The primary data for these studies are conversational speech taken from tape-recordings of sociolinguistic interviews of working-class native speakers of the four dialects studied.
The dialects were chosen to be very different from one another. They include exemplars of each of the ``Three Dialects of English''(Labov 1991) plus a fourth dialect which is quite different from any of these and lies well outside this classification. The Three Dialects, so-called, are the Northern Cities, Southern, and Low-Back Merged dialects. The particular communities of the speakers studied here are Chicago, Anniston (Alabama), and Los Angeles (Chicano), while the maximally different dialect is Jamaican Creole.
A second reason for the choice of these particular dialects is the availability of the data. Four of my colleagues kindly gave me access to tape-recordings of some of the best interviews which they had made in the course of fieldwork in these communities. I wish to express my thanks to them here. The interviews analyzed are parts of larger projects on various linguistic aspects of the various dialects. The Los Angeles Chicano data is part of Otto Santa Ana's (1991) dissertation on various phonological issues in that dialect, including cluster simplification, as well as vowel reduction. The Chicago interviews by Sharon Ash are part of the project on Cross-Dialectal Comprehension, another interview by Benjamin Wald is part of the data on which Labov, Yaeger and Steiner (1972) was based (though acoustic analysis of this speaker, Jim C., was not done there). The Anniston, Alabama, material comes from interviews done by Crawford Feagin in 1971 as part of the fieldwork behind her (1979) book, Variation and Change in Alabama English, and which she has drawn on in other work since then (1986, 1987, 1990, and to appear). Finally, the Jamaican Creole data is from sociolinguistic interviews done by Peter Patrick, who is completing his dissertation, a sociolinguistic study of urban Jamaican Creole in a neighborhood of Kingston.
The relation of the interviewer to the interviewee is partly controlled in this data. The interviews are all same-sex conversations (excepting Feagin's interview with James H.). The interviewers are in all cases trained sociolinguistic fieldworkers who grew up in a speech community near to the one studied. Otto Santa Ana is himself a Chicano American from Arizona, so while he is not a native member of the Los Angeles Chicano speech community, his personal background makes him much better able to get natural behavior from LACE speakers than an Anglo such as myself might be. Crawford Feagin was raised in Anniston, Alabama. Sherry Ash is herself from the Chicago area. And Peter Patrick spent his first 15 years in Jamaica and is marginally a native speaker of JC. In every case, the interviewer was a somewhat marginal member of the speech community studied, neither an insider nor an outsider.
Considering that non-working-class academics are generally outsiders to the working-class communities near where they grow up, and even more so far from home, this data is about as natural as can be gotten in this type of interview. Certainly interviewers without detailed knowledge of the community studied would have stimulated a very different set of behaviors, that are likely to be much less characteristic of natural conversation between native speakers of the particular dialect.
The interview situation is also controlled: each is a sociolinguistic interview. (For details, see Labov 1984:32-42.) The format of the sociolinguistic interview is designed to get the maximum quantity of unmonitored speech, preferably in the form of narratives of personal experience, from a single speaker, in a fairly short period of time. The sociolinguistic interview is a one-on-one conversation, with the interviewer leading a rather open-ended conversation along paths that appear to be interesting to the speaker. A good interviewer knows what topics are likely to be interesting to their subjects. The desired (and in these cases, attained) result is that the interviewee talks much more than the interviewer.
The equipment used to record the interviews consisted of either portable reel-to-reel 1/4'' tape-recorders (for Chicago and Alabama speakers), or an excellent quality cassette tape recorder (for the Jamaican Creole and Los Angeles Chicano speakers). Sound quality varies from excellent to good, though sometimes overlapping speech and intermittent background noises (particularly for Jim from Chicago, and Roasta from Jamaica) required particular words or phrases to be excluded.
The data that is derived from these tape-recorded interviews includes all the measurable vowels that occur in lengthy parts of the interview. The data thus includes both stressed and unstressed tokens and a range of consonantal contexts, which are two of the variables that will be focused on. It should be noted that studies of less-monitored speech are more likely to show increased frequency and degree of application of phonetic processes such as coarticulation and reduction. An important disadvantage of this method of data analysis is that the sample is skewed by the token frequency of the various features of interest. Even the law of large numbers, by which bias is minimized as the sample size increases, may not help, since the token frequencies may be systematically skewed by certain high frequency words. This disadvantage can be minimized by careful attention to the skewing factors. On the other hand, while the data may be skewed in some ways, it is real speech data, which relatively closely reflects the actual vernacular speech patterns which speakers of these dialects use when communicating with each other.