Abstract

Using a multi-trial perceptual learning task, two experiments explored relative contributions of data-vs, concept-driven processing to the development of correct percepts for the Korean Geulja which is similar to the syllable in English and consists of consonants and a vowel. Subjects were presented with three sets of triple Geulja at exposure durations for 50-60% accuracy levels. Three Geuljas in each triple were discriminable by one or two features, i.e., they were very similar at the global feature level. Two methods for presenting stimuli, developed by Chastain (1977), were employed: The positional and the horizontal sets. The former set in which three similar Geuljas of a triple were presented at the same position over three successive cards was presumed to encourage subjects to rely on concept-driven processing. The horizontal set in which three similar Geuljas were simultaneously presented in a card .,vas expected to encourage subjects to adopt data-driven processing. Contrary to the results reported by Chastain, the present experiments consistently found the superior performance on the horizontal over the positional set, especially at the third position of a triple as well as at the rate of reduction in recognition errors. These results indicate the dominance of the data-driven processing in the perceptual conditions where no feedback was given to the subject's recognition response to one of very similar stimuli presented at the threshold.