Learning from Environmental Regularities is Grounded in Specific Objects not Abstract Categories


This paper examines statistical learning in the presence of predictive regularities at multiple levels of abstraction. Participants were presented with streams of pictures where picture order was predicted by both object identity and the categories these objects belong to. In Experiment 1, we establish that participants do learn based on the specific objects and not solely at the abstract, categorical level. In Experiment 2, we discount the possibility that participants gain abstract knowledge in addition to more concrete, object-based knowledge. Moreover, we consistently find equal learning in those who viewed the atypical exemplars and those who viewed the typical exemplars of the categories. Overall, our results suggest that when learning from environmental regularities, object-specific information takes precedence over more abstract, category level information when both are predictive.

Back to Table of Contents