Duration discrimination of affricate silent intervals: Performance in fixed- and roving-standard tasks.

Abstract

Duration discrimination for the silence preceding the frication noise in the word /watsh/ was measured. The utterance was produced by a female talker and digitized at 20 kHz. The 200-ms affricate silence was edited to obtain: (a) seven tokens with silence ranging from 15–45 ms in steps of 5 ms (‘‘short’’ range); and (b) seven tokens with silence ranging from 100–200 ms in steps of about 17 ms (‘‘long’’ range). The ‘‘short’’ and ‘‘long’’ range were perceived as /wash/ and /watsh/, respectively. With each range, discrimination was measured in both fixed-standard and roving-standard 2IFC tasks, using a constant-stimuli method. Listeners discriminated 5%–60% decrements in the duration of the silence. Thresholds were interpolated at the 75%-correct level from probits fitted to five-point psychometric functions. For the ‘‘long’’ silence tokens, thresholds (expressed as Weber fractions) ranged from 0.08–0.18 and 0.07–0.13 for roving- and fixed-standard tasks, respectively. For ‘‘short’’ silence tokens, the corresponding Weber fractions ranged from 0.09–0.82 and 0.08–0.73. [Work supported by OCAST Grant No. HSO-005 and Presbyterian Health Foundation.]