘I’ should be some solid stuff. Yet the absence of any mass to an acoustic mode fits rather well with other intuitions and observations. When we fall asleep or expire we do not get lighter. The ‘spirit’ has traditionally been seen as weightless. The idea of a spirit that is unrelated to the laws of physics is rightly criticized but a weightless spirit that fits neatly into physics ought to be perfectly respectable.

The energy held in these macroscopic modes is generally only a tiny fraction of the total energy in the domain. The vibrational energy of a musical instrument is very small in comparison to the thermal energy at room temperature. That in turn is small in comparison to the chemical energy of the molecular bonds and even smaller in terms of the energy buried in nuclear forces and the rest mass of particles. Nevertheless, in biological terms vibrational energy is not trivial, nor is the energy of modes of crystalline order, which significantly affect latent heat.

Thus, there is no doubt that the structure of a cell is associated with asymmetry (discontinuity), in the way that any ordered solid structure is and can be expected to be inhabited by energy bearing modes. Moreover, the complex levels of order in a cell are likely to give rise to quite a range of modes. Many cellular substructures are quasi-crystalline, including the cell membrane, which is a form of liquid crystal with rigid order (asymmetry) in the direction normal to the cell surface and relative disorder (symmetry) in the plane of the surface. The question is not so much whether modes occupying all or large parts of the cell exist but whether they are stable enough and in a sufficiently appropriate domain to act as subjects.

Acoustic modes come in forms with complexity that reflects the complexity of the asymmetries or discontinuities that bring those modes with them. In a homogeneous fluid acoustic waves are longitudinal – like sound waves in air. Waves radiate out from a source with oscillation in the direction of propagation. At the surface of a fluid transverse oscillations, with oscillation at right angles to the direction of propagation, occur, as in sea waves. In solid objects more complex transverse oscillations occur where there is variation in restriction of movement, as when touching the centre of a violin string changes a transverse oscillation of the whole string into two oscillations of half the length.

As solid structures increase in complexity the number of modes increases, but so also does the tendency for modes to dissipate through damping. Acoustic waves in solids propagate best as standing waves in which modes are reinforced by constructive interference or resonance. A violin string will resonate if a finger is place lightly at a point that allows constructive interference (giving a harmonic) but for most positions of the lightly placed finger interference is negative and no sustained sound occurs.

For an acoustic, or vibrational, mode in a neuron to have a chance of being a subject of ‘customary’ experience it must, therefore, be both in receipt of the right sort of input and free from excessive damping, in order to generate an output that usefully reflects the input. It might be argued that all we require of such a mode is that it should have the right sort of input. However, if we are to build the having of manifest experience into a causal chain that includes the reporting of experience I think we need there to be a biologically effective output.

On this basis the first requirement for a cellular subject mode, that it occupies the domain of relevant signal input, suggest that it occupies the dendritic tree into which synapses feed, or a part thereof. It need not include the domain of all synapses because we can allow a cell to have some inputs that are not manifest to our ‘customary’ subject. We just want the domain to cover enough inputs to explain the richness of customary experience. The second requirement, that the mode is not unduly damped, probably chiefly implies that this is not a mode that requires movement of cytosolic water, since this is likely to lead to rapid dissipation and damping.

Further consideration of the implications of an acoustic mode as subject of customary experience can be divided into two levels. I will initially consider some broad issues. In the subsequent section I will turn to possible detailed mechanisms.

Key point: within a cell there are constraints on what could reasonably be considered a substrate for a human subject.

6.76 General requirements for an acoustic cellular subject

Attempts to define the requirements for an acoustic cellular subject can fall anywhere on a spectrum of stringency. If only bare minimum requirements are considered the plausibility of there being suitable modes in cells seems high. The down side is that no very specific predictions are required that might be falsifiable by experiment. At the other end of the spectrum if requirements are made much more stringent then the likelihood of appropriate modes existing may seem more remote. There is also a significant chance of discarding viable candidates. However, stringent requirements are more likely to lead to testable predictions.

If all we require of a subject is that it have a relevant input, on the grounds that any assumptions made about what sort of output it needs to have may be illusory, then I think there is a strong case for saying that we can be pretty sure that some such modes will exist in neurons. James’s problem may be considered solved. I do not think we should be satisfied with this but it is useful just to go through some of the historic arguments about what minimum is needed and how it would work in an account compatible with physics.

One of the traditional criticisms of Descartes’s account of the subject, or soul, is that it requires causation to pass from a ‘physical’ realm to some other mysterious ‘mental’ realm and back again. As I shall discuss in some final comments on Descartes’s legacy I think this is largely a misreading of Descartes based on lay concepts of ‘the physical’ that he quite clearly discarded. Descartes’s two types of substance differed in that one and not the other obeyed ‘mechanical’ dynamic rules. The dynamic rules of the soul did not. Even by Newton’s time it was clear that even ordinary matter does not just obey mechanical rules. Descartes had rightly realised that we need more than mechanical rules, but wrongly claimed that this is only true of souls. There is, however, a technical issue in relation to Descartes’s account, about conservation laws, that is relevant to what we think we might require of a subject in an account consistent with contemporary physics.

Descartes believed that motion remained constant, without needing to specify direction. Newton and Leibniz established that it is momentum that is conserved, which means that things cannot change the direction of their motion without something else changing its direction of motion to compensate. When we see a billiard ball bounce back off a cushion all we see is the motion of the ball. What Newton pointed out is that when the ball changes direction, in order to conserve momentum the ball must impart some tiny velocity to the table and the earth it rests on. Descartes had suggested that the soul might influence the body by causing ‘swerves’ in internal motion that did not require any physical law to be broken because the speed need not change, just the direction. Leibniz denied that this was possible because momentum had to be conserved and this involved direction.

It is unclear to me why this should really be a problem for Descartes. As I understand it, in Descartes’s model a large number of nerves fed in to the pineal soul, on which they exerted forces by tiny tugging movements. The effects of these tuggings were not on ‘parts’ of the soul and so not mechanical. The soul, informed by these tuggings was then able to return the influence in a similar non-mechanical way in the form of will, presumably with the result that centrifugal nerves controlled the body by further tuggings. As long as there are lots of nerves involved, which there are, this seems to pose no problem for conservation of momentum since both the incoming and outgoings tuggings are likely to more or less balance each other out whatever complex pattern of information they carry. Even if not, all that would be expected would be for the pineal to wobble around a bit. Thus, Descartes’s swerving need not violate any laws and is no more ‘ethereal’ than Newton’s gravity, or light. What probably did violate laws is his conception of free will, but that applies to any set of dynamic laws.

More recently, there have been mistaken claims about conservation of energy being violated by Descartes’s ‘swerving’. In fact swerving does not necessarily require any work to be done so energy need not be imparted. The earth swerves constantly through space under the influence of the sun but no energy is gained or lost by either party. A ball bearing running along a perfectly smooth circular track will continue going around and around with no work being done. A beam of light in a coiled fibreoptic filament will go around and around with no work being done. Causation in physics does not necessarily require work. A pattern can be causal through constraining the possibilities for unfolding of other dynamic dispositional patterns. In fact in modern physics this sort of ‘causation’ more or less completely replaces the traditional push and pull concepts.

In other words, shapes cause things. The shape of set of railway points causes a train to run on one track rather than another. The shape of a lens causes light to be focused to a point, the shape of a beach causes a wave to break.

In similar terms, the shape of the dendritic tree of a neuron causes many local changes in electrical potential at individual synapses scattered throughout a volume of brain tissue to co-determine a change in electrical potential somewhere else, at the cell body, through the complex conductance, capacitance and inductance properties of a branching patterns of tubes. Whatever events are occurring at the local level of individual ions passing through channels the overall shape of the tree makes a crucial causal contribution. It determines the output of all the local events. If we accept that ‘decisions’ in human brains is made through the operation of nerve cells then it is a sense the ‘passive’ structure of the dendritic tree that does the deciding. Without it the ‘active’ events such as movements of ions would just generate chaos.

The energy bearing modes that arise with asymmetries in field theory are seen as going hand in hand with those asymmetries. It is not that one causes the other. They are what James Ladyman calls ontologically co-dependent. This suggests that all that we require of an acoustic mode as subject in terms of relation to output is that it belong to a shape that determines decision making, which that of a dendritic tree does. It need not do any work. Moreover, there is no doubt that the electrical potentials associated with a dendritic tree will generate forces experienced by that tree, even if, as for the earth orbiting the sun, no work is done by these forces. Thus there seems to be no problem in attributing an experience of all the electrical inputs to this shape/mode entity.

Given that all that is required for an energy bearing mode to exist is for some local pattern of order to exist and cells have plenty of these, not least the pattern that is the liquid crystal of the cell membrane, then it seems pretty certain that there will be modes around that will have these minimal requirements for subjectivity. The objection might be raised that if each mode has a different sort of input, as it will do, then this does not explain why we talk about the input to one particular type of mode. However, as I shall discuss later, I think this is a false problem. Reports of experience are about experiences in a sense that reflects complex collation, just as the aboutness of ‘intentionality’ does. In a model with multiple subjectivities the idea that a report reflects a single experience has already gone by the board.

I have serious reservations about the above analysis and will argue for a much more specific analysis in the next section. However, perhaps my main worry is that the above analysis may in fact be on target to the extent that it may be impossible to formally refute it and thereby establish a more specific model. If all my discussion so far achieves is to show that the puzzle of how there could be a human subject in side a brain is not a puzzle because we can predict that there will be plenty, even if we can never know what they are, then I would be disappointed. Dissolving the problem of the subject would be an empty victory if it meant that we never knew what the subject was.

Key point: it might be argued that simply the existence of an individual membrane structure is enough to make a cell a credible receiving human subject. However, as indicated next, there are also reasons to think that there are tighter constraints on a model.

6.77 Specific models for an acoustic subject

The idea that all we need of an entity taking part in a causal chain is that it be a certain shape is, I suspect, salutory and relevant to exploration of the human subject, but I think it may be too easy. If we consider an energy bearing acoustic mode associated with a dendritic tree and ask what form any input from electrical potentials might have it begins to look as if we need a more precise account.

In the case of the earth orbiting the sun or the ball bearing in a groove or the physics of ‘swerving’ without work are well understood. We need a similarly clear account of a relation between electrical potentials and an acoustic mode.

In an important sense the apparent ‘swerving’ of electrical phenomena determined by a dendritic tree is different from the other cases because it does involve work. The end result is dependent on dissipation of energy derived from ATP to produce heat. Although the shape of the dendritic tree is important it may not be legitimate to connect that in any way to the relation of the shape to any associated acoustic modes. To have a solid causal story involving a mode itself it seems that we ought to have an account of how electrical perturbations could influence acoustic dynamics and, to give an account of the final electrical output, vice versa. We want a means of electromechanical coupling.

Electromechanical coupling has received little attention in neurophysiology with the exception of events in the inner ear and in a few papers by Iwasa and Tasaki and by Heimberg and Jackson on mechanical changes associated with action potentials. There is no doubt that acoustic effects can entrain electrical effects because that is how our hearing apparatus works, but acoustic events have not generally been considered relevant to integration of electrical signals in dendrites. To suggest that they are runs contrary to the accepted model but it has the attraction that it involves making specific predictions that might be testable by experiment.

If an electrical field is generated in the vicinity of any structure that has some degree of polarity the structure will experience an internal mechanical or elastic force that will tend to deform it. Polarity simply implies that negative and positive charges are not homogeneously distributed throughout the structure. Polarity is the norm in biological structures at the level of subcellular organelles and membranes in particular. The standard biological membrane is a back-to-back double layer of polar molecules. What this means is that an electrical potential gradient across such a membrane will tend to make it become convex or concave. The physics of this effect, known as the flexoelectric effect, has been studied in detail by Alexander Petrov.

In a complex membrane structure like a dendritic tree the effects of electrical potentials on internal elastic forces can be expected to be similarly complex. Crick predicted in the 1980s that depolarisation events at synapses should lead to local mechanical deformation or ‘twitching’. This twitching has subsequently been observed. In the body of a dendrite, which is essentially a branching tube, the shape changes associated with electrical field are more difficult to predict. A potential that would make a flat membrane become convex will have a more subtle effect on a membrane that is already tightly curved and thereby constrained. Perhaps the most likely change with electrical potential is a shortening of the membrane along the length of the tube. This effect has been documented by Tasaki and Iwasa in axons carrying action potentials. The tubular structure does not actually change shape but the membrane ‘bunches up’, rather like a pushed up sleeve, in response to a change in electrical potential.

A single local elastic deformation in response to an electrical potential, such as a twitching synaptic spine, would not provide us with an entity that receives a rich pattern of input. To be a plausible domain for an input that corresponds to customary experience we would seem to need an acoustic/elastic mode that occupied all or a significant part of a dendritic tree. It is difficult to know exactly what the requirements would be but I see arguments for three in particular.

Firstly, to reflect the input domain of the dendritic tree as a whole there should be an oscillatory mode bounded by that domain, perhaps most easily conceived as a resonant standing wave mode of the sort that occurs in the barrel of a wind instrument. Although not essential, it would seem to make sense for there to be a single dominant mode that took up most of the energy passed from electrical to acoustic form. Herbert Fröhlich proposed that energy was preferentially transferred to nano scale mode of oscillation within the membrane bilayer. Fröhlich’s mode look to be too high frequency to be biologically relevant. However, Tasaki and Iwasa’s observations suggest that the longitudinal change does oscillate at what might be a relevant frequency, even if for a single action potential this is for only a few cycles.

Secondly, for the relationships between potentials at different synapses along the dendrite to be relevant to the manifest input to the acoustic mode it seems likely that there would need to be some sort of match between the distance between synapses and the wavelength of the mode of oscillation. Thus, considering the rate of propagation of elastic deformation in dendrites and intersynaptic distances in the range of microns, one would expect the frequency of relevant dendritic acoustic oscillatory modes to be in the 1-10 megaHertz range.

Thirdly, to be part of a relevant causal chain, one might expect a relevant acoustic mode to influence the eventual firing of the cell by regulating the depolarisation of the cell soma. This is entirely feasible since excitable membranes carry ion channels that can be opened by physical deformation. This would make the dendritic acoustic subject an integral part of the causal path of cellular computation.

An alternative to a resonant, or standing wave, mode occupying the dendritic tree might be a travelling wave. There seem to be practical problems with this idea, however. To influence the firing of the cell one would expect the travelling wave to pass from the ends of dendritic branches down to the cell body. It is unclear how such a convergent wave could be synchronised. A travelling wave in the opposite direction almost certainly does occur as a back propagation phenomenon when the cell body fires, but that could not mediate the firing response itself.

These detailed suggestions are open to a range of criticisms but they do at least show that there may be grounds for making quite specific predictions about mediating pathways in dendrites directly relevant to experience that might be open to experimental testing. Thus, it may be worth looking to see whether the equivalent of Iwasa and Tasaki’s elastic change for the dendritic tree might give rise to a standing wave oscillation with frequency in the megaHertz range.

Key point: if a subject of customary human experience is an acoustic mode in an individual cell it may be possible to predict parameters such as frequency of oscillation and effective wavelength.

6.78 Marrying experience with fundamental physics

William James’s problem with the human subject was that however small we get we cannot find a ‘single thing’ that could be a subject in a brain until we get down to an atomic level too small to be relevant to biological function. The subject needs to be a single thing because things A, B, and C with inputs x,y,z, p,q,r, and e,f,g cannot in any sense consistent with a causal account be a subject with input x,y,z,p,q,r,e,f,g. Moreover, the existence of signaling pathways between a ‘network’ of A, B and C if anything makes them less likely to be a single subject because some of the inputs to A will be from B and an input to A is not the same as some antecedent input to B. Further, the existence of traffic of signals between A,B and C does not help to make them a subject since null and inhibitory signals are just as important in brains as activating ones.

What I have tried to argue is that the level of input we need is the electrical input to a dendritic tree and that there might just be ‘single things’ at this structural level that would experience these inputs. Estimates how many elements there are in a human experience range from about 50 to 50,000 and that turns out to be exactly the sort of range of inputs that a dendritic tree will have.

Historically the human subject has been assumed to be the whole human being but both Descartes and Leibniz realised that a human being, in the sense of a bodily structure, is an aggregate, not a single thing or ‘simple substance’ as they would have it. (In this usage ‘substance’ implies an individual item or token, not a type of material in the modern sense that gold is a substance.) Even the brain is an aggregate of interacting parts.

A single thing, or simple substance, in this sense was considered by Descartes and Leibniz to be unified or indivisible in a non-arbitrary way that did not apply to aggregates. Both had difficulty with explaining this because they were aware that simple substances appeared to need complexity. Something complex that was extended seemed to take on the properties of an aggregate so Descartes rejected the idea that a soul could be extended. (But note that Descartes’s conception of spatial extension was not ours.) Leibniz thought that everything was in a sense a soul and so rejected the sort of extension that implies mechanical interaction completely. He suggested that somehow some souls could come to dominate domains of other souls but did not give any idea of how this might marry with physics.

Descartes and Leibniz may have been pursuing the wrong track but I think not. The key question, I think, is what complexity is it that is unified or indivisible in a substance with no parts? I think the answer is the relation of the substance to the world. In a world view in which everything is an instance of operation of a dynamic dispositional pattern there is a sense in which dynamic relation to the world is all that there is to anything anyway. This unity or indivisibility of relation provides, I believe, a critical link between aspect of experience and aspects of modern physics.

In experience the human subject finds itself relating to a scene in a way that we cannot conceive of being divided up. It makes no sense to say that one half of my experience of a scene is disconnected from the other half. We could say that everything manifest in an experience is co-manifest. In modern physics unification or indivisibility appears in a different context. What distinguishes a ‘single thing’, or more correctly a single instance of operation of a dynamic pattern, from an aggregate is that its relation to everything else is indivisible or unified. The relation of a photon to all the slits in a diffraction grating is indivisible. Quantum theory tells us that all immediate interactions are of this type, whereas indirect interactions do not have this feature.

This parallel in indivisibility suggests to me that the confusion over the relation of the ‘mental’ to the ‘physical’ stems in part from a failure to be explicit about whether we are treating things as subjects or objects in physics. We set out in physics to study objects. However, to do so all we find we can do is analyse the relational dispositional properties of instances of dynamic patterns. For centuries we have concentrated on the indirect interactions that seem to occur between aggregates, for which each interaction can be considered separate from the next. But when we get to the stage of being able to analyse the immediate interactions of single dynamic packets we suddenly find ourselves in a territory that looks like the study of subjects. And that is surely what we should fully expect if subjects are themselves instances of operation of dynamic patterns. If we study the relation of a photon to a diffraction grating we seem to be studying the behaviour of an object but we should also expect that to be studying, from a reverse perspective, the dynamic relations of a subject. That these are indivisible should come as no surprise.

When considering aggregate objects there is no need to be fussy about whether or not these really are single things. In fact our language of objects has traditionally been loose in this regard. We probably have genetically programmed categories of ideas about ‘objects’ that are reflected in the philosophers’ concept of ‘natural kinds’. At the same time we recognise the arbitrary nature of object status by having the concepts of parts and wholes. Perhaps Leibniz’s most important contribution was to indicate that if we want a single unified ontology of the world that deals with both subjects and objects then this loose talk about parts and wholes needs to be abandoned. Attribution of single thing status may not matter for objects but it does for subjects. A subject is something which has only one indivisible dynamic relation to a manifest world. Modern physics has shown us that if we examine dynamic relations in enough detail we can identify this indivisible property ‘from the other end of the telescope’.

This way of defining individual entities in terms of a manifest relation to the world might seem to be an epistemological one. It is often said that modern physics shows that the basic fabric of the universe is information, not stuff. I think we need to accept that the ontological and epistemological are very closely bound together. However, in the light of previous comments on knowledge I think there remains an important distinction. Manifestation does not imply knowledge, and when it does seem to impart knowledge it seems that this is not quite what we have traditionally thought it was.

The overriding problem we have at this point in history is the apparent lack of any way of discerning the rules of correspondence that link the qualities of manifestations to the immediate dynamics that underlie them. It seems that ideas of dynamics can be inconsistent and a pessimistic view might be that the rules of correspondence are a hotch potch of what has evolved in our brains through natural selection. They may share the Heath Robinson style of construction we seem to see in DNA, reflecting retention of chance alterations that just happen to confer advantage. A more optimistic view might be that there are some understandable relations between ideas of qualia or of dynamics or values, like good and bad, and the dynamics that underlie them. The problem is that until we have a clear idea of where the manifestations are occurring we have little chance of beginning an analysis.

This view of manifestation as nothing more than what an entity’s dynamic relation to the world is like to it might seem to fall under what has been called epiphenomenalism. However, I think this would be misleading. Strictly speaking, an epiphenomenon, like the hoot of the steam whistle that occurs when a train starts up, is a dynamic causal chain that is a side branch from that which entrains the events of real interest. In the complementarity based view there is no side chain. The manifestation is an integral part of the dynamics and you cannot have the dynamics without it. Dynamics always did need manifestations, to make them dispositions to something, rather than just dispositional mathematical patterns. If manifestations are inherent to all immediate dynamic interactions then we have a fully parsimonious and consistent model.

There is a strange twist to this way of explaining the relation of ‘mental’ to ‘physical’. If the mental aspect of an interaction between subject and world is the experience then its content is not the subject, but everything other than the subject. The only account we could have of subjects is a third person dynamic or ‘physical’ account. It is in the very nature of the complementarity-based approach that any attempt to describe one side of the divide has to draw on the other side. And perhaps if everything we can encounter in the world is relational, and the relations are asymmetrical, it makes sense that everything has two different aspects. The two aspects belong to the same network of causal events but because all events are many-to-one, or universe-to-subject, there is no simple correspondence of ‘two sides to the same event’. Changing the point of view moves one to a different event. This might seem to make analysis impossible but I believe it just makes it very difficult. The key to making it a bit easier is to identify what dynamic entities human subjects really are.

In 1928, Rudolf Carnap, wrote in relation to the ‘mind-body problem’: ‘...there still remain, in the main, three hypotheses: mutual influence, parallelism, and identity in the sense of the two aspect theory...’ The complementarity-based view is closest to an identity theory but at least in philosophers’ terms it is not a true identity. If we predicate something of a manifestation we cannot take this to be a predicate of the complementary dynamics. The structure of our language is subtler than that. It is genuinely a fourth hypothesis, and the one that seems to me to be most satisfactory.

Key point: a complementarity-based view of the relation of experience to dynamics provides a way of resolving questions posed by Descartes and Leibniz in the context of modern quantum theory- based physics.

6.79 Unpacking the binding problems

It is very hard to conceive of how a set of electrical signals somewhere in a brain could give rise to a single unified experience of a vase containing three red roses and two white lilies. This difficulty is often loosely referred to as the binding problem. Thinking in terms of electrical potentials as inputs to an acoustic subject in a dendritic tree, even if off target, may help to clarify the nature of this difficulty. In fact the binding problem turns out to be a mixture of related but rather different issues that need to be picked apart before we can hope to make sense of it.

In philosophical circles the binding problem is usually taken to be the problem of how or why it is that elements of experience are all unified into a single scene, which relates to the remarks in the previous section. In neurobiology ‘binding problem’ has also been used to describe the problem of how in the above example the signals for red are connected up with the signals for a rose shape and the signals for white are connected up with the signals for the lily shape, so that we do not get white roses and red lilies or just pink flowers.

Despite superficially seeming similar these two problems are in a sense opposite. The neurobiological problem is actually a problem of segregation, not of unification. It is the problem is how the red goes with the roses and not with the lilies and vice versa for the white. It is the problem of how you keep the meaning of signals separate rather than having them all mixed up like a vegetable soup. As indicated earlier, a popular approach to this sort of binding problem is to suggest that it is achieved by synchronisation of signals. Thus the signals for red might be synchronised with the signals for roses and those for white with those for lilies.

There are a number of problems with the synchrony hypothesis. Firstly, it clearly has nothing to do with the solving of the philosophical binding problem, which is about how all the aspect are together. If synchrony can work to segregate it cannot also work to generate overall unity. This simple fact seems to have been missed by a number of theorists including Gerald Edelman. The second difficulty is that our experiences are very rich and if every relation between elements in a scene was to be assigned a different synchrony we would rapidly run out of options. The idea may indeed by self contradictory. Signals for red need to be synchronised with a rose, but the signals for that rose need to be synchronised with the next rose to indicate they are adjacent. But those signals need synchronising with signals for the lily next along, and those with the signals for white and … Even without this absurdity there are simply too many elements in a typical scene to encode each relation in a different phase relation or frequency. Finally, as mentioned before, synchrony of signals is only of causal relevance at a point where both signals arrive and interact and that is not what the electroencephalographic studies of synchrony measure.

The idea of segregational binding by synchrony may also draw on a false assumption about the meanings of signals in a brain. There seems to be a suggestion that activity of a cell or minimal group of cells would mean ‘red’ or ‘rose’. However, even the firing of a single retinal cell does not signify anything as simple as that. It will signify perhaps red at a particular point in the visual field. As signals pass through higher levels of collation their meanings become more and more dependent on the signaling state of other units. A single signal might well mean ‘whatever object type has been assigned third level salience status is red’. In a man made computer the meaning of a signal going through a central processor is totally defined by the state of all the other gates in the computer and although this is likely to be less true of a brain I think it is doubtful that we need to invoke extra features of signals to account for binding. Ultimately, the segregational binding problem must be solved at the point of integration of the signals that give rise to a manifest experience. The red goes with the roses and the white with the lilies because the signals arriving at a juncture have the appropriate spatiotemporal relations at that point.

In summary, I suspect that the concept of segregative binding that concerns neurobiologists may be a mirage. The only binding we need is the organised unification of the elements of a complete experience. Prior to that signals have their necessary relational significance built in to them from the outset and maintained during the collation process without needing to invoke any relation between signals at distant sites of an embarrassingly non-causal sort.

Key point: two different problems are often discussed under the heading of ‘binding problem’, one involving ordered segregation and the other reunification, and they cannot both have the same solution.

6.80 The syntax of binding

One argument against the sort of approach I have been developing is that traditionally neurons have been thought to work on the basis of what is known as ‘integrate and fire’, or what might be accurately called summate and fire. In this model there is no difference in significance between any of the inputs to a neuron. Their relations are irrelevant. All that matters is that there are enough to overcome a threshold and depolarise the cell. As far as I can tell, however, this hypothesis has never been a formally accepted as the most consistent with experimental evidence. If anything it has been something of a null hypothesis, being the crudest model available, against which evidence for more sophisticated computational capacity has to be judged.

The conceptual problem I have with the summate and fire assumption is that, as for a computer, it does not seem to require any signals to have any ‘meaning to anything’ anywhere in the brain. All that is required is that the brain contains units that are sensitive to a certain level of input. A unit with the job of responding to the presence of a bicycle might fire with signals for two wheels and some pedals or it might fire for two handlebars and a saddle but it would not need to distinguish between the inputs, just add up to 3. The only meaning for add and fire is yes or enough. We want a meaning based on relations actually in the causal chain – that has to be relations of interaction at integration.

The alternative is to suggest that cells fire not just because the number of inputs reaches a threshold but because the inputs relate in a specific way. A bare minimum of depolarising inputs may still be required but the generation of an output, and perhaps the speed of generation of an output would be dependent on the ‘fit’ of the pattern. There are clearly potential advantages in having units that can recognise or respond to patterns like geometric figures presented in different ways although one would expect the integrating dynamics to be more complex and therefore potentially more costly to the system.

Current research into integration of signals in dendritic trees is chiefly limited to the combined effects of a very small number of signals under conditions that may be very different from those in a normally functioning brain. Evidence for a summative effect is well established but there is also evidence for summation being dependent on spatial relations between synapses mediating signals. The great variety of dendritic tree architecture between cell subtypes also suggests that a purely summative integration is unlikely. If all that mattered was the number of inputs there would not seem to be any need for different tree architecture in different types of cells.

My understanding is that the information 'something coming nearer' is fed into a single large neuron in a locust and it is this neuron that works out the situation from the pattern of post synaptic potentials. As far as I can see there is nowhere else in a locusts's brain where the experience, if there is one, could have a physical basis.

At present there is inadequate evidence to form a firm conclusion about pattern based responses in dendritic trees but at least it seems that this is a possibility, at least for some subpopulations of neurons with particular dendritic structures. If there are pattern based responses then it becomes legitimate to talk of some sort of syntax of integration of signals – a spatial format that generates meaning from relations.

Both Barlow and Marr commented that the final integration of elements in a percept seems to follow economical rules reminiscent of those involved in the syntax of natural languages. The simplest illustration of this for me is the idea that to have a percept of a red square only one additional piece of data is required to make the whole square red, just like the ‘fill’ command in a computer graphics programme. There is no need to have a signal for every pixel of the square. In fact there is no need to have pixels for the square, just signs for the four sided outline.

The idea that signals arriving at a row of synapses could generate the sense of something like a red square may seem preposterous, and much of the time it does for me. However, the idea that the activities of cells in a network might generate a red square is just as preposterous and causally incoherent to boot. The perhaps greater intuitive resistance to the idea of a row of synapses encoding the red square idea is I think simply a reflection of the more explicit nature of the model. The network model tends to allow one to slip back into some sort of functionalist conception in which the signals mean red square because they play a role relevant to red squares. The fact that at the point where these network signals are supposed to give rise to an experience they have no causal relation to anything and the causal relations in the past and future are either unreliably ascertainable or contingent on something that may not happen tends to get forgotten.

If we accept that the pattern of incoming signals in a dendritic tree determines the pattern of manifest experience we then have to try and work out what the correspondence rules might be. In a sense we have a task like trying to find out how a string of digital binary signals encode a picture displayed on a screen in a graphics software package. The key difference, however, is that we are not dealing with the sort of mechanism that generates pictures on screens for subsequent viewing through an optical apparatus. For this reason there is absolutely no reason to think that signals would be allocated to pixels in a display. The absence of pixels and voxels in experience is likely to be important in two ways. Firstly it provides economy (the economy of words) and secondly, pixels and voxels have boundaries that need to be encoded. If those boundaries are also encoded in a pixellated or voxellated way then we have infinite regress. If they are not encoded in voxels then there is no need for the primary image to be encoded in voxels – so it makes no sense to have voxels with boundaries with no voxels.

The rules may be as far removed as we like from any rules we know of that are used to generate pictures, as long as they are capable of operating in a consistent way capable of generating the range of manifest images we experience.

As indicated in the section on meaning I think there are strong reasons for following Marr and Barlow in thinking that the signals that are integrated into an image in experience work together rather in the way that words work together in a sentence. Thus I would see the signals that encode an experience of five red roses to be in a quasi-linguistic form: some signals for rose, some signals for ‘five of them’ and some signals for ‘make them all red’. Since no intermediary mechanism is proposed between the signals and what they seem like there is no problem about how the correspondence might be achieved in physical dynamic terms. There are no dynamics involved, just a correspondence between two incommensurable aspects of an interaction: what goes on and what is manifest.

As mentioned, the only constraint on the rules involved would seem to be that if many signals are integrated we do not start to get inconsistencies developing. The signals for red must not alter the perceived shape. The signal for five must allow the roses to be different in position and detailed outline. It would all seem horribly difficult but of course with words it does seem to work.

The requirement of consistency may be a bit more elastic than might first seem to be the case. Escher has demonstrated that we can experience pictures that seem consistent as a whole but in fact contain sensed relations that cannot coexist – like rings of stairs that are all going down. A number of traditional optical illusions that seem to create wavelike movements or distorted shapes that cannot coexist demonstrate the same phenomenon. Nevertheless, all these effects require quite subtle combinations of lines and in most cases we do notice that something is not right. The great majority of our experiences seem to be based on signals that create consistent appearances. Interestingly, at least from my own experience, it seems that in delirium or under the influence of drugs like anaesthetics or tyramine from cooked cheese, experience can take the form of impossible structures and movements that are terrifying in their tendency to suggest that they will swallow up the subject into the same impossibility. Thus the ‘syntax’ of binding signals into an experience seems capable of generating very complex consistent appearances but does have the capability to generate impossibilities.

Key point: binding of individual features into a manifest experience is likely to work in a way more like the way words work together in a sentence than the way pixels combine to form a picture.

6.81 Possible dynamics for binding

Trying to build a model for how electrical potentials at synapses might encode elements of experience for an acoustic dendritic mode with which they interact may seem unduly detailed speculation. However, I think the exercise is useful as a way of highlighting the constraints on what might be possible in a broader range of models.

Neural membranes work on the basis of cyclical depolarisation and recovery. Depolarisation effectively wipes out any preceding signals, so it seems reasonable to think of signals encoding elements of experience in time frames bounded by depolarisation. Even without depolarisation the effects of a single post synaptic potential will rapidly decay. For this reason I see rather little scope for elements of experience being encoded in time. Finely detailed timing may be relevant to the function of specific types of neuron in auditory pathways but the evidence that I have seen suggests that collative mechanisms extract patterns independent of fine time relations. It seems likely that for higher level collation for customary experience detailed temporal aspects relating to the sensory input itself will be ‘cleaned out’. Poeppel has argued that for auditory signals the complex temporal aspects of the primary signals are replaced by more standardized temporal coding at higher levels

Synchronisation of electrical activity is commonly found in the cortex and has been implicated in experience. I think the most likely explanation for this is that integration in cells involved in customary experience involves signals that have previously been locked into a standardised time sequence, so that the significance of each is independent of time of arrival (since all arrivals are synchronised) and purely dependent on spatial relation. This may prove to be oversimplified but it is hard to see how very much information can be encoded in time of arrival anyway. In the traditional summate and fire model time of arrival is not even relevant. Whatever the dynamics of integration are there does not seem to be much scope for arrival falling into more than a very small number of time subframes within the frame determined by firing.

For these reasons I think it is reasonable to assume that electrical signals at synapses must encode elements of experience through their spatial relations. This might seem implausible if one looks at the structure of a dendrite, which, close up, looks a bit like the stem of a meandering tradescantia plant with odd shaped knobbles at irregular intervals. However, one might think the same when first faced with a clarinet, that has holes in unevenly spaced places and lots of wiggly external apparatus attached to these. As modern architects have shown, very simple dynamic effects like holding up a roof may best be achieved by a mass of strange detailed substructure.

Encoding in spatial relations fits naturally with an interaction with an acoustic mode, since acoustic modes are inherently associated with patterns of spatial relation. The question is how we should expect the relation to give rise to meaning.

Having discarded the optical relation that links pixels to meaning there is another well understood relation that might be expected to serve better in the context of a waveform. That is the Fourier transform, of which the Gabor transform is a special case. The advantage of this sort of relation is that a change in a single term will have an effect on the whole, rather than just a change in a pixel in a corner. This is reminiscent of language where a change in tense of one word can completely change the meaning of a statement. Moreover, these transforms create asymmetries rather than adding bits, which seems in keeping with the general thrust of fundamental physical relations.

On the other hand, Fourier transformation does not readily generate discontinuities, of the sort we seem to experience when perceiving an object. It seems we are looking for something that may not, and probably should not be expected to, have a counterpart in the mathematical relations we find useful in analysing dynamics alone. We seem to need something that will complement the collational processes described by Hubel and Weisel. If a cell in visual cortex has responded with a signal that means a line from top left to bottom right then we need a way of signals input to the subject giving rise to the sense of a line. This may seem mathematically a tall order but it may be that we are not even dealing with mathematics in the usual sense, since all that is being generated is the sense of a line, not a line in causal dynamic terms. We have no right to think that any particular relation is implausible if it going to be that unfamiliar in its nature. We need to remember that nothing in the outside world is actually ‘like a line’ because dynamic space is not like anything in that sense.

We may seem to be peering into a dark bottomless pit here in terms of trying to understand a relation that maybe is not even expected to be mathematical. I think we need to take seriously the nature of the problem but I do not think we need to give up entirely. It seems reasonable to suggest that in a system like a brain that sends around discrete signals we are looking for a way of economically encoding elements of experience like shapes and colours that may bear some resemblance to language. It seems likely, for instance, that we encode a nearly round shape with a signal for roundness plus further signals to indicate deviation from roundness. This means that in an experience not involving close attention to one element there may be a lot of signals encoding many elements in outline only. For a following experience in which we focus hard on one object we may have many more signals relevant to that object, in terms of distance, outline, colour composition etc. and fewer signals allocated to background scenery.

The next issue that comes to mind is that such a system would seem to need to allocate different sorts of significance to different signals. Just as words are divided into nouns, adjectives and verbs and then into mass nouns, count nouns, animate nouns etc. there needs to be a basis for a signal giving colour to a shape rather than, for instance, size. This suggests to me that there must be what one might call registrational rules that determine how an incoming electrical signal relates to the putative acoustic mode or equivalent. An obvious way to do this would be for the mode to occupy a complex manifold such as a branching tree and have different parts of the tree relating to each other in different ways. In a standing wave in a branching structure any signals into the main ‘stem’ would be likely to have a different significance from those at the tips of branches. Another analogy that may be relevant is that of the left thumb hole in a woodwind instrument. Opening this hole does not just change the pitch of the note played up one semitone, as most holes do, but instead it has the capacity to change the note by an octave.

Unfortunately, as yet we have virtually no experimental evidence to go on in terms of even the dynamic implications of relations between signals but there is some evidence from studies on post-synaptic integration in individual neurons which suggests that the impact of one signal on another is dependent on proximity in relation to dendrite branches. As might be expected there seems to be more interaction between nearby signals than between distant ones. That may illustrate another way in which integration may follow rules that are perhaps the opposite of what pixel based signals would be like. If proximity means silence then in a sense we would expect ‘large’ to be encoded in a small distance.

Again, the difficulty we are faced with is that all experimental work is likely to tell us is the dynamic effect of signal integration, not what the pattern is like to the subject. Nevertheless, unless there is some rule based relation between these two, however upside down or inside out it might seem then it is hard to see how the system would function usefully.

Key point: Formulating a dynamic account of binding of features in experience is a novel and daunting exercise. However, an acoustic model may give some simple clues as to how testable models might be built.

6.82 Dimensions and degrees of freedom

An issue that has long been of interest to epistemology in general is the degree to which a human subject is circumscribed by parameters that predetermine its experiential scope. It seems rather likely that the number of dimensions we sense the world in is laid down by parameters of the subject. As indicated above, a cell may have 50,000 degrees of freedom in its input in the sense of having 50,000 synapses, each of which might mediate a signal. However, these degrees of freedom will be different from the number of dimensions of the experience, in the sense that some will encode alternative values for one aspect of experience and others will encode alternatives for other aspects. These will not be the dynamic dimensions of time and the three spatial dimensions but will include colour, temperature, odour and even emotional value. On the scheme proposed , the human subject will have its particular sort of experience partly because a highly organized pattern of many signals is available to it, but also because it will have a complex ‘way of receiving’. A negative charge can only receive a push or pull signal from another charge. An acoustic mode residing in a complex architecture is likely to be influenced by or ‘receive’ signals is much more varied ways.

One point that arises out of this is that if signals encode meaning in a spatial array it is not possible for ‘dimensional matching’ to apply. Time, colour and acceptability will all have to be encoded in space. This emphasises the fallacy of having a true homunculus with inner relations resembling external world relations. In a Turing machine the tape carries the input in one dimension but the programme written into the machines memory stacks will triage signals in a way that can allocate them to as many dimensions as are needed. We should expect the cellular subject to do something similar if the dynamic implications of the integration are going to be useful in the way that they are in a computer.

It is also conceivable that cellular subjects are set up to experience things in a much more specific way, of relevance to how we have adapted to our environment under the forces of natural selection. Arnold Trehub has written intriguingly on the illusion that the moon is larger at the horizon than at the zenith of the sky. The interesting conclusion is that we may have our experience of the world preprogrammed to fit in with a flat surface beneath our feet and a wide flattened dome over our heads. Because this dome is not actually spherical we misinterpret the size of the moon. Although one can think of other explanations like comparison against trees on the horizon, Trehub shows that none of these is convincing. If such a preprogrammed framework for experience exists it may of course be imposed at the collational stage and not at the experiential stage but I think both possibilities must be considered.

A further issue about the syntax of binding is to what extent we should expect it to be the same for each cell. One would certainly expect there to be similar ground rules at least for cells with parallel functions. Yet if each cell is designed to respond in a different way to a pattern of signals ‘broadcast in workspace’ then presumably there will be differences in the dynamics of integration and these may well be reflected in differences in what is experienced. This need not be a problem because for any given cell, as long as it experiences signals for red square the way it experiences signals for red square then there will be no confusion. This leads to a rather surprising variation on the ‘inverted qualia debate in which it is proposed that what I sense when presented with red is what you sense when presented with green. I think there is a real possibility that different cellular subjects will each experience a given colour differently. This is not something that will ever be open to test but this does not, I think, render it meaningless, because one day we may be able to see the structural differences that go with different sorts of experience.

If, as suggested earlier, cellular computation varies according to whether it is adapted to responding to a 1 in 100 subset of inputs or a 1 in 2 set of inputs, then we might reasonably suppose that the sort of experience entailed, if there is one for both, would be quite different. Similarly one might expect things to be different for cells that work on a more or less additive basis and cells that compare subsets of inputs through some form of subtraction, perhaps involving inhibitory potentials at synapses. An experience based on a mode tuned to interact in a particular way with certain patterns of input would be more relevant to the latter. Subtraction would seem to be necessary for the sense of agency – comparing input with expected input. Maybe this sort of cell gets customary experience, including a sense of appropriateness or belonging to self.

Key point: binding of features in experience seems require not only degrees of freedom in terms of numbers of inputs, but also in terms of allocating these to different sorts of experiential meaning.

6.83 How long is an experience ?

There has been a longstanding debate as to whether during waking hours we have a continuous experience of the world or whether we have a sequence of brief experiences. In strictly experiential terms the nature of this question may be unclear. We may feel that we are experiencing continuously but that may almost be a tautology since it is hard to make sense of experiencing not having an experience.

What is of more interest is the question of the temporal aspect of the relation between the immediately proximal dynamics that are associated with a manifest experience and the pattern of the experience. A continuous model would seem to imply that there is some continually present subject for which the pattern of experience is determined by the input at any given point in time. A discontinuous model would imply that input is ‘chunked’ into episodes in some way, either by a feature of the input signals or a feature of the subject. That would include the possibility that the subject is itself very short lived and only ever has one pattern of experience.

The immediate problem with the first, continuous, model is that when we come to immediate dynamic interactions there is no such thing as a point in time. Everything occurs within a finite spatial and temporal domain. The spatial domain receiving an input in a neuronal dendritic tree is defined by the structure of the tree. That leaves the temporal domain unspecified. In order for there to be any dynamic interaction at all we need a finite time span but that time span must be short enough to separate what we experience as the present from experiences in the past that no longer obtain, other than as material from memory. If we consider other models for the domain of the subject the problem is the same.

This problem is hardly ever considered by those in neuroscience but it is very familiar to those involved in fundamental physics. Fundamental physical processes involve complex harmonic oscillations and, put simply, an oscillation has to have time to go through at least two cycles before it can have a frequency. Interactions that might appear to be instantaneous must involve a finite time. Thus when a radio wave is generated by a tuned circuit several cycles are involved in what would appear to be an instantaneous indivisible event.

It seems that we need some temporal span to define what gets into the experiential pattern of ‘now’. In terms of signal input at synapses, I can think of no basic physical parameter that would provide this. It may be of note that at the quantum level, as when a photon passes through a diffraction grating, the temporal span of interaction is defined by the entity being informed by its environment – the photon. The wave function for the photon defines a brief span during which interaction is non-trivial. This suggests that maybe we should be looking for a time span for an experience that is determined by the subject.

If a subject of customary experience is an acoustic mode in the dendritic tree of a neuron then we have a temporal framework, since the mode will have a frequency, but if it is a continuous standing wave we do not seem to have any specific time span in the way we do for a photon. There would be nothing to distinguish, or move from, one now and the next.

One possibility is that the acoustic wave is not a standing wave, but rather a single pass travelling wave passing from one end of the dendritic tree to the other. This poses various problems, including the question of why such a wave should propagate in a particular direction at a particular time.

Another possibility is that the mode is a standing wave but is short lived. The obvious possibility is that the mode exists between somatic spiking but each time the soma spikes it is abolished, to be reset when the cell repolarises. This would make the mode operate within the time frame of the post-synaptic integration that we think is the basis of neuronal computation. It seems reasonably plausible that spiking should abolish an acoustic mode since it will produce a major transient change in both electrical and mechanical state of the cell, which will include the dendrites because the spike back propagates throughout the dendritic tree as well as forward propagating down the axon.

This is the model that seems most plausible to me at present. In metaphorical terms I would liken it to the lifting of the sustain pedal that is indicated at the end of each bar of a number of Chopin waltzes. During the bar the whole piano resonates to the chord harmonies but at the end of the bar all the acoustic modes in the strings are damped, to be recharged when the notes of the next bar are played.

Another possibility that may need to be considered is that although the acoustic mode that receives electrical inputs may be continuous the individual quanta of energy that make it up may have brief time spans. Quanta will be continually being gained and lost by the mode. However, it is quite difficult to see how this would work, or even whether it is theoretically a valid option.

A model of the subject that indicates that it exists for no more than about 20 milliseconds may be worrying to those who believe in the idea of an enduring self. However, we know that whatever part of the brain provides the substrate for a subject it is at least in part transient. The molecular composition of nerves is undergoing constant turnover. The sense of enduring selfhood can only be based a narrative fed in by signals from elsewhere in the brain and there seems to be no way in which we can be sure these are reliable. There is no way that a transient self can have a direct sense of being transient because it could have no sense of not being there before or after it existed. Similarly, for any model of this sort, we would not expect interruptions of experience to be experienced since there would not be any incoming signals encoding ‘this is an interruption to your experience’.

Key point: the continuous nature of experience may be illusory. A likely possibility is that experiences, and perhaps the subjects of experience last for a few milliseconds.

6.84 How do we talk about experience?

A central tenet of the complementarity based approach I am proposing is that an account of ‘what it is like’ or what is manifest in phenomenal experience for any subject within a human brain adds nothing to a sufficiently detailed biophysical dynamic account in terms of what causes what. The ‘phenomenal’ account is merely a first person version of the third person version that is the dynamic account. That is not to say that what is manifest in experience is epiphenomenal or non-causal, simply that the causality would be exhaustively covered by an ideally detailed biophysical account.

The situation where this claim seems most in danger of breaking down is when we talk about aspects of our phenomenal experience. It might seem that if we are specifically talking about phenomenal aspects, those aspects must contribute something specific to the causal chain. I hope to show that there is no problem here, or at least that the problem is soluble, and in the process explore some reasons why we might have come to think of a ‘mind-body problem’ in the first place.

I will consider three aspects of phenomenal experience that present rather different issues. Firstly, there is perhaps the simplest and most intuitively puzzling question of how we can talk not just about types of sensory input but the detailed characteristics of these inputs, like the shimmering intensity of a primrose yellow satin cloth, without the actual phenomenal feel contributing to the causation of our talking. Secondly, there is the more general issue of how we can ‘talk about’ phenomenal experience at all if the phenomenal aspect is ‘redundant’. At this point I will wheel in the philosophers’ zombie, who, although ‘physically identical’ to us, has no phenomenal experience so would seem ill-placed to talk about it, and still more ill-placed to express puzzlement over it, yet, I will argue, can be expected to do just that. Thirdly, there is the question of how we can give an account of the richness of a single experience, which raises some practical issues that probably fall within the scope of current experimental psychology.

A point to establish at the beginning is that it would have been a mistake to entitle this section ‘How can we talk about experience?’ as if talking about experience were an ‘option’ that we can ‘choose’ of our own ‘free will’. If we are asking how we are asking for an account of what causes what and ‘option’ and ‘choice’ are, to my understanding, pseudodynamic concepts that have no place in such an account. All that we can ask is how talking about phenomenal experience comes about. This breaks down in to two rather different questions: What goes on in a brain in an instance of talking about experience? and How did a combination of genetic evolution under Darwinian pressures and cultural evolution leave us with nervous systems that talk about experience via such goings on?

If we couch the problem in bald monochrome terms of talking about sensory inputs it does not seem to amount to much. If a nervous system gets in input from an auditory sense organ that signifies ‘a sound’ then producing speech responses that correlate with such inputs presents no obvious difficulty. Computers running voice recognition systems seem to do a lot better than this without us worrying about them talking about the phenomenal character of their inputs.

The problem becomes more apparent if we start to think of complex examples of ‘what something is like to us’ such as the shimmering of yellow satin, or the interminable deep gnawing of root canal toothache, or the way Casals playing Bach sounds like him talking to you personally. What biophysical account is going to capture these different characters?

The first response to this is that we have reason to believe that the brain is complex enough to have a wide enough repertoire of biophysical goings on for there to be different goings on for every finest nuance of phenomenal experience and pathways to generate different outputs to go to speech centres for each one. However, to get the range of phenomenal characters we have to ascribe a particular character to a complex pattern which, for reasons I have given already, I think must constitute a single fundamental complex indivisible relation between a subject entity and all the goings on that impinge on it causally. It may be significant that although the language of biophysics should be able to describe such a relation it may well not do so in such terms simply because the equations involved would be insoluble by current computational means. They may even be what Roger Penrose has defined as non-computable and this may have deeper implications that I shall return to later.

In short, there should be a dynamic biophysical sense in which a particular pattern interacting with a subject ‘is like something’ in just as fine detail as in phenomenal terms it ‘is like something’. All the detail of the phenomenal pattern can be accounted for in terms of detail of the interaction of manifest with dynamic so we do not need the phenomenal to do extra work (quite apart from the fact that if it did it would have to be classified as extra dynamics).

But that does not quite seem to deal with the fact that when we talk about the phenomenal aspect we seem to be talking about the phenomenal, not the dynamics that, in James Ladyman’s terms, are ‘ontologically co-dependent’ with it. But here we have to go back to the discussion of what we mean by ‘about’. In the section on language I came to the conclusion that for something to be about X is ultimately for it to seem to be about X. The causal relations to X need not fulfill any consistent expected role. X may not exist, as for a unicorn or Edna Everidge. Moreover, we only have one causal relation to consider here, for both the phenomenal and dynamic aspects of a particular experience. I am pretty sure that there is no fact of the matter, in causal terms, whether when I talk about what a toothache is like I am talking about the phenomenal or the dynamic version. In terms of what it will seem to be about to a subject within my brain, it will seem to be about what I am familiar with, which is the manifest phenomenon.

So that seems to settle the matter. If there is both a phenomenal and a dynamic ‘what it is like’ and things are about the aspect they seem to be about and the phenomenal is there, waiting, manifest, then this was a false problem. Except that I do not expect anyone to be satisfied with that. There still seems to be a nagging sense that the talk really must be dependent on the nature of the phenomenology. This is where I think the philosophers’ zombie (PZ) comes in handy. We need to have some reason for thinking that the PZ version of me will ‘puzzle’ over the distinction between the phenomenal and the dynamic when for the PZ there is no phenomenal. The above analysis may be right up to a point. Moreover, I have already explored instances of our concepts appearing to be about something but there being nothing that could be that something. The PZ may have a pseudoconcept of phenomenality. But we want some clue as to why that should be.

Key point: a simple causal account of how we talk about experience does seem to be available, but without some supporting justification it seems implausible.

6.85 The difference between knowing and self-knowing

In the section on knowledge I described how our understanding of the goings on in the outside world is based on inferences about operation of dynamic dispositions made by our ‘knowing machines’. Although I have reservations about ‘enactive’ theories of experience I think there is no doubt that our understanding of dynamic processes occurring in space and time relies heavily on moving both ourselves and bits of the world around to see how the dispositions play out. We use ourselves as rulers. We use rulers as rulers. We build up a knowledge of goings on in a spatial framework using inputs from vision, touch, auditory, vestibular and proprioceptive senses, in combination with a sense of time based largely on internal biological clocks, but also external clocks.

Using these resources we build up a concept of a ‘physical’ world in which patterns can be placed in space because we have parallel pathways that give us spatial knowledge. In one way or another all of these resources have to be calibrated by moving about. In contrast, our sense of time does not require any moving about.

It is clear that the human brain has the capacity to infer some aspects of its own internal dynamics. I can infer that I was thinking about coffee this morning. I can be aware of my ‘mind’ generating full blooded ‘inner voice’ versions of the chorus of ‘Those Were The Days’ with the initial slowing rubato varied over a wide range that I can then recall. I can ‘action-replay’ in my mind Mo Farah crossing the finishing line of the 10,000 metres faster or slower. What is common to all of these is that I have a sense of the timing of the internal dynamics but any sense of space refers not to the internal dynamics but to some external dynamics that they may be ‘about’.

Moreover, even the sense that these internal goings on are dynamic in a physical sense that involves space is an assumption based on educated argument. My memory of seeing the branches of a tree a few minutes ago I construe as being dependent on an ‘input’ to my sensory apparatus because if I cover my eye to block the input route I see no branches. However, my contention that the content of an experience is determined by the content of an input, rather than a throughput or output, is merely an extrapolation from this. If I consider my experiences all I seem to be able to deduce from what they are like is their temporal sequence. They are, perhaps, just a sequence of patterns in nowhere.

This lack of any spatial reference for knowledge of internal dynamics suggests that we have, as part of our knowing machine, a self-knowing machine that draws inferences in a quite different framework from that used for the outside world. What is inferred is not even ‘goings on’ or dynamics, but simply a sequence of patterns. It is not difficult to see how this would work. Instead of moving around, a part of the brain A can gain information about the brain’s internal sequences by comparing the sequence of inputs from one part B with the sequence of inputs from another part C that also influences the sequence of inputs to B. Instead of probing the world by changing spatial relations and seeing what comes in a part of the brain may change its own sequences of outputs to other parts and see what comes back in. If B always sends an input to A just after C does then the inference can be that B outputs are caused by C outputs, except that the involvement of ‘input’ and ‘output’ can only be inferred through considered analogy with our relation to the outside world.

We therefore seem to have brains that can infer what is going on in the world in two quite different frameworks of reference, with perhaps just enough clues to make us think that these frameworks ought to be aspects of some single general framework, since the vast majority of the time internal patterns and external dynamics seem to correlate in a reliable rule-based way.

Note that the discussion in this section has so far said nothing about the phenomenal side of patterns. It is an analysis that can apply equally to a computer that we do not expect to have any relevant phenomenality. My computer tells me what the clouds are like in Zermatt, through a net webcam. It also tells me that ‘Microsoft Word has unexpectedly quit due to an error’. The computer will tell me where Zermatt is by flagging the webcam on Google Maps but it will not tell me where Microsoft Word is.

Computers are designed to have their knowing and self-knowing machines operate in mutually exclusive domains. However, it is likely that our own knowing and self-knowing machines evolved through the usual random Heath Robinson biological pathway of borrowing bits from here and there wherever it seems a little bit advantageous. (Over a hundred proteins with quite different functions in the immune system all derive originally from ‘borrowing’ copies of one gene and then tweaking them.) In addition there seems to be something soft wired about the human brain that naturally leads to a tendency to ‘explore’ in novel ways. Perhaps these two aspects are even linked. Perhaps it is the soft-wired nature of our knowing machines that makes us unique in having cross-talk between world knowing and self-knowing functions. Perhaps in other creatures essential housekeeping self-knowing routines are kept out of the Baarsian ‘workspace’ of customary consciousness. Even in us some aspects of world knowing, like vestibular and proprioceptive senses, and of course the internal sense of time, seem barely to make it to the level of phenomenal experience. All sorts of housekeeping self-knowing functions like temperature and adrenaline regulation are, of course, also carried out by the hypothalamus.

From this I think we can conclude that there are perfectly clear operational/computational reasons why a zombie human species should evolve with a nervous system that deals with patterns within two quite different reference frameworks and, because soft wiring may help behavioural adaptation during life, Darwinian forces may well deliver a species that sometimes tries to analyse one sort of data in terms of the other framework. Thus we can fully expect a zombie human species to have identified the mind-body problem as the divide between inferred spatiotemporal dynamic patterns in the world and sequences of ‘patterns in nowhere’. We might also expect in time for a zombie Noam Chomsky and a zombie Daniel Dennett both to deny that it is a problem, for more or less opposite reasons. Others will continue to puzzle and find new reasons for doing so, and call it a ‘hard problem’.

One might think that computers do not puzzle, but when my computer, instead of saying ‘MS Word just quit’ goes into a crash loop I think it may be doing much the same as the consciousness studies buff. It keeps asking a question in the wrong context and keeps getting no answer it can make use of so keeps asking the question again – which seems pretty much like puzzling.

But, you may say, the zombies are distinguishing two sorts of pattern in dynamic terms. They are not distinguishing the dynamic from the phenomenal. However, I do not think it is that simple. For patterns generated internally all the zombie will be able to deduce is ‘what they are like’ as patterns and when they occurred. That may not be what we think we interpret ‘phenomenal’ to mean but at this point I think the arguments in the previous section can be allowed to kick in and complete the explanation. If there is a phenomenal what it is like for some subject in a brain then ‘phenomenal’ will seem to refer to that what it is likeness. (If not, it will not.) This may still be a pseudoconception but we have a very satisfactory set of reasons for understanding how it might come about.

Still concerned that the story is not finished? How could a zombie raise the arguments I am raising about the difference between what it is like phenomenally and what it is like dynamically? Perhaps the problem is that our conception of non-phenomenal what it is like involves inappropriate introduction of senses of physical interactions that come with inputs – like pushes and pulls and maybe not blueshes and pinklls. The assumption that we can intuit a ‘veridical’ God’s eye view of what physical dynamics are like is very hard to erase.

We should expect pseudoconception here because the standard talk is of what things are like, when they are not. Conflation of manifestation and dynamics is the default we all slip into. All we have is manifestation so we interpret what it is like or what things are like in manifest terms. We also have the pseudoconcept of me, which is very much not what we portray it as – a human body.

6.86 Some further zombic considerations

The concept of a philosophers’ zombie (PZ) has been open to various interpretations and modifications but the basic definition is of a creature that is physically identical to us at all levels but who has no phenomenal experience.

The first point that this raises is that if, as I have suggested, physical dynamics, or goings on, consist of instances of operation of patterns of disposition to entrain experiences under appropriate conditions then this definition is self-contradictory. The ‘physically identical’ zombie’s brain will by definition instantiate appropriate conditions for experiences so a non-experiencing zombie is not physically identical to us. Although I think this is an important point I do not think it necessarily affects the usefulness of considering whether a PZ would talk about experience if a PZ had dynamics that were mathematically identical to ours but which did not entrain experiences according to the same rules. In fact our reliable evidence for what appropriate conditions for experience are extremely limited, maybe just to our ‘own current experience’ so we can reasonably entertain a full range of possibilities from solipsism to panpsychism, including the idea that some human brains support experiences but some do not. The implausibility of this is not really a concern here because we are simply trying to clarify, by breaking things down into components, our own ability to talk about our experiences.

The next point is that if my suggestion that there are multiple local subjects of experience in brains is true, or in fact if any theory such as that of Baars, in which a subdomain supports experience, is true, then in a sense we should consider our own human frames to be PZs. I think it is extremely implausible that human beings have experiences, even mine. I think human beings are populated by experiencing subjects somewhere inside their brains. Given that as far as we know the manifest effects of experience (qualia if you like) are never carried over from one causal interaction to the next one the system as a whole ought to be considered a PZ. The discussion may all seem to get circular at this point but I think there are still useful lessons to draw.

Daniel Dennett has repeatedly expressed his puzzlement over the idea that ‘phenomenal’ means anything over and above what science can describe and so denies that the PZ makes sense. His position might seem to be similar to that laid out in my second paragraph above. However, it is quite different in that he seems to deny that there is such a thing as a first person account of how the world is felt by a subject. This is driven by his insistence that there is no receiving Cartesian subject involved. Instead he sees experience as arising in a functionalist way from computational events seen in terms of ‘role in the world’ or input-output relations. I have given reasons why I think this cannot work. My position is not that there is nothing in addition to the purely mathematical patterns of physical laws, but rather that the additional effects that these patterns dispose to, that must be seen as intrinsic to physics, should not entail further dispositions that go beyond those of the said physical laws.

Dennett complains that he has found no clear exposition of what phenomenal means and, in a way not uncommon with philosophers, suggests that this may entitle him to be sceptical about there being any meaning. However, what in effect Dennett seems to want to have is a description of the phenomenal in the sort of dynamic terms that we use for talking about tables – perhaps ‘What does it do?’. As indicated, because we have no dynamic framework for knowing about our experience we cannot expect to have such a description – in a sense that is what the definition of the phenomenal is that Dennett is searching for.

So I return to the idea that if the PZ has the same internal self-knowing machine as we do, with no spatial reference frame, in addition to an external world knowing machine with such a frame then both should be equally in a position to discuss the curious dichotomy between the dynamic world outside and the first person aspect of what experience is like. But we still have nagging doubts.

What I think may be useful here is to introduce the piecemeal PZ. The piecemeal PZ is relevant partly because it has features in common with computers, but also because it has a wider relevance. The real challenge to the sort of model I have been proposing is a PZ in which each cell has been replaced not by a ‘non-experiencing’ unit that has exactly the same dynamics, with all inputs been integrated through an interaction with some single dendritic entity such as an acoustic mode, but by a series of units that integrate information piecemeal in the way that computer processers do. If the result of all this piecemeal integration is, in computational terms, identical to that of the human dendritic tree then the tendency to talk about phenomenality should not be changed because the system as a whole will have indistinguishable dynamics. However, there would be no place where there was, even in dynamic terms, a first person perspective on a full pattern of experience.

The piecemeal PZ is of crucial relevance because a neuroscientist might argue on the basis of current knowledge either that post synaptic integration in a dendritic tree should indeed be seen as a series of piecemeal integrations, either at an ionic/molecular or a dendritic branch level or that there is no fact of the matter about this. Without the proposal of an acoustic mode to interact with all incoming electrical signals one would probably be very sceptical about the idea that there was any real first person perspective on a full input pattern in dynamic terms.

Worryingly, this argument suggests that we have no reliable evidence for our sense of ‘seeing a full scene’ having any real basis at all. It could be just an illusory ‘narrative’ clunking through a piecemeal machine, experienced by nobody. Yet, this does not seem to make sense either. How do we find a way of showing that we cannot be piecemeal PZs?

I do not think that as yet there are well constructed arguments. Nevertheless, a clue may come from Roger Penrose’s claim that human thought is non-computable. That is to say that the input-output relations somewhere in the brain simply cannot be accounted for in the piecemeal terms of a computer or Turing machine. For those, like Penrose himself, who take his evidence for non-computability as strong, this could be an important foothold in the above dilemma. I am uncertain, but encouraged that at least Penrose feels the argument is strong.

It may turn out that we can show that what happens in input integration in a cell, perhaps especially a cell that is performing some sort of comparison task, rather than an additive one, simply cannot be achieved with piecemeal steps. There is at least one simple argument available. Piecemeal steps, if set up in a simple convergent fashion, as would be true for a series of molecular or dendritic branch events, would not achieve an input-output relation that depended on all the internal relations between input elements. Granny would not be reliably distinguished from Auntie Betty. To make sure that all input interrelations were reflected in a comparison we would need a sequence that made use of a push down stack or Turing tape. That is to say that results from comparisons of single inputs would have to be put into a memory store for retrieval for further checking against other inputs. Even if this would not be non-computable in Penroses’s terms it would be totally implausible in terms of known neurophysiology.

This point allows us at least to say that there are very good reasons for thinking that a piecemeal PZ with our neuroanatomy simply would not evolve. If we accept that complex pattern comparison is something that neurons do at a single pass (and there does not seem to be much alternative even at a network level within the time constraints) then we do not have to worry about a piecemeal PZ puzzling over its non-experience because such a thing cannot exist. In other words, the existence of a single indivisible relation between inputs and some receiving entity in a dendritic tree may actually be a necessary condition for anything like a human brain existing. In dynamic terms there must be a first person view of a whole scene and if we consider that in terms of inputs in terms of how they are influencing the subject we have a meaning for phenomenal that will seem to be about what experience ‘is ‘like to me’ for any me to which such an input is like something – which is all that we are asking for.

We can also consider the possibility that Penrose is indeed right and that integration in a single cell does not just require its own micro-Turing machine to compute but is actually non-computable. This would not be so surprising since we are dealing with the interaction between a field and a mode which in some interpretations at least will involve an infinite number of degrees of freedom. I strongly suspect that the precise dynamics of this are non-computable. We then have the prospect of a brain being a network of units capable of solving non-computable problems, which presumably would imply that the whole system could solve non-computable problems.

Maybe we have considered the evolution of intelligence from the wrong perspective. Maybe the cell is the sophisticated processor, capable of solving non-computable problems. Perhaps vertebrate and insect brains are just add-on Turing machine type systems that provide more sophisticated collation of sensory inputs, motor co-ordination and memory storage.

At this point I will leave the general topic of how we come to talk about our experience. I have satisfied myself that an organism with the brain dynamics that we have would be expected to puzzle over the ‘mind-body problem’ just as we do. It is extremely hard to construct a knock down argument for this but my conclusion is that if we follow through on all the uncertainties raised by our intuitions they can all be satisfied once we come to see what the context of the discussion really is. As a spin off I am tempted to think, like Penrose, that we may have reason to believe that we have underestimated the computational power of the individual integrating units in a brain (which I see as cells and Penrose, together with Hameroff, has seen as groups of cells or ‘hyperneurons’).

Key point: detailed exploration of the context of the problem suggests that there is a plausible explanation for our ability to talk about our phenomenal experience without having to invoke causal processes beyond the biophysical dynamics we expect to obtain.

6.87 How do we talk about richness?

The foregoing analysis may raise a query about my previous claim that the content of our experiences is likely to be encoded in space rather than time. In the last section I have emphasised the fact that we can only use time to probe our internal patterns. However, there is no conflict here, just some complexity. My contention is that the content of an experience, both in terms of what it seems to be about and what it is like to have, is encoded in the spatial relations between inputs to a subject. That content will include senses of space, time, colour, smell and loneliness. For each instance of a pattern temporal relations are essentially those of synchrony. However, the timing of each these spatially distributed input patterns can be inferred by a self-knowing system with a built in clock and will give rise to a sense of when the experience was had. Much of the time this is used to tell us when an external event happened but it can equally be used for self-examination. This encoding of when we have experiences in signals handled by the self-knowing machine will then give rise to signals decoupled from the timing they are ‘about’ and fed in synchronously with other signals to a spatial input pattern to a subject that will encode ‘I thought about Delhi on Wednesday’.

Another way of looking at this model is to say that if the content of an experience in terms of what it is about and what it is like to have is encoded in its spatial relations then there is no room left to use these spatial relations to tell us where the experience is occurring. A spatial pattern cannot be both about what it is about in the outside world and about its own distribution.

A possible conclusion to draw from this is that although our self-knowing machines may be able to probe the timing of experiences it may not be able to probe the structure of the patterns involved. This creates a new problem for talking about experience if by structure we mean to include what might be called ‘richness’. It seems that although we do not know where our experiences are we do seem to have some idea how complex or detailed they are.

There is a caveat here, however, in that there is a major debate going on as to what that level of richness is – is it accounted for by, say, 50 bits, or 50,000 bits? We seem to have good reasons to limit the range to about this but it is a very wide range. The “Grand Illusion’ school would have it that we grossly overestimate the richness of a single phenomenal experience. On the other hand Victor Lamme maintains that our (visual) experiences are just as rich as they seem. If that needs as many bits as a photoshop file it looks to be beyond the capacity of the visual input system but I suspect even Lamme would agree that we could bring the bit count down to somewhere near 50,000.

On the model that I am following there is no way in which the number of elements in an experience could be known to the self-knowing system because there is no probing mechanism available. However, even in the case of knowing when our experiences are I suspect that the inference mechanism does not depend solely on direct real time comparisons. As for mental arithmetic I think we are likely to build inferences up from ‘off the peg’ calculations that have been acquired during infancy and childhood and maybe even fetal life and sleep. Thus when I multiply seven by eight I find fifty six readily available long before my brain has time to add up to fifty six. If I ask how many sides a triangular tile has you will say three without needing to count.

If we postulate that there are certain banks of cells that have the job of feeding signals into a Baarsian workspace, to be experienced by banks of subject cells, and we allow for connections via the sort of mordant loop mechanism mentioned earlier then I think we have the basis for assessing the ‘likely richness’ of experiences. If our self-knowing systems have built up tagging systems that associate the firing of a particular cell with the firing of twenty cells one level down in a hierarchy and each of those with a hundred cells a further level down, including the sorts of cells we use for counting numbers, then the system may give the ‘inferential guess’ that a current experience is being fed by twenty thousand inputs.

In fact we struggle to guess even that well. I might look at the carpet and think that since there are four diamond patterns in my attention I must at least have sixteen inputs to do the lines and five to do the coloured diamonds and background and so on. It becomes pretty clear that this is not a direct probing of any one experience, since our attention is constantly wandering, but rather a surmised reconstruction from multiple experiences held in short term memory overlain by a lot of rather dubious theoretical steps.

This raises the point that whatever probing of the richness of experience goes on, it is likely to be fairly second hand. The possibility also arises that the self-knowing machine only has access to cells one or two steps further back in sequence to the cells actually sending inputs to subjects. Maybe the self-knowing system will attempt to probe a bank of cells holding data that are not manifest all the time but are ready for immediate use.

This sort of discussion is quite closely relevant to neuropsychological studies, such as those of Lamme’s group, that try to distinguish between patterns present in phenomenal experience and patterns to which we have ‘cognitive access’ in the sense of being able to retrieve for verbal reporting or other behavioural responses. I am doubtful that with current resources that we will be able to determine the true richness of a single experience, for the reasons given but I do think these studies can give useful information about the way data are sorted and stored in the steps immediately leading up to manifestation in customary experience.

In conclusion, I think that there can be ‘dirty’ ways that the self-knowing system can guesstimate how rich an experienced pattern might be. Moreover, these may not be that much dirtier than the ways the system judges timing for most everyday purposes. A much more precise mechanism for evaluation of timing does seem to be available for judging ongoing internal events when required, but this is almost certainly just a minor elaboration of the mechanism used for judging the timing of external events. The assumption is added in that the experiences coincide with the external events, which only gives trouble for odd phenomena like echoes and thunder and lightning. If dreams actually unfold at the rate they seem to unfold then presumably they are coming in to experience through a route that uses the same monitoring as for sensations from outside. If not, then that would suggest an alternative route. I am unclear on the consensus on this issue.

The analysis so far may be all that there is to be said but there is a lingering sense in which we seem to know our experiences are rich, at least in very broad qualitative terms, directly. That would seem to imply that this is not an issue of what the knowing machine can infer so much as what the knower itself can be aware of. This raises the issue for the next section – that of manifest evaluation.

6.88 The problem of manifest evaluation

A key feature of the concept of manifestation in experience is that the input to the system will be experienced in a way that reflects the mode of interaction. Thus, a negatively charged particle will experience the influence of another negatively charged particle as having a quality related to what we call a push, rather than a pull. A signal input to a human subject that is interpreted as blueness will have that interpretation because of the relation involved. What is a more difficult issue is how the subject will interpret a pattern of inputs that we associate with some significance or value as a pattern.

The simplest example might be a cell rather of the sort studied by Hubel and Weisel which has inputs that each signify one of three lines forming a triangle. The cell may have the function of firing when three lines are present in the appropriate relation – joining at the vertices. This firing will give rise to higher levels of collation handling signals that carry the concept of triangle and these may entrain an inner voice percept of the word triangle. What is unclear is whether or not we should expect our first cell to have a sense of triangularity.

At first sight it would seem this should be so. However, the alternative possibility is that the sense of triangularity will only be had by cells further down the collation path that are receiving signals that, together with signals for some lines signify ‘this is a triangle’. The issue becomes a little easier to conceive, perhaps, if we substitute grandmother for triangle. Does a cell receiving a diagnostic pattern of features of granny’s face have a sense of this being grandmother or is that sense encoded further down the line only after our first cell has diagnosed her identity?

Another example would be an assessment of value in terms of good or bad. If faced with two pears, one perfectly ripe and the other rotten, does a cell receiving all the relevant visual pattern information get a sense of one being good and the other bad or is this encoded further on. Is our customary experience of a good and a bad pear the manifestation of an input at a point where visual signs are provided along with evaluation signs, such that we sense the evaluation in the same way as we sense colours, or is the evaluation generated in the process of receiving the colour signals.

Why I think this may be a difficult question to answer is that if we are dealing with a single indivisible relation between subject and input then the distinction between interpretation of individual signals and evaluation of patterns of signals becomes unclear. On the other hand if the integration of signals involves some sort of comparison involving effectively a subtraction process then it may be that different aspects of the relation of subject to input are interpreted according to very different rules.

One of the things that I think may cause confusion in any analysis of these issues is the potential mistake of thinking that a subject to which something is manifest must have the function of recognising that something. If there are multiple subjects then this need not be the case and I think this is a strong argument in favour of multiple subjects operating in parallel. This confusion is evident in some discussions of what are called grandmother cells. The original concept of a grandmother cell was a cell which specifically fired when grandmother was in view. It sometimes seems to be assumed that this must therefore be the only site of experience of grandmother. However, taking into account the fact that signals from the primary visual cortex tend to get sent to thousands of receiving cells, it seems reasonable to assume that thousands of sites have an input indicating grandmother but rather few of these will have the job of responding.

Our experiences appear to be very rich and that means that a very large number of responses might be relevant for any one experience. If there is only a single subject in a brain it is hard to see how it could mediate all such responses at once. Thus faced with a visual experience of two pears relevant responses might include the identification of the type of fruit, whether one looks good to eat, whether the other looks good to eat, whether they are near enough to reach and so on. It seems that our brains generate and handle all sorts of responses of these types in a very short space of time. A single subject would have to do that serially, presumably with arrangements of inputs changing to reflect the response required each time. With multiple subjects all the different responses can be achieved in parallel.

The implication of this is that any given subject may be making all sorts of evaluations in its interaction with its input but may only be set up to respond to one of these. Put simply, a subject may sense that one pear is good to eat but respond by indicating that it is near enough to grasp. A cell whose function is to recognise motorbikes may still have a sense that grandmother is in the room. It will just not respond.

Some might argue that it is inappropriate to assign specific tasks to individual cells in this way and that such tasks should be seen as being carried out by nets of cells. I have a strong suspicion that this is little more than an obfuscating tactic that has become popular because it avoids having to have a clear theory. Computational tasks in the brain can only be carried out by post synaptic integration, which happens in each cell separately. There may well be redundancy in the sense that several cells perform the same task, but they are still each performing that task. If the task is divided up into subtasks performed by different cells each cell is still performing a task and some downstream cell will still have to complete the task on the basis of input from the first cells.

Puzzles about the sites of manifest evaluation of triangles and grandmothers’ faces are perhaps simply a reminder that we understand little about the coding language that makes proximal patterns in nerve cells seem to be about external dynamic patterns in spacetime. Puzzles about manifest evaluation in terms of good and bad may raise a deeper issue. It is of interest that we attribute such values both to world events and to our inner events. We say the day in the park was pleasant, but also that the experience was pleasant. We say the pain was terrible. I find the noise of a drill unpleasant and I also find my tinnitus, which is no the noise of anything, unpleasant. Moreover, assessment of pleasantness or unpleasantness does not seem to be something a knowing machine could infer by some computation.

I wonder whether we need to consider the possibility that certain sorts of evaluation at least are not ‘intentional’ or ‘about’ some distal pattern in quite the way that much of experience is so much as direct interpretations of the proximal pattern that is manifest. If a single celled organism is a subject of experience it will presumably benefit from evaluating its input for its own sake, rather than as an indirect indication of some distant events. Moreover, I wonder whether that evaluation might include the issue of richness discussed earlier. If a knower wants to know how rich its input is there is a sense in which it would be inappropriate for it to rely on some supporting knowing machine to do the job. The knower has the answer directly manifest. The knower may have neither a spatial nor a temporal reference frame in which to assess its pattern but that may still leave the possibility of assessing if it is rich, or good, or perhaps true. Moreover, I have talked throughout of knowing machines drawing inferences but for these inferences to involve evaluations as simple as recognising a triangle we have to invoke the computational powers of the individual cells that actually do the work in a connected net.

My sense is that maybe these queries link up with the query about non-computability. Earlier I suggested that truth was some internal matching or consonance but did not enlarge. What I think may be important is that we are not looking just for a matching of the sort

X = Y , otherwise written as X-Y = 0

We seem to be wanting a matching across levels of abstraction, so that a set of data can be compared with some conceptual criteria about some property to be evaluated. We seem to be looking for something more like the relation between a variable, a function and an argument, as in

X = A(Y) or X – A(Y) = 0

Meaning that the actual pattern X is that we should expect from applying some concept A to a context Y.

If all this is to be done with an input pattern we seem to need X, A and Y all to be encoded in inputs but in such a way that each is interpreted as playing different roles. The very simple analogy would be the fingering of G’ on the oboe, which consists of index, middle and ring fingers of the left hand encoding generic G and the left thumb half covering a hole to encode the higher octave version ‘ to give G’.

My limited knowledge of logic suggests to me that this may imply an interrelation between function, argument and value within a single computational step of a sort not possible within the limited lambda calculus made use of in Turing machine style computers. This may be ill-informed but I have a suspicion that the real difference between us and computers lies in the immediate biophysical relations within an individual cellular computational (or perhaps super-computational) unit.

6.89 A workable Panexperientialism.

The recent philosophical literature frequently makes it appear that panexperientialism, or panpsychism, is an outmoded and ridiculous conception. Most often it is simply referred to as something that is so implausible that any argument that supports it must be considered a blind alley. The reasons for rejecting panexperientialism are thus rarely given.

The alternative belief, that experience is confined to higher animals, may stem in part from a slightly liberalised version of a religious view that only man has knowledge, and thus only man has experience. This was, after all, Descartes’s view. What may be more significant, however, is the common assumption that having manifest experience is associated with a different pattern of causal dynamics to those one would have without it. This is bound in with the idea of free will. To many the idea of having a ‘mind’ or psyche in the sense of having experience is intimately associated with the concept of agency. If something can perceive it can change what goes on from the way things were going to be. Thus to allow a rock to have experience is to allow a rock to be able to change goings on from the way they were going to be.

As indicated previously, this idea of being able to change things from the way they were going to be is incoherent. They were not going to be that way, as it turns out. We have no reason to think that the existence of manifest experience changes what was going to happen. I think it makes sense to think that those dynamics associated with manifest experience can only be those dynamics if there is manifest experience. Experience is not an optional extra. It is an intrinsic part of the goings on. But that does not mean that it alters the dynamic rules of the goings on from those reasonably well defined in physics.

On this basis I cannot think of any reason to object to panexperientialism, because it would not predict anything unexpected otherwise. The most significant criticism of panexperientialism is that it makes no testable predictions at all. It could reasonably be called redundant or even meaningless as a hypothesis. The reason why I think it is neither is that it has an important element of parsimony in that it releases us from requiring any explanation why some things should have experience and not others.

It also provides a prima facie case for thinking that we have no reason to seek a single subject of experience in a brain. We are more likely to be looking for those subjects whose experience matches most closely in content the sort of experience we discuss as ‘our experience’ and which seems to inform not just our verbal behaviour but our considered reactions to stimuli of various sorts.

There does, nevertheless, seem to be a legitimate question regarding experiences that might occur in the dendritic trees of cells not immediately involved in the high-level multimodal collation that seems to be involved in ‘customary’ experience. It might be that essentially all neurons have rather similar acoustic modes, or whatever might turn out to be a more accurate account of the subject, but that for neurons in primary visual cortex their input produces an experience of utmost banality. One would probably not even expect it to seem visual, since inputs need only signify that they relate to vision if other sings are dealing with sound or touch. On the other hand it might be that a large proportion of neurons just do not make use of the sorts of modes that might host an experience even remotely similar to customary ones.

In this respect one might speculate that neurons using a summate and fire integration might not make use of coupling to an acoustic mode whereas those making use of a comparative pattern-based integration might do so. It might be that in the cortex only the pyramidal-type cells made use of acoustic modes, with interneurons managing on simple electrical summation.

While it is tempting to think that there might only be a few specialised cells with the sort of acoustic modes that could interpret an input of signals in terms of customary experience this raises the uncomfortable need for a sudden acquisition of an entirely new functional property in a biological membrane. To an extent it seems more plausible that if acoustic modes do modulate electrical events in cell membranes that they have done so since the evolution of motile single celled organisms. A twist on this story would be the idea that single celled organisms may be endowed with acoustic membrane subjects but that in higher animals all but a very few cells lose this property which is left to those cells at the highest level of integration, rather like Leibniz’s dominant monads. I think this unlikely, but it is always dangerous to assume consistency in the way biology behaves!

Some advocates of panpsychism like to suggest that sentience or consciousness of some form occurs at many different levels, from fundamental particles up through cells to organisms and even social groups. My view is that as soon as one is dealing with an aggregate the idea of sentience or consciousness becomes unworkable. So human beings as a whole are not, after all, conscious: only their cells. This is something that many find hard to accept but I see no reason to stick to the idea of a whole conscious human creature just because that is our culture-informed intuition. We need to change our culture, in the way we often do as science advances. Modern biology has moved a long way from equating a creature, in the sense of a group of cells that tend to move around together, with a life unit. Pregnant mothers, bee hives, transplants and slime moulds all in different ways show that there need not be a one to one relation. If we are looking for a fundamental life unit a cell, or perhaps in more general a membrane bound biochemical microsystem, looks a more promising candidate. Nicholas Humphrey has suggested that higher organisms differ from protozoa in that their ‘souls’ have become buried within sophisticated information transduction machines. To avoid evolutionary ‘jumps’ it is easiest to suggest that the inner soul is a single cell, just as it is for protozoa. It may even be that around inner cellular souls in brains are many cells that have lost the level of sentience of protozoa, just as queen bees are surrounded by workers that have lost the ability to procreate.

When it comes to potential ‘experiences’ of inanimate entities, despite the fact that we will never know what they are like, there is a tantalising sense that their content ought to bear some understandable relation to the dynamics involved. Leibniz talked of the simplest of monads as having appetition, as if one would expect even a photon to have a sense of something like desire. Although this may be a ‘meaningful’ speculation it is, however, fraught with difficulties. As indicated in section 1 one might expect the entire lifespan of a photon to be one Whiteheadian occasion of experience and at least in its own frame of reference this will be a span of no temporal duration at all. In contrast an occasion for an electron might last billions of years. The problem is not so much that this raises any more counterintuitive issues than the situation for the putative human subject, but it becomes extraordinarily difficult to identify what ‘understandable relation’ could mean. It seems very unlikely that we would be able to predict likely relations for human subjects on the basis of a more firmly established relation for simpler entities, as is the usual practice in science.

Some people may always object to the idea that inanimate entities could have any sort of experience. However, I think the earlier discussion in relation to zombies shows that it this objection may have no legitimate traction in terms of metaphysics. It seems that any metaphysics based on the sort of causal relations we think of as obtaining in a ‘physical’ world has to assume that in some way the pattern of casual powers of the rest of the universe are manifest to each and every dynamic dispositional packet, in a way that depends on the dynamic dispositional law that packet instantiates. Causal connections must, at least in some operational sense, be manifestations. And after all, even the excursion beyond solipsism to the postulation of experiences for other people, is only justified by parsimony and a desire not to be alone in the world. If nothing else, and despite the fact that I try to avoid ‘-isms’, I think we need to allow the world to follow some form of ‘panmanifestationalism’ to complement a pandynamicism.

Key point: Leibniz was on the right track.

6.90 The nub of complementarity

Perhaps the key message of the complementarity-based view of the world is that we should take manifestations seriously. The second message is that manifestations are not some strange anomaly amongst a material world but rather they are intrinsic to any coherent causal account of the world. They are the effects from which all causal patterns must be inferred. Manifestations may seem to be insubstantial in that they can only be described in terms of relations to other things, but then the same applies to ‘physical’ causal dynamics. The inter-relational nature of both is something we need to see as natural and inevitable.

The distinction between ‘physical’ and ‘mental’ in the sense of ‘material’ and ‘phenomenal’ has not served us well. Phenomenal experience is intrinsic to physicality – it is the read out. Moreover, mental activities are physical in the sense of having dynamics. Experience has a ‘what it is like’ aspect, but then so does the outside world in the sense of being manifest. The distinctions between the dynamic and the manifest and between the dynamics of the outside world and the manifest signs we have of them, on the one hand , and the dynamics of our neural processes and the rather different signs we have of these, on the other, are more subtle but ultimately more robust.

What we need to develop in the scientific and philosophical programme of understanding relations in the world is a comprehensive account that includes both relations outside the human body and those within. Understanding the events within is made difficult by the complex collational process that links distal to proximal events in the genesis of manifest ideas. We need to understand, at least in broad terms, how the human knowing machine generates the phenomenon of aboutness and build it in to the causal dynamic account.

We also need to try to find out what the rules are that govern relations between proximal events in our brains and the manifest ideas they are associated with. So far it has been difficult to get started on such a programme because we do not know what those most proximal events are. My analysis suggests that the answer may be complicated but I see it as possible in the not too distant future to begin to formulate what might be called the final neural correlates of consciousness – the dynamic patterns immediately associated with experience.

I think there are strong reasons for thinking that these most proximal patterns take the form of a determinate field interacting with a dynamic quantised energetic mode. To have a place in a relevant biological causal chain I think that mode has to be acoustic but I would be very prepared to accept that there may be a better option. Although I see the subject as a quantized mode I do not think that at the level of dynamics relevant to customary experience formal quantum-mechanical analysis is necessary, any more than it is for understanding the behaviour of a violin string.

I see a panexperientialist view as being the most parsimonious but I also see it as being something that can never be confirmed experimentally. That means that we can never discover what it might be like to be a photon or a free electron. Nevertheless, on the basis of a more operational ‘panmanifestationalism’ I think it may be reasonable to build theories relating to the minimum requirements for a mode that might have the sort of narrative of experience in space and time that we are familiar with. I think it entirely reasonable that we might be able to synthesise modes in a laboratory with these properties but we will never know whether or not the experience feels the way it does to us. I suspect it may be relatively easy to build arguments that make it implausible that the modes occupied by fermions like electrons should have experiences in any way similar to ours, particularly if it becomes reasonably clear what a human subject mode might be. Descartes conceived of thinking (experiencing) things as massless and unextended in the sense of not excluding other things from their space. Although it would probably be largely fortuitous he might turn out to be right.

I believe that there are overwhelming reasons for thinking that there are many human subjects in a single brain and that these may be both dispensable, as individuals, and very ephemeral. In that regard I differ markedly from Descartes and Leibniz but in many other respects I think these two thinkers reached insights that are still out of reach to many in the relevant fields of science and philosophy.

My complementarity-based view rejects the traditional concept of objects much in the way that the structural realism of Worrall and Ladyman does. However, where Ladyman would deny the existence of individuals in any sense my view retains the concept of individuals as dynamic subjects. The ‘object-world’ is fused into a universal totality of relations but we need that totality to be manifest to individual subjects to explain first person experience. That does not mean that the subject has to be something separate, with traditionally conceived qualitative properties. The subject is an asymmetry in the universe that brings with it a relation. The subject is in a sense nothing more than the dynamics of its relation to the rest of the universe. That relation is one of ‘perception’ as Leibniz indicated. What both Descartes and Leibniz may have underestimated is the complexity of this perception relation. It is perhaps doubtful that we are yet anywhere near the clear and distinct ideas that Descartes felt should be the basis of our understanding of the world. Our perception may take the form of Leibniz’s apperception in the sense of having salient content, but it remains a Heath Robinson affair of internal signs for external dynamics built through the vagaries of evolution.

Sections 7 & 8 under editing

9 A Twenty-first Century Monadology

The following is a twenty-first century version of the Monadology of Gottfreid Leibniz, based on the translation by Robert Latta. Wherever possible Leibnz’s original text has been preserved, to illustrate just how much of it is consistent with, and presages, a coherent modern view of the universe. Some additions have been made to form a specific link between Leibniz’s ideas and modern physics. Some changes have been introduced to ease potential conflicts between Leibniz’s view and recent knowledge. These may distort Leibniz’s position but it is hoped that Leibniz would accept such ‘poetic license’ as being in the spirit of his purpose. Perhaps the greatest change is in the suggestion that ‘creation’ and ‘annihilation’ of certain Monads may occur continually and that not all Monads are necessarily permanent.

The Monadology has been re-caste in a framework which does not make use of words such as god, divine or deity, as much as anything because there is little agreement about what such words mean in a modern world. A naïve anthropomorphic view of god is deliberately avoided. However, an aspect of Leibniz’s approach that is maintained is the proposal that nothing knowable exists or evolves except as a relation to the universe as a whole. In a sense all things of which we are aware are merely patterns or asymmetries within the Universe, or in Spinoza’s terms, Nature, and have no meaning except in the context of the whole. If Nature is equated with God, as in Spinoza’s view, then there is no difficulty in maintaining a concept of God. Nevertheless, this does represents a significant departure from the usual interpretation of certain aspects of Leibniz’s own view.

In some areas Leibniz’s eighteenth century language may lead to a false impression that sweeping statements about the role of the universe in causation, in the actions of individual elements, and in the relationship between possibility and reality, are being made in order to justify a theistic view. However, I suggest that careful reading of the original text shows that Leibniz is addressing issues that are just as important to a coherent formulation of physics as they are to any metaphysical overview, and that his arguments are cogent in both contexts.


1. The Monad, of which we shall here speak, is nothing but a simple substance, which may in some cases enter into compounds. By 'simple' is meant 'without parts.' This Monad is a single mode of harmonic oscillation, occupied by quanta of energy.

2. And there must be simple substances, since there are compounds; for a compound is a collection or aggregate of simple things.

3. Now where there are no parts, there can be neither extension nor form in any sense that would allow divisibility. These Monads are the real atoms of nature and, in a word, the elements of things.

4. There is no way in which a simple substance can be disassembled by natural means.

5. For the same reason there is no conceivable way in which a simple substance can be assembled by natural means, since it cannot be formed by the combination of parts.

6. Thus it may be said that a Monad can only come into being or come to an end all at once; that is to say, it can come into being only by creation and come to an end only by annihilation, while that which is compound comes into being or comes to an end by parts.

7. Further, there is no way of explaining how a Monad can be altered in quality or internally changed by any other created thing; since it is impossible to change the place of anything in it or to conceive in it any motion between parts of a sort which could be produced, directed, increased or diminished therein, although all this is possible in the case of compounds, in which there are changes among the parts. The Monads have no windows, by which it is meant that nothing can enter between parts of the monad so as to have a different relation to different parts of the monad. Other things can only have a single relation to the monad as an indivisible whole.

8. Yet the Monads must have some qualities, otherwise they would not even be existing things. And if simple substances did not differ in quality, there would be absolutely no means of perceiving any change in things. For what is in the compound can come only from the simple elements it contains, and the Monads, if they had no qualities, would be indistinguishable from one another, since they do not differ in quantity. Consequently, space being a plenum, each part of space would always receive, in any motion, exactly the equivalent of what it already had, and no one state of things would be discernible from another. Everything would be symmetrical.

9. Indeed, each Monad must be different from every other, either in position or momentum or both. For in nature there are never two beings which are perfectly alike and in which it is not possible to find an internal difference, or at least a difference founded upon an intrinsic quality. In this regard, it is now known that there are two forms of Monad, obeying two distinct sets of laws, corresponding broadly to the intuitions of M. Descartes. Monads obeying Fermi statistics are modes that can only be occupied by one quantum of energy (a fermion), which excludes all other quanta from its domain. This exclusion is the ultimate source of the compound phenomenon known as matter, or for M. Descartes, res extensa. Monads obeying Bose-Einstein statistics (bosons) are modes that can be occupied by any number of quanta. These latter Monads carry forces. They do not exclude other Monads from their domain. They are known to us as, for example, the elements of light, sound and spirit: for M. Descartes, res cogitans. All our knowledge of the Universe comes from these Monads of force. Thus the Monads of matter can only be known to us be inference. We may never be able to know their internal features, such as the perceptions and appetitions of which we will soon speak. Thus, this Monadology will be chiefly concerned with those Monads of force and in particular the Monads of spirit, with which we are directly familiar.

10. We assume also as admitted that every created being, and consequently the created Monad, is subject to change, and further that this change is continuous in each. A Monad is always in a form of progression.

11. It follows from what has just been said, that the natural changes of the Monads come from an internal principle, since an external cause can have no influence upon their inner being. This may seem puzzling, but what is meant is that there can be no mechanical intermediary which ‘steers’ or ‘rearranges’ the Monad; rather the Monad will always progress in harmony with every aspect of the Universe, as shall be described below, according to its internal principles.

12. But, besides the principle of the change, or progression, there must be a particular pattern of changes that constitutes, so to speak, the specific nature and variety of the simple substances. This is their mode of oscillation.

13. This particular pattern of changes should involve a multiplicity in the unit: in that which is simple. For, as every natural change takes place gradually, something changes and something remains unchanged; and consequently a simple substance must have internal relations varying in many ways, although it has no parts.

14. The passing condition, which involves and represents a multiplicity in the unit or in the simple substance, is nothing but what is called Perception, which is to be distinguished from Apperception or Consciousness, as will afterwards appear. In this matter the Cartesian view is extremely defective, for it treats as non-existent those perceptions of which we are not aware in the consciousness we normally discuss. This has also led the Cartesians to believe that human minds alone are Monads, and that there are no spirits of animals nor other Entelechies. As explained, this idea of Perception is to be most clearly understood for Monads of force. A related idea may apply to Monads of matter, but this may never be knowable.

15. The activity of the internal principle which produces change or passage from one perception to another may be called Appetition. It is true that the appetite cannot always fully attain to the whole perception at which it aims, being limited, for instance, by the speed of light fixed in the space-time metric, but it always obtains some of it and attains to new perceptions.

16. We have in ourselves experience of a multiplicity in simple substance, when we find that the least thought of which we are conscious involves variety in its object. Thus all those who admit that the soul is a simple substance should admit this multiplicity in the Monad; and M. Bayle ought not to have found any difficulty in this, as he has done in his Dictionary, article 'Rorarius.'

17. Moreover, it must be confessed that perception and that which depends upon it are inexplicable on mechanical grounds, that is to say, by means of figures, motions or any intermediary agents. And supposing there were a machine, so constructed as to think, feel, and have perception, it might be conceived on an enlarged scale, while keeping the same

proportions, so that one might go into it as into a mill. That being so, we should, on examining its interior, find only parts which work one upon another, and never anything by which to explain a perception. Thus it is in a simple substance, and not in a compound or in a machine, that perception must be sought for. Further, nothing but this (namely, perceptions and their changes) can be found in a simple substance. It is also in this alone that all the internal activities of simple substances can consist.

18. All simple substances or created Monads might be called Entelechies, for they have in them a certain perfection; they have a certain self-sufficiency which makes them the sources of their internal activities and, so to speak, incorporeal automata. Thus, there is no way that physicists can deny that anything wonderful that might be called intelligence, understanding, animacy or spirit must arise from the internal principles of Monads. It cannot arise merely from their ‘interaction’ since this interaction is merely an account of the operation of the internal principles.

19. If we are to give the name of Spirit to everything which has perceptions and appetites in the general sense which I have explained, then all simple substances or created Monads (at least Monads of force) might be called spirits; but as feeling is something more than a bare perception, I think it right that the general name of Monads or Entelechies should suffice for simple substances which have perception only, and that the name of spirits should be given only to those in which perception is more distinct, and is subserved by memory.

20. For we experience in ourselves a condition in which we remember nothing and have no distinguishable perception; as when we fall into a swoon or when we are overcome with a profound dreamless sleep. In this state a spirit does not perceptibly differ from a bare Monad; but as this state is not lasting, and a spirit comes out of it, a spirit is something more than a bare Monad.

21. And it does not follow that in this state that there is a simple substance without any perception. That, indeed, cannot be, for the reasons already given; for it cannot continue to exist without being affected in some way, and this affection is nothing but its perception. But when there is a great multitude of little perceptions, in which there is nothing distinct, one is stunned; as when one turns continuously round in the same way several times in succession, whence comes a giddiness which may make us swoon, and which keeps us from distinguishing anything. Indeed it now appears unlikely that Monads of spirit have a continual existence; rather they may come in to existence and disappear quite regularly. Monads of matter (Fermions) in most cases last for almost the lifetime of the Universe but Monads of force are transient. Some Monads of force may last for the lifetime of a living cell, but those associated with our consciousness might last either the waking day, or perhaps a much shorter time, being constantly re-created and annihilated during thought.

22. Since there is nothing in our understanding of the Universe that allows us to distinguish the present from the past or future it must be that the present for a Monad is its entire life span.

23. This suggests that the Monads of our consciousness last a fraction of a second.

24. And if we had in our perceptions nothing marked and, so to speak, striking and highly-flavoured, we should always be in a state of stupor. And this is the state in which the bare Monads are.

25. We see also that nature has given heightened perceptions to Monads within animals, from the care she has taken to provide them with organs, which collect numerous rays of light, or numerous undulations of the air, in order, by uniting them, to make them have greater effect. Something similar to this takes place in smell, in taste and in touch, and perhaps in a number of other senses, which are unknown to us. And I will explain presently how that which takes place in a spirit represents what happens in the bodily organs.

26. Memory provides the spirit with a kind of consecutiveness, which resembles reason, but which is to be distinguished from it. Thus we see that when the spirits of animals have a perception of something which strikes them and of which they have formerly had a similar perception, they are led, by means of representation in their memory, to expect what was combined with the thing in this previous perception, and they come to have feelings similar to those they had on the former occasion. For instance, when a stick is shown to dogs, they remember the pain it has caused them, and howl and run away.

27. And the strength of the mental image that impresses and moves these spirits comes either from the magnitude or the number of the preceding perceptions. For often a strong impression produces all at once the same effect as a long-formed habit, or as many and oft-repeated ordinary perceptions.

28. In so far as the concatenation of their perceptions is due to the principle of memory alone, men act like the lower animals, resembling the empirical physicians, whose methods are those of mere practice without theory. Indeed, in three-fourths of our actions we are nothing but empirics. For instance, when we expect that there will be daylight to-morrow, we do so empirically, because it has always so happened until now. It is only the astronomer who thinks it on rational grounds.

29. But it is the partial knowledge of necessary and eternal truths that distinguishes us from the mere animals and gives us Reason and the sciences, raising us to a degree of knowledge of ourselves and of the Universe. And it is this in us that is called the rational mind.

30. It is also through the knowledge of necessary truths, and through their abstract expression, that we rise to acts of reflection, which make us think of what is called I, and observe that this or that is within us: and thus, thinking of ourselves, we think of concepts of being, of substance, of the simple and the compound, of the immaterial, and of the Universe, conceiving that what is constrained in us is in it without constraint. And these acts of reflection furnish the chief objects of our reasonings.

31. Our reasonings are grounded upon two great principles: firstly that of contradiction, or impossibility, in virtue of which we judge false that which involves a contradiction, and true that which is opposed or contradictory to the false.

32. And secondly, that of sufficient reason, in virtue of which we hold that there can be no fact real or existing, no statement true, unless there be a sufficient reason, why it should be so and not otherwise, although these reasons usually cannot be known by us.

33. There are also two kinds of truths, those of reasoning and those of fact. Truths of reasoning are necessary and their opposite is impossible: truths of fact are contingent and their opposite is possible. When a truth is necessary, its reason can be found by analysis, resolving it into more simple ideas and truths, until we come to those that are primary.

34. It is thus that in Mathematics speculative Theorems and practical Canons are reduced by analysis to Definitions, Axioms and Postulates.

35. In short, there are simple ideas, of which no definition can be given; there are also axioms and postulates, in a word, primary principles, which cannot be proved, and indeed have no need of proof; and these are identical propositions, whose opposite involves an express contradiction.

36. But there must also be a sufficient reason for contingent truths or truths of fact, that is to say, for the sequence or connection of the things which are dispersed throughout the Universe of created beings, in which the analyzing into particular reasons might go on into endless detail, because of the immense variety of things in nature and the infinite division of bodies. There is an infinity of present and past forms and motions which go to make up the efficient cause of my present writing; and there is an infinity of minute tendencies and dispositions of spirits, which go to make its final cause.

37. And as all this detail again involves other prior or more detailed contingent things, each of which still needs a similar analysis to yield its reason, we are no further forward: and the sufficient or final reason must be outside of the sequence or series of particular contingent things, however infinite this series may be. In short, the history of a spirit is not algorithmic.

38. Thus if there is a final reason of things, it must be in a necessary entity, in which the variety of particular changes exists only eminently, as in its source; and this entity we call the Universe.

39. Now as this entity is a sufficient reason of all this variety of particulars, which are also connected together throughout, there is only one Universe, and this Universe must be considered sufficient.

40. We may also hold that this supreme entity, which is unique, universal and necessary, nothing outside of it being independent of it, this entity, which is a pure sequence of possible being, must not be constrained by anything and must contain as much reality as is possible.

41. Whence it follows that the Universe is absolutely perfect, if perfection is nothing but symmetry and certainty of being, in the strict sense, leaving out of account the asymmetries, limits or bounds in things which are limited. And where there are no bounds, that is to say, in the Universe, perfection is absolutely unconstrained.

42. It follows also that created beings derive their perfections, certainties or symmetries, from the influence of the Universe, but that their imperfections, uncertainties or asymmetries, reflect their own nature, which is incapable of being unconstrained. For it is in this that they differ from the Universe. An instance of this original imperfection of created beings may be seen in the natural inertia of bodies.

43. It is farther true that in the Universe there is not only the source of existences but also that of essences, of ways of being, in so far as they are real, that is to say, the source of what is real in the possible. For the understanding of the Universe is the region of eternal truths or of the ideas on which they depend, and without it there would be nothing real in the possibilities of things, and not only would there be nothing in existence, but nothing would even be possible.

44. For if there is a reality in essences or possibilities, or rather in eternal truths, this reality must needs be founded in something existing and actual, and consequently in the existence of the necessary Universe, in which essence involves existence, or in which to be possible is to be actual.

45. Thus the Universe alone (or this necessary being thing) has this prerogative that it must necessarily exist, if it is possible. And as nothing can interfere with the possibility of that which involves no constraints, no negation and consequently no contradiction, this is sufficient of itself to make known the existence of the Universe a priori. We have thus proved it, through the reality of eternal truths. But a little while ago we proved it also a posteriori, since there exist contingent beings, which can have their final or sufficient reason only in the necessary being thing, which has the reason of its existence in itself.

46. We must not, however, imagine, as some do, that eternal truths, being dependent on the Universe, are arbitrary and depend on a divine will, as Descartes, and afterwards M. Poiret, appear to have held. Necessary truths depend solely on the Universal laws of understanding and are their inner object.

47. Thus the Universe alone is the primary unity or original entity, of which all created or derivative Monads are products and have their birth, so to speak, through continual asymmetries of the Universe from moment to moment, limited only by the scope of the Universal being thing, of whose essence it is to be unconstrained.

48. In the Universal laws there is Power (or energy), which is the source of all. There is also Knowledge, whose content is the variety of the ideas, and finally Will, which makes changes or products according to the principle of the best. These characteristics correspond to what in the created Monads forms the ground or basis, to the faculty of Perception and to the faculty of Appetition.

49. A created thing is said to act outwardly in so far as it restores perfection, or symmetry, to the Universe, and to be passive, in relation to another, in so far as it influenced by this restoration of perfection, or symmetry. In this way Monads of force tend to be active and Monads of matter inactive. Activity is also attributed to a Monad, in so far as it has distinct and rich perceptions, and passivity in so far as its perceptions are confused or limited. Thus the Monads of matter are considered passive and the most active of Monads are those of spirit, rich in perceptions arising from memory.

50. And one created thing is more perfect than another, in this, that there is found in the more perfect that which serves to explain a priori what takes place in the less perfect, and it is on this account that the former is said to act upon the latter.

51. But in simple substances the influence of one Monad upon another is only notional, and it can have its effect only through the mediation of the Universal laws, in so far as the Universal laws, in regulating other Monads from the beginning of things, should have regard to that one Monad. For since one created Monad cannot have any mechanical influence upon the inner being of another, it is only by this means that the one can be dependent upon the other.

52. Accordingly, among created things, activities and passivities are mutual. For the Universal laws, applied to two simple substances, find in each reasons which require them to adapt the other to it, and consequently what is active in certain respects is passive from another point of view; active in so far as what we distinctly know in it serves to explain what takes place in another, and passive in so far as the explanation of what takes place in it is to be found in that which is distinctly known in another.

53. Now, as in the Universal laws there are an infinite number of possible universes, and as only one of them can be actual, there must be a sufficient reason for this.

54. And this reason can be found only in the fitness, or in the degrees of perfection as defined by the Universal laws, that these worlds possess, since each possible thing has the right to aspire to existence in proportion to the amount of perfection it contains in germ.

55. Thus the actual existence of the best that the Universal laws of probability describe is due to this, that these laws make it certain.

56. Now this connection or adaptation of all created things to each and of each to all, means that each Monad has relations which express all the others, and, consequently, that a Monad is a perpetual living mirror of the universe.

57. And as the same town, looked at from various sides, appears quite different and becomes as it were numerous in aspects; even so, as a result of the infinite number of Monads, it is as if there were so many different universes, which, nevertheless are nothing but aspects of a single universe, according to the special point of view of each Monad.

58. And by this means there is obtained as great variety as possible, along with the greatest possible order; that is to say, it is the way to get as much perfection as possible.

59. Besides, no hypothesis but this (which I venture to call proved) fittingly exalts the greatness of the Universal laws of connection, which are thought.

60. Further, in what I have just said there may be seen the reasons a priori why things could not be otherwise than they are. For the Universal laws, in regulating the whole have had regard to each part, and in particular to each Monad, whose nature being to represent, nothing can confine it to the representing of only one part of things; though it is true that this representation is merely confused as regards the variety of particular things in the whole universe, and can be distinct only as regards a small part of things, namely, those which are either nearest or greatest in relation to each of the Monads. It is not as regards their object, but as regards the different ways in which they have knowledge of their object, that the Monads are limited. In a confused way they all strive after the symmetrical and unconstrained, the whole; but they are limited and differentiated through the degrees of their distinct perceptions.

61. And compounds are in this respect analogous with simple substances. For all is a plenum (and thus all matter is connected together) and in the plenum every motion has an effect upon distant bodies in proportion to their distance, so that each body not only is affected by those which are in contact with it and in some way feels the effect of everything that happens to them, but also is mediately affected by bodies adjoining those with which it itself is in immediate contact. Wherefore it follows that this inter-communication of things extends to any distance, however great. And consequently every body feels the effect of all that takes place in the universe, so that he who sees all might read in each what is happening everywhere, and even what has happened or shall happen, observing in the present that which is far off as well in time as in place: sympnoia panta, as Hippocrates said. But a spirit can read in itself only that which is there represented distinctly; it cannot all at once unroll everything that is enfolded in it, for its complexity is infinite.

62. Thus, although each created Monad represents the whole universe, it represents more distinctly the domain or body which specially pertains to it, and of which it is the entelechy; and as this domain expresses the whole universe through the connection of all matter in the plenum, a spirit also represents the whole universe in representing this domain or body, which belongs to it in a special way.

63. The domain belonging to a Monad (which is its entelechy) constitutes along with the entelechy what may be called a living being, and along with the spirit what is called a cell. Now this domain of living being or of a cell is always organic; for, as every Monad is, in its own way, a mirror of the universe, and as the universe is ruled according to a perfect order, there must also be order in that which represents it, i.e. in the perceptions of a spirit, and consequently there must be order in the cell, through which the universe is represented in the spirit.

64. Thus the organic domain of each living being is a kind of divine machine or natural automaton, which infinitely surpasses all artificial automata. For a machine made by the skill of man is not a machine in each of its parts. For instance, the tooth of a brass wheel has parts or fragments which for us are not artificial products, and which do not have the special characteristics of the machine, for they give no indication of the use for which the wheel was intended. But the machines of nature, namely, living cells, are still machines in their indivisible parts. It is this that constitutes the difference between nature and art, that is to say, between the Universal art and ours.

65. And the Universe has been able to employ this wonderful power of art, because each portion of matter subdivided into myriad parts, of which each has some motion of its own; otherwise it would be impossible for each portion of matter to express the whole universe.

66. Whence it appears that in the smallest particle of matter there is a world of creatures, living beings, animals, entelechies, or souls.

67. Each portion of matter may be conceived as like a garden full of plants and like a pond full of fishes. But each branch of every plant, each member of every animal, each drop of its liquid parts is also some such garden or pond.

68. And though the earth and the air which are between the plants of the garden, or the water which is between the fish of the pond, be neither plant nor fish; yet they also contain plants and fishes, but mostly so minute as to be imperceptible to us.

69. Thus there is nothing fallow, nothing sterile, nothing dead in the universe, no chaos, no confusion save in appearance, somewhat as it might appear to be in a pond at a distance, in which one would see a confused movement and, as it were, a swarming of fish in the

pond, without separately distinguishing the fish themselves.

70. Hence it appears that each living body is full of other living beings, each of which has its dominant entelechy or soul.

71. But it must not be imagined, as has been done by some who have misunderstood my thought, that each soul has a quantity or portion of matter belonging exclusively to itself or attached to it for ever. For all bodies are in a perpetual flux like rivers, and parts are entering into them and passing out of them continually.

72. Thus the soul changes its body only by degrees, little by little, so that it is never all at once deprived of all its organs; and there is often metamorphosis in animals, but never metempsychosis or transmigration of souls; nor are there souls entirely separate from bodies nor unembodied spirits.

73. It also follows from this that there never is absolute generation nor complete death, in the strict sense, consisting in the separation of a spirit from a cell. What we call births are developments and growths.

74. Philosophers have been much perplexed about the origin of forms, entelechies, or spirits; but nowadays it has become known, through careful studies of plants, insects, and animals, that the organic bodies of nature are never products of chaos or putrefaction, but always come from seeds, in which there was undoubtedly some preformation; and it is held that not only the organic cell was already there before conception, but also a spirit in this cell.

75. The animals, of which some are raised by means of conception to the rank of larger animals, may be called spermatic, but those among them which are not so raised but remain in their own kind (that is, the majority) are born, multiply, and are destroyed like the large animals, and it is only a few chosen ones that pass to a greater theatre.

76. But this is only half of the truth, and accordingly I hold that if an animal never comes into being by natural means, no more does it come to an end by natural means; and that not only will there be no generation, but also no complete destruction or death in the strict sense. And these reasonings, made a posteriori and drawn from experience are in perfect agreement with my principles deduced a priori, as above.

77. Thus it may be said that not only the soul (mirror of an indestructible universe) is indestructible, but also the animal itself, though its mechanism [machine] may often perish in part and take off or put on an organic slough [des depouilles organiques].

78. These principles have given me a way of explaining naturally the union or rather the mutual agreement of the soul and the organic body. The soul follows its own laws, and the body likewise follows its own laws; and they agree with each other in virtue of the pre-established harmony between all substances, since they are all representations of one and the same universe.

79. Souls act according to the laws of final causes through appetitions, ends, and means. Bodies act according to the laws of efficient causes or motions. And the two realms, that of efficient causes and that of final causes, are in harmony with one another.

80. Descartes believed that souls or spirits cannot impart any force to bodies, because there is always the same quantity of force in matter. Nevertheless he was of opinion that the soul could change the direction of bodies. Leibniz believed that this could not be so because of conservation momentum.

81. However, if souls are modes of oscillation carrying force they can exchange momentum with matter. This exchange would have no intermediary, and as such could be considered consistent with Leibniz’s concept of pre-established harmony.

82. In rational animals, cells, so long as they are outside the nervous system, have merely ordinary or sensuous Monads; but for those that are within the nervous system, their sensuous souls are raised to the rank of reason by dint of the complex information they receive and their specialised use of the Universal rules of knowledge.

83. Among other differences which exist between ordinary Monads and spirits, some of which I have already noted, there is also this: that Monads in general are living mirrors or images of the Universe of created things, but that spirits also reflect the Universal rules of knowledge.

84. It is this that enables spirits to enter into a kind of fellowship with the Universe, and brings it about that in relation to them the Universe is not only what an inventor is to his machine (which is the relation of the Universe to other created things), but also what the inventor is to his apprentices, or indeed a father is to his children.

85. Whence it is easy to conclude that the assemblage of all spirits must be the highest embodiment of the Universal rules of knowledge, that is to say, the most perfect State that is possible, under the most perfect of rules.

86. This highest embodiment of the Universal rules is a moral world in the natural world, and is the most exalted within the Universe; and it is in it that the glory of the Universe really consists, for it would have no glory were not its greatness and goodness known and admired by spirits. It is also in relation to this that this highest embodiment specially has goodness, while the Universal rules are manifest everywhere.

87. As we have shown above that there is a perfect harmony between the two realms in nature, one of efficient, and the other of final causes, we should here notice also another harmony between the physical realm of nature and the moral realm of grace, that is to say, between the Universe, considered as origin of its own mechanism and as origin of the ethic of spirits.

88. A result of this harmony is that things lead to grace by the very ways of nature, and that this globe, for instance, must be destroyed and renewed by natural means at the very time when the government of spirits requires it, for the punishment of some and the reward of others.

89. It may also be said that the Universal laws must in some way be reflected in the laws of human behaviour. It is therefore reasonable that sins bear their penalty with them, through the order of nature, and even in virtue of the mechanical structure of things; and similarly that noble actions will attain their rewards by ways which, on the bodily side, are mechanical, although this cannot and ought not always to happen immediately.

90. However, it seems that the laws of the Universe have given us a sense of good and bad, yet we do not as yet know how we should best apply these principles. The best hope is that greater understanding of these laws will guide as to the best action.