Consonant-Vowel Coarticulation in the Buckeye Corpus

Theories of vowel perception, production, and acquisition have moved on from the early simple assumption of a single target for each vowel in auditory space, and yet modern variations continue to employ this notion in more sophisticated garb. The vowel prototype is lately conceived as the result of an auditory processing function incorporating information about both vowel dynamics and flanking sounds. Vowel sounds most often are flanked by consonants which are coarticulated with the vowel. It is commonly presumed that listeners apply an unknown auditory transformation to undo the coarticulatory effects when processing speech cognitively. This presumption requires coarticulation to be systematically predictable from neighboring consonants, which some studies have not found to be the case. The present study examines the vowel formants of 37 speakers from the Buckeye corpus of American English. It is found that consonant-vowel coarticulation effects are large and ubiquitous, but are not sufficiently consistent across speakers to be predictable from the consonants. As a result there are no “ideal” vowels to be found in the data. There appears to be no way to transform such acoustic data by relying on the consonants that could yield a set of ideal vowel prototypes.Theories of vowel perception, production, and acquisition have moved on from the early simple assumption of a single target for each vowel in auditory space, and yet modern variations continue to employ this notion in more sophisticated garb. The vowel prototype is lately conceived as the result of an auditory processing function incorporating information about both vowel dynamics and flanking sounds. Vowel sounds most often are flanked by consonants which are coarticulated with the vowel. It is commonly presumed that listeners apply an unknown auditory transformation to undo the coarticulatory effects when processing speech cognitively. This presumption requires coarticulation to be systematically predictable from neighboring consonants, which some studies have not found to be the case. The present study examines the vowel formants of 37 speakers from the Buckeye corpus of American English. It is found that consonant-vowel coarticulation effects are large and ubiquitous, but are not sufficiently consiste...