Abstraction and specificity in preschoolers' representations of novel spoken words

Four experiments explored long-term auditory priming for novel words (nonwords) in preschoolers. In Experiment 1, 2.5-year-olds more accurately identified novel words that had been presented just twice in an initial study phase than nonwords that had not been presented, showing auditory priming for nonwords. Experiments 2, 3, and 4 revealed that the sound representations underlying auditory priming in young children, as in adults, include both abstract and token-specific information about the sounds of new words. In Experiment 2, 3-year-olds showed priming for studied nonsense syllables that changed both token and recorded context from study to test, compared to entirely new test syllables. In Experiment 3, 3-year-olds more accurately identified nonsense syllables that were the same tokens in the same context at study and test than syllables that changed token and context from study to test. In Experiment 4, 3-year-olds more accurately identified the same-token syllables from Experiment 3, even when those syllables were presented in isolation, spliced out of their original contexts. Thus children's rapidly formed representations of new spoken words include both components abstract enough to identify the same sound sequence across changes in word token and changes in phonetic context and components specific to the originally presented token. We argue that the powerful perceptual learning mechanism underlying auditory word priming has the right properties to play a central role in the development of the auditory lexicon.