Statistical models of executive function test performance in children

Research into executive functions (EF) has increasingly adopted a multivariate approach in which groups of tests are administered to adults (e.g., Miyake, et al., 2000) or children (e.g. Huizinga, Dolan & van der Molen, 2006) in order to assess the structure of these processes. Within such tasks participants perform multiple trials to assess each component EF skill. Most studies simply aggregate scores to identify the underlying construct. My approach is to analyse performance using random effects models on successive trials of a task, both with and across time points. This enables me to examine the dynamics of performance and influence of extraneous processes like test order on the development of these skills. I examine a study in which 89 children were administered a battery of EF and social understanding at three time intervals, six months apart. The analyses reveal that children’s success on individual tasks is strongly influenced by their previous performance on test trials and the order in which tasks were administered. At the very least such findings raise questions about the validity of research in which large numbers of tests are administered to children within a battery. I suggest that executive task performance must be cast within a theoretical framework constructed as a dynamic relation between the participant, the test and the experimenter.