Buckeye Corpus

We recorded 40 people, each being interviewed for about one hour. The group of speakers is stratified on age and gender - 20 old, 20 young, 20 male, 20 female. They were all Caucasian, long-time local residents of Columbus, Ohio. The style of speech is unmonitored casual speech. The speakers responded to a newspaper advertisement seeking people to participate in a focus group on current local issues. Only after the recordings were made were they told the true purpose of the recordings. The acoustic signal is clear of noise, digitally recorded in a quiet room with a close-talking head-mounted microphone. The words spoken by these people are entered in word label files (as used by XWaves or WaveSurfer) and have time stamps that indicate the exact acoustic chunk that corresponds to each word.

Phonetic transcription of the citation forms and actual pronunciations are stored with the word labels. The phones - the actual sounds uttered by the speakers in producing the words - are also stored in label files with their time stamps.

Further details
Collection of speech was completed in Spring 2000. Forty speakers were recruited from the Columbus, Ohio community, all natives of Central Ohio (i.e., born in or near Columbus, or moved there no later than age 10). The sample design is stratified for age (under thirty and over forty) and sex. Class was not strictly controlled in order to attract participants; most speakers are middle class to upper working class.

Speakers were recruited using three methods: (1) advertisements in local free newspapers in four neighborhoods of Columbus and the Ohio State University newspaper; (2) referrals from other speakers; (3) recruitment of friends and neighbors. speakers were screened during a short phone call to make sure they were members of the target population. Potential speakers were told that the research team was interested in people's opinions. Qualified speakers came to the Ohio State University campus to have a conversation about everyday topics such as politics, sports, traffic, schools. Use of this procedure was approved by the Internal Review Board, and no speaker expressed concern after being debriefed on the true purpose of the study.

After a significant amount of piloting different protocols for eliciting large amounts of unmonitored speech, a modified sociolinguistic interview format was chosen. Interviews were conducted in a small seminar room by the (male) postdoc and (female) graduate assistant (see speech samples). To control for the possible influences of the interviewer's sex, cells are balanced so that each interviewer conversed with half of the speakers in each cell.

From the 40 speakers, about 300,000 words of speech were collected, from which the corpus of aligned speech was created. This large sample should ensure that the estimates of the forms and frequency of phonological variation are representative of the population under study. Furthermore, there should be a large number of tokens of many variant forms appearing in different phonetic environments, allowing for the control of phonetic environment in studying variation.