Errors and latencies for same-different judgments of dichotic CV syllables at various stimulus onset asynchronies (SOAs)

Abstract

Three studies were conducted using this paradigm. The major independent variables were SOA and the number of feature contrasts (FC) between the syllables to be compared. Errors and latencies were found to decrease systematically with SOA and with FC. The effect of FC on latencies persisted over at least 500-msec SOA, which argues against comparisons of “holistic” phonemes after encoding is completed. The third study included variations in pitch and predictability of this variation (blocked versus random) as additional factors. They were found to have dramatic effects at very short SOAs but surprisingly little influence at longer SOAs. The results are discussed with reference to a two-process model of dichotic interaction: “integration” at very short SOAs versus “perceptual noise” at longer SOAs. Evidence for integration was seen in the frequency occurrence of incorrect “same” responses at short SOAs. Dichotic syllables that are not integrated are assumed to be compared after being categorized, the major evidence being a negative relationship between intelligibility and incorrect “different” responses.