Slone, L. & Johnson, S.
University of California, Los Angeles
Much research has documented learners' ability to segment auditory and visual input into its component units. Two types of strategies have been proposed to account for this phenomenon: statistical strategies, in which learners track statistical relations (e.g., transitional probability, co-occurrence frequency) across elements, and chunking strategies, in which learners group elements together into units. In a series of five experiments, we investigated adults' (Experiments 1-3) and infants' (Experiments 4-5) performance on a visual sequence-learning task. Experiments assessed learning of two types of items for which statistical and chunking models provide contrasting predictions: embedded items - features that occur within larger features - and illusory items - sequences that are never presented to participants, but have the same statistical structure as other items that are presented. The pattern of results obtained was most consistent with the competitive chunking approach of Servan-Schreiber and Anderson (1990): both infants and adults primarily engaged in chunking, however adults were able to maintain representations of both embedded items and the units within which these items were embedded, depending on the difficulty of the task. Implications for theories and models of statistical learning will be discussed.