Files
Abstract
In US English, the merging of the voiceless labiovelar glide [hw] and its voiced counterpart [w] has been an ongoing process over the past century, originating in central port cities on the Atlantic seaboard and gradually spreading to include the bulk of the continental US, with the exception of the Southeast. At this point, the so-called wine-whine merger shows evidence of nearing its completion, with [hw] absent in the speech of most younger speakers and declining over time such that middle-aged individuals produce it at lower rates than older individuals. Notably, however, although [hw] is interpreted by outsiders as a feature of Southern speech, there is evidence that those who use it do not believe it to index Southern identity: [hw] was used more frequently by Southern speakers in Bridwell (2019) with low attachment to their local area, and in more formal tasks such as reading.
While previous studies have sought to analyze the wine-whine merger through current interview data, at the present time [hw] pronunciations are rare enough that it is difficult to collect sufficient data to identify its social correlates. It is therefore of interest to analyze speech from a time when [hw]-usage was more widespread in order to identify which groups were among the first to adopt the merger. For this purpose, the current study analyzes data from the Digital Archive of Southern Speech (DASS), an audio corpus of interviews collected from 1968-1983 as part of the Linguistic Atlas of the Gulf States, and transcribed and forced-aligned by researchers at the University of Georgia. DASS contains 372 hours of recorded speech from 64 individuals (born 1886-1965) across eight Southern US states and is encoded for numerous social attributes. These characteristics make it possible to measure the state of the wine-whine merger at a transitional point in the American South, and to compare its progress among speakers from a range of backgrounds.
To create the dataset for the current study, 7702 words containing morpheme-initial were extracted from DASS and coded by the authors as instances of [hw] or [w] on the basis of the audio and spectrogram in Praat. From a variety of social and linguistic attributes, a model of best fit identified age, land region, semantic content, vowel, ethnicity, and education as contributing variables to [hw]-production. Results not only agreed with findings from previous studies related to geographic region, age, and semantic content, but also introduced two new variables as significant predictors of [hw]-use: ethnicity and education. These showed that Black speakers, particularly those born after 1930, used [hw] less than non-Black speakers, and that college-educated speakers used [hw] more than their peers with a high school education or less. This last finding is particularly significant in that it suggests that [hw] was not viewed as a regional variant in the 1970s, but rather associated with standardness and prestige, an attitude which may explain unexpected patterns of [hw]-use in more recently collected data.