A finding reliably demonstrated in past research is that statistical learning mechanism facilitates the process of learning language. What remain poorly understood are the effects of multiple speakers in infants and adults learning a statistical artificial language. This study sought to examine the effects of two different speakers in adults because previous literature has suggested that infants lack the ability to segment words when the speech stream consists of two different speakers. Therefore, our experiment sought to understand if 1) adults could successfully segment words across two different speakers and 2) if they can generalize segmentation to a novel voice. Contrary to the infant study, it was found that adults could successfully segment and identify words even when exposed to different speakers. However, adults had difficulty in generalizing to a novel voice when exposed to a single talker. These results support the role of the exemplar theory and raise the possibility that adults are not that experienced language processors as previously expected.
Kang, Hosung, "Adult statistical word segmentation across two speakers" (2017). 2017 Undergraduate Awards. 13.