Lany, J. 1, 2 , Wang, T. . 2, 3 & Hay, J. 4
1 University of Liverpool
2 University of Notre Dame
3 St. Mary's College
4 University of Tennessee, Knoxville
Infants preferentially learn high-transitional-probability (HTP) sequences as object labels, relative to ones with low TPs. This could mean that HTP sequences are represented as candidate words, or instead that HTP sequences are simply easier to encode. We tested whether TPs can confer word-like status by familiarizing 20-month-old English-learning infants with a corpus of Italian sentences containing HTP and LTP words, and then using those words in a label-learning task. Infants become resistant to learning words that are not typical of their native language as labels as they gain native-language proficiency. Thus, we hypothesized that if tracking TPs results in identifying candidate words, HTP Italian sequences will lose their advantage as infants learn more English words. However, if HTP sequences are just easier to encode, infants should always show a HTP advantage. Indeed, in Experiment 1 (N=38), only infants with relatively small English vocabularies learned the HTP Italian words. In Experiment 2 (N=30) we replicated this effect, and also found that infants generalize HTP but not LTP words to referents of the same kind, a hallmark of lexical but not associative knowledge. These findings suggest that HTP sequences are represented as candidate words, and that TPs are relevant to word learning.