Cornell Cognitive Studies Symposium

Statistical Learning across Cognition

Perception of Multimodal Sequential Structure in Human Infants

David J. Lewkowicz
NYS Institute for Basic Research


Humans and non-humans are endowed with general sequential learning mechanisms that enable them to encode fixed serial sequences either on the basis of inter-element associations or on the basis of their ordinality. Statistical learning is a special and more sophisticated form of sequential learning because it permits the detection of regularities in sequences that are not always fixed. Evidence indicates that sequential learning mechanisms, including statistical learning, emerge early in human development. Some of this evidence shows that infants can encode fixed serial lists but this evidence can be interpreted as showing only that infants can form associations between adjacent list members. What has been missing until now is evidence that infants can encode fixed sequences in terms of their ordinality. In my talk I will present the first evidence that infants can, indeed, perceive list ordinality as well. The second part of my talk will focus on infant perception of another instance of sequential specification: rhythm. I will discuss results from studies in my laboratory on infants' perception of bimodally specified rhythmic sequences and the special problems that its bimodal representation might present from an integration point of view. I will illustrate the perceptual power of rhythmic temporal organization for infants and will use these findings as well as those from our work on infant serial list learning to consider the theoretical significance of sequential perceptual abilities for cognitive development in general.


Back to main page