Statistical Learning of Novel Words is Unaffected by the Additional Assignation of Referents

Abstract

Statistical learning is a key area of interest in psycholinguistic research; in cross-modal learning situations, it is still an open question how learning in one modality affects learning in the other. In the present study, we examined whether the addition of object referents to words in an unbroken speech stream would improve, impair, or have no effect on the learning of the words that comprised the stream. The contemporaneous assignation of object referents to words in an unbroken speech stream models real-world word learning scenarios. We assessed word learning and learning of word-object associations using segmentation and object-label recognition tasks. We found that the object referents had no effect on word learning: participants were above chance to learn the words in all conditions and they did not receive a learning boost with the addition of pictures. However, participants were also well above chance to recognize an object referent when given its word label even if they did not successfully segment that word label in the speech stream. There was no significant correlation between word learning and word-object association learning, indicating that the two abilities are independent of each other. Implications of this intriguing finding and further questions it raises are discussed herein.