Dichotic forward and backward "masking" between CV syllables.

Abstract

Dichotic CV syllables were presented at stimulus onset asynchronies (SOAs) between 0 and 480 msec in two ear−monitoring tasks (’’forward’’ and ’’backward masking’’). The design included indentical pairs and orthogonally varied stimulus intensities. A ’’similarity effect’’ (higher recognition scores when the target syllable contrasted on fewer features with the ’’masking’’ syllable) was found over the whole masking range (ca. 300 msec). Other findings included strong effects of both target and mask intensities, a ’’lag effect’’ but no right−ear advantage, and generally opposite effects on identical and nonidentical pairs. Identical pairs showed no masking at all. A model is suggested to account for the pattern of errors in dichotic listening: confusions are assumed to arise from nonspecific capacity limitations (’’perceptual noise’’), separately for each stimulus (hemisphere), and, at short SOAs, responses are assumed to be frequently assigned to the wrong channel. In addition, fusion may occur at very short SOAs. It is further argued that little direct interaction between dichotic inputs may take place; that central convergence occurs at a late stage in processing; that the assumption of a feature processing stage as distinct from a phoneme recognition stage may be unnecessary; and that the backward ’’masking’’ range cannot be equated with absolute processing time. Subject Classification: 70.30.