Tummeltshammer, K. 1 , Wu, R. 2 & Kirkham, N. 1
1 Birkbeck, University of London
2 University of Rochester
Studies demonstrate that infant visual attention and learning are influenced by the statistical structure of visual features. Such evidence has lead to the hypothesis that infants attend preferentially to statistically reliable features and away from random or noisy features. We tested this hypothesis by presenting 9-month-olds with two visual streams side-by-side: patterned objects (with two perfectly correlated parts and a third uncorrelated part), and random objects (with three uncorrelated parts). Following familiarization, infants saw the patterned objects split apart in a manner either consistent or inconsistent with their correlated parts. Results show evidence of learning in 15/24 infants, who preferred the inconsistent over the consistent splitting event. A median split divided infants into strong and weak learners. No differences emerged in total looking time to patterned and random streams during familiarization; however, a Group x Block interaction indicates that strong and weak learners distributed attention differently over time. Strong learners decreased looking to the patterned stream (and increased looking to the random stream), while weak learners looked indiscriminately between the two streams. The same analysis of Wu, et al.'s (2011) Exp4 shows that cueing infants' attention to the patterned stream produces similar looking trends between strong and weak learners.