Proceedings of the Annual Meeting of the Cognitive Science Society

The limited ability to simultaneously perform multiple tasksis one of the most salient features of human performance anda defining characteristic of controlled processing. Based onthe assumption that multitasking constraints arise from sharedrepresentations between individual tasks, we describe a graph-theoretic approach to analyze these constraints. Our resultsare consistent with previous numerical work (Feng, Schwem-mer, Gershman, & Cohen, 2014), showing that even modestamounts of shared representation induce dramatic constraintson the parallel processing capability of a network architecture.We further illustrate how this analysis method can be appliedto specific neural networks to efficiently characterize the fullprofile of their parallel processing capabilities. We presentsimulation results that validate theoretical predictions, and dis-cuss how these methods can be applied to empirical studiesof controlled vs. and automatic processing and multitaskingperformance in humans.