CUES TO WORD SEGMENTATION FOR ADULT LEARNERS

Abstract

An important step in acquiring a language is the ability to segment words from speech streams. Typical speech contains many cues to word segmentation, but cues are not always consistent. In studying the efficacy of particular cues, it has been suggested that some non-linguistic information, such as music, may actually help with word segmentation. Although it is traditionally accepted that music and language are treated as separate types of information by the brain, recent evidence suggests that there may be shared structural, though likely not semantic, properties. The current study was designed to compare the effects of cues to word segmentation on learning rates in order to determine if tonal information could provide a benefit beyond that provided by regular speech cues. Participants listened to a speech stream of pseudo-randomly repeated nonsense words. Speech streams were of four types: monotone, prosody-enhanced (final vowel lengthened), tonally-enhanced (each syllable "sung" on a particular tone), and tonal-word (every "word" "sung" in the same series of three tones). On a forced-choice test participants were asked to choose which in a pair of syllable strings most resembled a word from the exposure stream. Learning was measured by the number of correct responses on the forced-choice test. Results showed a significant facilitory effect of the prosodic cue (i.e., final vowel lengthening), but no effect of either tonal condition, suggesting a privileged status for language-specific cues to word segmentation. Failure to replicate previous findings of tonal facilitation are discussed in relation to the detrimental effects of two unexpectedly high between-word transitional probabilities as well as a potential lack of statistical power.