Kapatsinski, V. , Olejarczuk, P. & Redford, M.
University of Oregon
Intonation contours are interesting for investigating perceptual category learning because of their temporal extent and variability. We present a series of experiments on perceptual learning of intonation contours, in which random deflections from a novel contour prototype were presented to English-speaking adults and children between the ages of 9 and 11 (a la Posner & Keele 1968). Participants associated contours in a category with a creature that 'talked like that'. At test, they judged whether contour exemplars (both familiar and novel) were likely to be said by one of the familiar creatures or some novel creature they haven't seen before. All participants accepted previously trained exemplars, prototypes, and new exemplars provided they were as close to the prototype as the old exemplars. This result suggests a region around the prototype in which all exemplars are equally good, inconsistently with both exemplar and prototype models (Smith & Minda 2000) but in line with a ?window? model of intonation contour targets (cf. Keating 1990). In addition, as found by Gibson & Gibson (1955) for scribbles, children were likely to accept novel exemplars that were somewhat farther away from the prototype than the training exemplars, forming wider ?windows?, or regions of perceptual equivalence.