Perceptual cue learning

Abstract

This paper describes results from an on-going perceptual training study in which native listeners of English are trained perceive a categorical difference between Polish alveopalatal and retroflex sibilants. The stimuli consist of naturally produced, modified two-dimensional stimuli varying by fricative noise and formant transition cues. Subjects are trained to categorize using only the fricative dimension, formant transition dimension, or both dimensions. Testing consists of a two-alternative forced-choice discrimination paradigm and stimulus labeling. Initial results from pilot work and subjects in the fricative dimension training group demonstrate that some listeners can easily attend to either dimension independently for categorization where training focuses attention and sharpens boundaries while others rely heavily on the vocalic cues, only attending reliably to frication noise cues with several training sessions or through verbal instruction. A handful of subjects did not show any improvement due to training; most of these subjects also showed no initial categorization strategy. Moreover, sensitivity to the fricative dimension and boundary increases for all learners (measured in d′), with no corresponding increase or decrease in sensitivity to vocalic cues. Further results from other training conditions will also be reported. [Work supported by NIH-NIDCD.]