Studying socially stratified linguistic variation in the Edinburgh Phonetics Recording Archive

Phonetic research on language variation and change has recently been facilitated through the use of large-scale speech corpora. However, recordings of speakers from the same speech community, made over long periods of time, are often not readily available, and collecting such recordings can be expensive and time-consuming. Pre-existing recordings, even if not purposebuilt for a particular project, can thus be valuable resources for language variation research. Here, we describe our recent work with the Edinburgh Phonetics Recording Archive to further our understanding the features impacting phonetic variation in Scottish English. Recorded by staff and students at the University of Edinburgh starting in the late 1940s, the Archive contains many thousands of recordings, including over 700 recordings of the ‘Arthur the Rat’ passage, a phonetically-balanced text long used in collecting data on accents of English (MacMahon 2010). In addition to speech recordings, the archive contains a considerable amount of demographic information, such as place and date of birth and secondary school attended. In this paper, we describe our initial analysis of a subset of this corpus consisting of 63 speakers born in Edinburgh between 1887 and 1948 and recorded between 1949 and 1966. Due to the large volume of data in the corpus, we employed the Munich AUtomatic Segmentation tool (MAUS: Schiel et al. 2011) to provide a first-pass segmentation, which was then hand-corrected to remove major errors. We began by focusing our attention on the acoustic realization of the over 14,000 vowel tokens in our corpus. Although the vowel inventory of Edinburgh English is sometimes described as similar to that of Glasgow (Hughes, Trudgill & Watts 2012), considerable socially stratified variation has also been evidenced within Edinburgh proper (Johnston 1984). The Arthur the Rat corpus provides us with an opportunity to explore this variation in a more quantitative fashion. Our initial analyses provide acoustic evidence for social class correlates of fine phonetic variation in Edinburgh speech. For example, we observe an increase over time in the fronting of the TRAP vowel, and we further find that males, and speakers who attended fee-paying schools, also produced a more fronted TRAP vowel (Fig. 1). A less complex example is the fronting of GOOSE, which appears to be temporally stable (i.e., unrelated to speaker year of birth) but is much further front for speakers who attended state schools than fee-paying schools (Fig. 2). In ongoing work, we are exploring differences in socially-stratified variation in monophthongization, Voice Onset Time, and the realization of the Scottish Vowel Length Rule. We hope this project can inspire additional work with archival recordings that promise to expand our understanding of the dynamics of language variation and change.