A model comparison approach reveals individual variation in the scope and control of attention

Abstract

The maintenance capacity of working memory is known to be severely limited in scope. However, the reason this capacity varies across individuals remains unknown because it has proven difficult to estimate the maximum capacity of an individual’s “scope of attention” (SoA) separate from their ability to achieve this maximum capacity due to temporary lapses in “attention control” (AC). The present study accomplished this separation by using a maximum likelihood framework to extract latent constructs representing SoA and AC from a whole-report version of the visual-array task. The results of two experiments (N = 145 and N = 189) showed that model fit was significantly greater when the model allowed both AC and SoA to vary across individuals relative to a model in which only AC was allowed to vary (and SoA was fixed). More importantly, the individual estimates of SoA and AC derived from this variable model suggested that (1) the observed range of SoA was found to be small across individuals, with 91% able to maintain a maximum of 3 or 4 items; (2) the consistency with which AC could be deployed was only weakly correlated with the magnitude of SoA; and (3) AC and SoA were both found to be significant predictors of fluid intelligence. Altogether, the present study clarified the nature of maintenance capacity and suggested that SoA and AC both need to be included in a mechanistic account of complex cognition.