Second International Conference on Spoken Language Processing (ICSLP'92)

Banff, Alberta, Canada
October 13-16, 1992

Best Exemplars of English Velar Stops: A First Report

Katharine Davis, Patricia K. Kuhl

Department of Speech and Hearing Sciences, University of Washington, Seattle, WA, USA

Studies on vowels by Kuhl and her colleagues have shown that certain stimuli in a vowel category are rated as "better exemplars" (prototypes) than others. These stimuli also exhibit a "perceptual magnet effect" in that acoustically similar stimuli are perceptually drawn towards the prototype [1,2]. Vowel prototypes are language specific, both in adults and in 6-month-old infants [3]. A new series of studies has extended this work to consonants. The initial goals were twofold: 1) identify the preferred exemplars of naturally produced English voiced and voiceless velar stops in initial position, and 2) examine the contribution of various acoustic cues contained in these stops (pre-release lead voicing, silence between lead voicing and the release burst, aspiration, burst duration, and burst locus. Ten adult speakers, five male and five female, recorded twenty different words with initial velar stops before the vowel /ae/. The 200 tokens (twenty words, ten /g/-initial and 10 /k/-initial, by ten speakers) were digitized and edited to include only two pitch periods in addition to the initial consonant. The resulting stimuli were presented to eight subjects in two judgment tasks: 1) a two-choice (/k/ or /g/) identification task which also measured reaction time, and 2) a category goodness rating task wherein subjects were asked to rate each stimulus as to how "good" an exemplar of its category (/k/or/g/) it was. Acoustic parameters of the stimuli were correlated with subjects' identifications and goodness ratings to determine which variables had the most effect on subjects' goodness judgments. Results showed that the subjects' judgments were strongly influenced by the acoustic variables. The variables contributing the most were, 1) the individual speaker who produced the stimulus, 2) the durations of both lead and lag time, and 3) the presence vs. the absence of lead time in the /g/'s.

Full Paper

Bibliographic reference.  Davis, Katharine / Kuhl, Patricia K. (1992): "Best exemplars of English velar stops: a first report", In ICSLP-1992, 495-498.