ASA 124th Meeting New Orleans 1992 October

5aSP9. Internal phonetic category structure depends on multiple acoustic properties: Evidence for within-category trading relations.

Philip Hodgson

Joanne L. Miller

Dept. of Psychol., Northeastern Univ., Boston, MA 02115

Two acoustic properties underlying the distinction between the presence and absence of a stop consonant, as in say vs stay, are silence duration and first-formant (F1) onset frequency: little or no silence and a high F1 onset frequency specify say, whereas a longer silence and a lower F1 onset frequency specify stay. Previous research has established that these two properties enter into a trading relation at the phonetic category boundary. Whether they also enter into a trading relation within the category is explored in the current investigation. Two series of stimuli having low (230 Hz) and high (430 Hz) F1 onset frequencies, respectively, have been created. Silence duration in each series varied from 0 to 996 ms, resulting in stimuli ranging from say through stay to an exaggerated version of stay. A preliminary two-choice study (with just the stimuli between 0 and 136 ms) confirmed the expected trading relation at the say--stay boundary, with the boundary for the high F1 series being located at a longer silence duration than that for the low F1 series. In the main experiment, listeners were presented randomizations of all the stimuli and asked to judge each for its goodness as an exemplar of the word stay, using a rating scale (1--10). It was expected that for each series, only a limited range of stimuli would be given high ratings. At issue was the location of these stimuli. If silence duration and F1 onset frequency trade within the category as well as at the boundary, then the best exemplars for the high F1 series should be displaced towards longer silence durations relative to those for the low F1 series. The results confirmed the predictions. These findings indicate that the internal structure of phonetic categories is determined by multiple acoustic properties. It is suggested that this complex acoustic-phonetic mapping derives from the common articulatory origin of these properties. [Work supported by NIH.]