Cornell Cognitive Studies Symposium

Statistical Learning across Cognition

Visual Statistical Learning in Infancy

Scott Johnson
Cornell University


Interest in statistical learning in infancy has centered principally on auditory input, but we recently showed that infants as young as 2 months of age can extract probabilistic sequences in visual stimuli (Kirkham, Slemmer, & Johnson, 2002). Our task tested infants' learning of simple transitional probabilities between adjacent shapes in a sequence, and the question arises whether more complex patterns can be learned as well. In joint work with Natasha Kirkham and Jonathan Slemmer, we are exploring this question currently in two separate lines of research. In the first, 5- and 8-month-olds are observed for evidence of extracting a higher-order pattern across two different sets of discrete stimuli ("ABB" vs. "ABA"). Eight-month-olds appear to learn the "ABB" pattern, but not the "ABA" pattern. Five-month-olds provide no evidence of pattern learning. In the second study, 8-month-olds are observed for evidence of learning transitional probabilities of discrete shapes that each appear in a distinct location. The infants show some evidence of learning to "bind" each shape to its location. These new findings will be discussed in terms of the facility of infants to learn some patterns more readily than others, and what this might mean for developing perceptual and cognitive systems.


Back to main page