04 Jun 2002

Jim Scobbie (QMUC)

A basic assumption of generative phonology, whether constraint-based or derivational, is that the grammar must generate all and only the members of a coherent and definable set of well-formed representations. This implies that two different speakers of the same language or dialect will have identical grammars despite their differing exposure during acquisition to phonetic, phonological and lexical experiences. In fact, given the sheer amount of language-specific detail which has to be mastered, it might be thought unlikely that speakers form identical mental representations of their shared language. The crucial question though is how trivial these differences are. Are they relevant just at a low level of idiolect --- responsible for a type of variation which can tell us nothing of broader cross-linguistic patterns? Or are they more significant, suggestive that our universal linguistic endowment should be approached as a flexible statistical problem-solving faculty rather than as a computational system with a definable list of universal rules and a definable set of target representations?

In order to approach this issue experimentally we have undertaken both experimental phonetic analysis and traditional phonological research. We have adopted an approach based on sociolinguistically-structured groups of subjects who can be expected to vary in their systemisation of the phonetic or phonological phenomena under consideration.

In previous LabPhon meetings, structured groups of subjects have been used to demonstrate socially-functioning, phrasal, non-lexically-contrastive aspects of phonetic competence by Docherty, Foulkes, Milroy and colleagues [1]. They have also been used by Scobbie, Stuart-Smith and colleagues [2] as a tool to disambiguate the social from the lexically contrastive functions of a family of phonetic features active in conveying a contrast.

The reason here for using a socially heterogenous subject pool is to model more closely the normal linguistic situation in which we live. Standard experimental phonetics and laboratory phonology tend to rely on subjects who are highly literate and who speak standard varieties, or similar. Such data will tend to support the assumption of a "coherent and well-defined" system, because standard linguistic varieties have a set of homogenising norms associated with them with which speakers comply. But such a well-defined situation is not applicable to many (or most?) speakers of English (to be specific), who master production systems which typically range from non-standard, local, vernacular varieties to something more like the standard, as well as the ability to handle yet further dialectal variants perceptually.

In this paper we will mainly adopt an approach based on phonetic analysis of data from a sociolinguistically-structured pool of subjects. We undertook a study of speech from 12 young adult speakers from the Shetland Isles, an insular community with significant numbers of incomer families. The speakers can be said to have been exposed to the same contrast system: specifically the English stop voicing contrast in word-initial position. They have not all, however, been exposed to just one phonetic system for cueing that contrast, insofar as it is possible to make this separation, and the various target systems are in competition at a phonetic level. The choice is between short-lag aspiration vs. long-lag aspiration and pre-voicing vs. short-lag aspiration.

All subjects were born and raised in the islands, but the family backgrounds of the speakers differed: the shet/shet group have parents who were themselves born and brought up in Shetland; the scot/shet group have parents from elsewhere in Scotland; the eng/shet group have parents from England. The latter two groups represent 2nd generation incomers, whose phonetic and phonological systems are based on exposure to incompatible and competing initial stop-voicning systems, that of their non-Shetlandic parents, (short-lag vs. long lag VOT) and that of the Shetlandic community (prevoicing vs. short-lag VOT).

We measured the VOT cue to post-pausal, word-initial stop voicing, based solely on wordlist data (1229 tokens). A number of distinct systems were observed. They display patterns of prevoicing, short lag aspiration and long lag aspiration.

A standard Shetland system of fully voiced vs. unaspirated stops is observed for three shet/shet subjects and one shet/eng subject.

The fourth shet/shet subject is on the boundary between aspirated and unaspirated /p t k/ and on average she has voiceless unaspirated /d g/.

One shet/eng subject has prevoiced /b d/ but voiceless unaspirated /g/. His /p t k/ are on the boundary between aspirated and unaspirated, probably falling into the latter category.

One shet/eng subject and three shet/scot subjects have standard English VOT.

One shet/scot subject and one shet/eng subject have strongly aspirated stops for /p t k/ and prevoiced stops for /b d g/, a system that is highly marked, but communicatively unambiguous in this community.

We will discuss these systems in more detail. Our results raise interesting questions for theories of the phonetics/phonology interface (and the issues mentioned above) as well as exemplifying the susceptibility of markedness in the face of functional factors.

References:

Docherty, G., P. Foulkes, J. Milroy, L. Milroy, & D. Walshaw (1997) Descriptive adequacy in phonology: a variationist perspective. Journal of Linguistics 33: 275-310

Scobbie, James M., Claire Timmins, Jane Stuart-Smith, Fiona Tweedie, Nigel Hewlett and Alice E. Turk (2000) Fieldwork in the urban jungle: an empirical phonological study of Glasgow English. 7th International Conference on Laboratory Phonology (Labphon VII), Nijmegen. June 29th - 1st July 2000.

[back to PWorkshop Archives]

<owner-pworkshop@ling.ed.ac.uk>