Psychophysical procedure and the perceptual magnet effect: Comparisons of fixed and roving AX discrimination of /i/

Abstract

Recent experiments by Iverson and Kuhl [ 553–562 (1995)] have demonstrated that the perception of /i/ is influenced by category goodness. Listeners exhibit a perceptual magnet effect characterized by high sensitivity to acoustic differences near poor exemplars of /i/ and low sensitivity near excellent exemplars of /i/. The present study examines whether this effect is influenced by psychophysical procedures. Listeners were asked to discriminate pairs of stimuli from an /i/ to /e/ continuum, and the task was varied in two ways: (1) Each block of trials had either one pair of tokens (fixed discrimination) or pairs of tokens from the entire stimulus range (roving discrimination), and (2) the acoustic difference between each pair of tokens was either 30 or 60 mels. The results demonstrated that the peak in discrimination at the /i/-/e/ boundary diminished with fixed discrimination tasks and 30-mel differences between tokens, supporting previous findings [Macmillan et al., 1262–1280 (1988)]. However, the magnitude of the perceptual magnet effect seemed less influenced by these manipulations. The results suggest that distortions of sensitivity at boundaries and within phonetic categories may arise from different mental processes.