Computational and empirical studies of task switching

This thesis is concerned with the study of executive function using the task switching paradigm. It is argued that theoretical accounts of task switching fall into two categories: a) those that attribute the reaction time cost of task switching to interference caused by previous performance of a different task ('task carryover' accounts) and b) those that attribute the 'switch cost' to the duration of one or more executive control processes that are specific to task switching ('extra process' accounts). These accounts have different implications for the study of executive function with the task switching paradigm: according to extra process accounts, control processes involved in task switching may be investigated directly, whereas according to task carryover accounts there can only be an indirect relationship between the operation of control processes and the duration of switch costs. A parallel distributed processing (PDP) model is presented, implementing a version of the task carryover account. This model simulates reaction times when subjects switch between word reading and colour naming in response to Stroop stimuli, providing a good fit to a large body of empirical data. Further simulations with random parameter settings show that the model simulates this data as a result of its architecture and general processing principles, rather than the specific parameter settings that were chosen. Six empirical studies are then reported, investigating task switching performance when subjects simultaneously perform a working memory task. The model is used to relate patterns of switch costs in these experiments to the underlying control processes and strategies that may be employed by subjects. It is concluded that task switching may provide a suitable methodology for the study of executive function, but that the relationship between switch costs and executive function is complex and indirect