More cues or more languages? word segmentation using statistical learning in multilinguals, bilinguals, and monolinguals

Abstract

ABSTRACT This study aims to extend statistical learning (SL) research to multilinguals and provide an insight into what could facilitate word segmentation. We studied how the number of cues available in the input as well as the number of languages spoken influence SL and word segmentation. We used two SL tasks: one involving the tracking of transitional probabilities (TPs) between syllables of words, and another involving the tracking of two congruent cues-syllables and tones – in an artificial tone language. Data was collected from monolinguals, bilinguals, trilinguals, and quadrilinguals. Our results indicate that all language groups demonstrated similar SL capacity when segmenting words using TPs of syllables. However, when an additional cue was added, bilinguals, trilinguals, and quadrilinguals outperformed monolinguals. Interestingly, quadrilinguals also outperformed bilinguals. Performance was best for all groups when the input afforded two cues. This study suggests that while experience with multiple languages does not affect core SL ability, it enhances the tracking of multiple cues. The study further indicates that SL is affected by the number of cues available in the input as we found that performance was facilitated by the presence of two congruent cues.