ISCA Workshop on Plasticity in Speech Perception (PSP2005)

Senate House, London, UK
June 15-17, 2005

Distributional Information and Infants' Use of Phonemic Contrasts for Word Learning

Erik Thiessen

Carnegie Mellon University, Pittsburgh, PA, USA

When engaged in word-learning tasks, infants between 14 and 17 months appear to be lesss sensitive to phonetic distinctions than they are in discrimination tasks (Stager & Werker, 1997; Werker, Fennell, Corcoran, & Stager, 2002). For example, while children can hear the difference between 'bih' and 'dih,' they treat 'bih' and 'dih' as interchangeable labels in word learning tasks.

By 20 months, the disconnect between phonetic discrimination and word-learning has largely disappeared (Werker et al., 2002). One explanation proposed for this is that capacity increases between 14 and 20 months, allowing older children to attend to subtler phonetic distinctions in word-learning tasks. Another explanation is that characteristics of childrens' lexicons determine what distinctions they will use in word-learning tasks (e.g., Schafer & Mareschal, 2001; Shvachkin, 1954; Walley, 1993). To test these competing claims, we replicated Stager and Werker's (1997) results with new stimuli. Children familiarized with an object called a 'daw' accepted 'taw' as a label for that object.

In two additional conditions, children not only learned a label for the daw-object, but the names of two additional objects. In one of the conditions, the other two objects were labeled as 'tawgoo' and a 'dawgoo.' In this condition (the Uninformative condition), 'daw' and 'taw' occurred in similar contexts in the two additional words. In the other condition (the Informative condition), the two additional objects were labeled a 'tawgoo' and a 'dawbow.' In this condition, the two additional objects had dissimilar names, and provided examples of 'daw' and 'taw' in distinct contexts.

In the Uninformative condition, infants continued to treat 'daw' and 'taw' interchangeably, and accepted either as a label for the daw-object. In the Informative condition, infants dishabituated when the daw-object was referred to as 'taw.' Experience with 'daw' and 'taw' in distinct lexical contexts enabled infants to use the distinction in a wordlearning task. This suggests that the distribution of phonemes in the words with which infants are familiar plays a role in their subsequent use of those phonemes.

These results indicate that the change in children's use of phonemic contrasts in word-learning tasks between 14 and 20 months may be explained by characteristics of their lexicon at those two ages. At 20 months, children know more words than 14- month-olds. Their larger vocabulary may provide them with more evidence about the distribution of phonemes in the words of their language.

Bibliographic reference.  Thiessen, Erik (2005): "Distributional information and infants' use of phonemic contrasts for word learning", In PSP2005, 243 (abstract).