Karaman, F. 1 , Lany, J. . 2 & Hay, J. 3
1 Usak University, Turkey
2 University of Notre Dame
3 University of Tennessee, Knoxville
Infants can use transitional probabilities (TPs) to extract words from speech and subsequently link them to meaning. However, the role TPs play in memory and future word learning remains unclear. Using a combined word-segmentation + word-learning task, in Experiment 1 we examined 22- to 24-month-old English-learning infants? (N=64) ability to map recently segmented words onto novel objects either immediately after familiarization with an Italian corpus (control condition) or following a 10-min delay. While control infants successfully mapped both high and low TP (HTP; LTP) words to novel objects, infants in the delay condition only mapped the HTP words. These results suggest that HTP words are not only remembered better than LTP words, but that they continue to serve as candidate object labels after a delay. In Experiment 2 (N=32), we probed differences in learning HTP vs LTP labels by violating the TPs of labels between familiarization and training/test. Results suggest that strong co-occurrence statistics likely drive learning and memory for HTP words, whereas syllable frequency may support learning of LTP labels. Although infants may take advantage of multiple types of regularities during word learning, when it comes to retention, higher-order structure like TP may be privileged in memory.