Do newly formed word representations encode non-criterial information?

Lexical stress is useful for a number of language learning tasks. In particular, it helps infants segment the speech stream and identify phonetic contrasts. Recent work has demonstrated that infants aged 1;0 can learn two novel words differing only in their stress pattern. In the current study, we ask whether infants aged 1;0 store stress information in their representations of words even when it not required for the task. To this end, we taught infants novel, three-syllable word-object pairings. At test, we manipulated the word by presenting infants with forms that shared the stress pattern of the familiar words but differed in the segments, and forms that shared the segments of the familiar word but differed in the stress pattern. Our findings reveal that infants' representations of new words include word-level stress information and do not simply contain the information critical for distinguishing between different forms.