My object of investigation is the grammaticalized expression of the notion “Speaker (S) wants Hearer (H) to do something, whereby S requests H´s agreement for that”, which I will be referring to as the requestive (see also Kuteva 2009). The requestive as category is an explication of a notion introduced by Jespersen ([1925] 1968) and for the first time explicitly defined in detail in Krabbe (in preparation), who assumes validity of the following sequence of stages in the grammaticalization of the verb please into the politeness adverbial marker please: (i) full verb (“my words please you”), (ii) conditional clause with the verb please (“if it pleases you”), and; (iii) politeness adverbial marker (“please, open the window”).

In the present investigation I will treat adverbial politeness markers such as please in English as dedicated requestive markers, i.e. linguistic expressions which are specialized for encoding a polite request; this does not necessarily rule out the existence of other, non-dedicated means of expressing a polite request in the same language, cf. the use of a modal verb in the past in an interrogative sentence in English (e.g. Could you open the window?).

On the basis of preliminary, empirical observations from languages spoken in Europe, Africa, Asia and Papua New Guinea as well as from the communicative behaviour of the speakers of these languages, I will propose that there might be a correlation between directness/indirectness of communication on the one hand, and presence/absence of dedicated grammaticalized requestives.

More precisely, I will hypothesize that there exist three major groups of languages as far as grammaticalizing the requestive notion is concerned:

1. Languages with no dedicated grammaticalized requestives and no other politeness markers (e.g. Papua New Guinea languages, African languages). The communication behaviour of the users of such languages is characterized by directness,
2. Languages with dedicated grammaticalized requestives (and often with additional, non-dedicated, i.e. multi-purpose linguistic expressions that have the function of politeness, among other functions, e.g. English as well as other European languages). The communication behaviour of the users of such languages is characterized by moderate directness;
3. Languages with no dedicated grammaticalized requestives but with other politeness markers, which often amount to a highly elaborate honorific system, whereby the requestive semantics is distributed – as a rule – across the whole syntagm containing the request (e.g. Korean). The communication behaviour of the users of such languages is characterized by indirectness.
   

Within urban ethnic minority communities, the adult migrant generation (Gen1) is widely treated as having relatively fixed or 'fossilised' non-native features in their English, whereas the local- born generation (Gen2) is argued to show little or no retention of such features (Chambers 2002; Sankoff 2002). Implicitly, a major accent shift is assumed to occur abruptly at the boundary between Gen1 and Gen2, i.e. non-local-born vs. local-born. Using data from first and second generation members of a bilingual Punjabi community in West London, and focusing in particular on the Punjabi-derived feature of [t]-retroflexion, I argue that:

1. Gen1 adults show sensitivity to the dialect contact situation and are able to modulate their use of retroflexion to some extent,
2. retroflexion is retained by many Gen2 speakers,
3. by the time retroflexion reaches younger Gen2 speakers, its use is sensitive to different linguistic (e.g. position and articulation) and social (e.g. class and network) factors, and
4. time depth and gender emerge as particularly significant forces in this shift in usage.

Rather than disappearing abruptly in the British-born group, retroflexion gradually acquires a new social function among younger Gen2 speakers. The findings suggest that dialect shift is not exclusively driven by the cognitive mechanisms of native acquisition but also by demographic and social factors (paralleling recent critiques of abrupt creolisation). In closing, I briefly discuss how the complex range of ‘hybrid’ dialect repertoires found across individuals points to distinct locations and trajectories within a shared social space.