Abstract

The field of psychology frequently trades on the ambiguous relationship between theoretical models of individuals and theoretical models of groups. Making the inference from one to the other implies being able to put variance in data in its just place. In research on cognitive development, where we are concerned with dynamic processes rather than static ones, it can be even more difficult to decipher the noise from the true picture. Due to reliance on traditional statistical methods to extract significant results, most discussions of learning and development are based on aggregated group data, even though developmental processes reflect changes in an individual brain. In this talk, I will illustrate some of the problems with group averages for really understanding a developmental process, focusing on the example of how young children develop social awareness. Then I will give an overview of a new method I have adopted from the animal learning literature to overcome these challenges: Bayesian change point analysis. Developmental change in a child’s understanding is evident in a cumulative record when the slope of the curve changes. Using Bayesian methods of analysis allows us to determine both the presence and the absence of change points in individual records, and to look at the types of profiles represented across a group. Results show striking variability in developmental-change patterns. Bayesian change point algorithms open up new ways of quantifying variability in development, both within and between individuals.