Control, automaticity, and working memory: A dual-process analysis

Cognitive control has become a central concept in numerous areas of psychology; however little effort has been devoted to integrating the various approaches to defining and measuring this key concept. The main goal of the current research was to compare two approaches to the study of cognitive control, working memory (WM) and the process dissociation (PD) procedure. Traditional span measures were used to formulate two groups with different levels of WM capacity who then performed two PD memory tasks (verbal and spatial Proactive Interference) and two attention tasks (Color-Word and Spatial Stroop). Estimates of control and automaticity were computed using standard PD equations and multinomial models of performance, which were used to examine the configuration of processes underlying performance across different task domains. For the memory tasks, estimates derived from both the standard PD equations and the multinomial models suggested that control operated independently of more automatic processes in all tasks and that differences between the span groups were restricted to the more controlled processes. For the attention tasks, low spans were significantly slower than high spans in all conditions suggestive of general slowing. When adjustments were made to equate the span groups on response latency and estimates recomputed, the remaining span differences were completely isolated to the automatic estimate. This suggests that WM-related control processes are being indexed by the automaticity parameter in the PD Stroop model. This claim was supported by correlations among the estimates, the WM measures, and a measure of fluid intelligence (Ravens). These analyses showed that both the control estimates from the memory tasks and the automaticity estimates from the Stroop tasks shared strong correlations with each other and with the criterion measures. This is in contrast to the automaticity estimates from the memory measures which showed no reliable relationships with the criterion measures. These findings help to explain the greater interference shown by low spans in a variety of memory and attention tasks. More generally, the results suggest that integrating the WM and PD approaches can lead to insights into the nature of cognitive control.