Abstract

The study of the processes involved in speech segmentation has gained special relevance in recent years by trying to establish what type of information listeners use to segment the speech signal into words. An event-related brain potential experiment was conducted in order to understand how two of these cues (statistical and stress cues) interact. The experiment consisted of the presentation of artificial speech streams in which words were marked either by statistical cues alone, or by a combination of statistical and stress cues. As a baseline, comparison streams were also created with the same syllables but organized in random order. Results showed an N400 component that marks the on-line segmentation of speech into words, and an increased positivity (P2 component) for languages that include both types of cues. Possible implications of these results for the process of speech segmentation are discussed.