What Does 'Experience' Refer To?
Any discussion of the nature of experience must presume that the word ‘experience’ actually has a referent – it refers to something. This is by no means universally accepted. My old neurophysiology tutor, Ian Glynn, found no reason to assume it, and Ian is not a man of narrow perspective. Whether experience has a referent might seem a technical linguistic matter but, on the contrary, I think it raises issues about just how careful we need to be when trying to understand our relationship to the world.
My own view is that it is useful to at least consider that the word experience might refer to something, although in the sense of a type of interaction or event rather than an object. However, even if we can agree on that it is not the end of the matter. We also have to be clear what we mean by ‘refer’. As indicated in the section on meaning in the essay Reality, Meaning and Knowledge, the relation of ‘referring to’ is not a simple one. It is not a causal relation, so there need not be any fact of the matter what a word refers to (however much analytic philosophers might think otherwise). Different instances of usage of a word by different people relate to different subsets of phenomena for all sorts of reasons. Agreed usage changes; the word bat was thought to apply to one sort of animal but we now know it applies to two quite unrelated sorts of animal. Even for central scientific terms like species or electron there are further problems. Both of these terms define subsets of phenomena within some broader subsets but break down within other broad subsets. (There is no fact of the matter whether or not two animals living a million years apart belong to the ‘same species’.)
All this is, nevertheless, fairly unremarkable and manageable in practice for the usual subject matter of science. The problem with experience is more serious because, even if it is a type of event, it is a peculiar type of event we call ‘mental’ (see section on mental events) which we cannot characterise in the way that we characterise external events like a man walking. We can give an experience a time and a rough position, inside a brain, but we cannot give it a size or shape.
To take this further it is instructive to consider how a brain comes to associate the use of a word with one or more of its own internal events. This must require memory. To refer to my experience an hour ago of enjoying a piece of cake my brain needs to have formed a pattern of modified connectivity that allows me to repeat a ‘thin’ version of the pattern sensations I had in association with the idea that these are a reliable representation of an event in my brain an hour ago. If, as seems likely, my original experience was part of a computational process that modulated my behaviour at the time then to also lay down a memory it seems that we need to postulate that branches of the axons feeding sensory data into the experience also fed the same data into a memory forming module that integrated these data with an ‘episode retrieval tag’. That indicates that we should not expect our ‘memories of experiences’ to be caused by the experiences (qua events) they represent, or at least if they are then something else is likely to be causing the immediate response to the sensory data. This should not be very surprising since it is not as if there were some unexplainable perfect match between our memories and our experiences – far from it.
This means that our memories and associated concepts of experiences are in fact based on something a bit like the input to a second digital camera placed behind the viewfinder window on a digital SLR camera. We are logging the fact that a certain pattern of signals was being output from somewhere in the brain of a sort capable of giving a ‘thick’ experience by inputting into the same part of the brain that gets the thin version when the memory is retrieved.
This conclusion has two important implications. One is that memories, whether ‘of a world event’ or ‘of an experience’ record patterns of output rather than input. A memory of an experience is a logging of the content of an output, not of the event that output generates. The other is that it gives no indication of how many events that output of content will be fed into. We already have the apparent requirement that it is two and since the brain seems to work with a high level of at least partial redundancy (and most axons go to 10,000 places) the default assumption is that it is a very large number.
This does not, however, mean that we should think that memories are ‘about’ outputs or that when I talk of 'the experience I had' I am talking about an output. For one thing, our sensed ideas are never about the signals immediately responsible for them. The situation is more complicated because the memory, or the word usage, is about an inferred event that includes the concept of 'it happening to me'. As indicated, the excitations of the cells whose output feeds in to an experience are not the events we are interested in, merely precursors to those events. It seems likely that similar, if not identical events might occur in sensory neurons during sleep but that influences on downstream neurons mean that the signals never generate experience.
This analysis shows just how careful we need to be in assuming that the relation of a word referring to something has a simple and uniform interpretation in terms of causal events. After all, philosophers like Fodor and Dretske have spent a lifetime trying to make a start at defining what that might be and have come to no clear consensus. It looks as if there is a peculiar difference between the pathways involved in remembering a mental event, as compared to remembering an external event. For the latter it seems as if signals come in from the event itself and cause the laying down of the memory. For mental events it seems likely that this does not happen, rather that the memory is formed on a causal sidebranch that reasonably reliably receives an equivalent pattern of signals to that which feeds into the mental event being ‘remembered’.
This might seem to cast some special doubt on the validity of our knowledge of mental events. It certainly seems to threaten the idea that our accounts of our own experiences cannot be wrong. It is often suggested that if I sense that I am seeing blue then I cannot be wrong about the fact that I sense that, even if I am looking at something green. I suspect this claim is very naïve, for the simple reason that all our knowledge is much more fragile than that (for reasons given in Reality, Meaning and Knowledge). The specific problem in this case is that we have forgotten to make sure we know what we mean by ‘I’.
The reality is that our knowledge of external events is just as indirect and inferential as it is of mental events, even if knowing about mental events involves a rather salient form of causal sidebranch. For external events we do not ‘get signals directly from the event’ either. We have to use inferences that are just as indirect and sometimes much more so. In a sense all our information about world events comes from sidebranches that involve the bouncing of photons off or emission of sound waves by protagonists in events that we never get direct causal influences from. Moreover, it is only by collation of multiple patterns of signals that we can infer the events that are responsible for the differences between those patterns. If we push physics to the limits and consider how we know black holes, far corners of the universe, single quarks, or Higgs bosons the amount of sidebranching gets very serious indeed.
All that said, the idea that a memory of a mental event is formed on a sidebranch that completely bypasses the event itself, does look to be special. What may mitigate it is the fact that a memory, as recalled, includes not only the pattern of content equivalent to that of the past event but a sense that that content contributed to further ongoing events within the ‘personal narrative’. The inference of the existence of the mental event that is the true content of the remembering is almost certainly dependent on collation of feedback signals drawn from processes that do indeed follow ‘experiences’ on the other sidebranches from that reaching the memory module. Just as we have feedback loops that make us feel the owners of our actions we almost certainly have feedback loops that confirm that input patterns to be remembered really did have the sort of effects on our emotions or behaviour we have come to expect.
The catch, however, is that this feedback is unlikely to give any clue as to how many mental events receive the input pattern that is copied to the memory module.
What we probably need to accept is that we have good reason to think that what we call one experience is likely to consist of a very large number of events all dependent on the same output pattern from a bank of upstream neurons. To give another analogy, our sense of events is based on something a bit like a ‘hit-counter’ linked to a website. Every time information is sent out from the website it counts a hit. However, the output might be to a computer used to project on a screen in a theatre with an audience of a thousand or copied on to another site later to be visited by a million.
Thus the limitation on our knowledge of our mental events, in addition to not knowing what size and shape they are, crucially includes not knowing how many there are based on a particular pattern of content. The intuitive claim that 'I know that I only have one experience at a time' is not supported by neuroscientific evidence!