Exploration of the mental architecture of working memory: An internet-based study

Abstract

The structure of working memory has been studied but so far the results have been inconsistent. The main aim of the current study was to examine the mental structure of working memory using exploratory factor analysis. In addition, it examined the reliability of online cognitive testing. Ten working memory tasks representing four different task paradigms, namely simple span, complex span, running memory and n-back, were used. The selected tasks allowed content-based (verbal/visuospatial) and/or function-based (updating, maintenance, processing) latent variables to emerge. The tasks were administered online to a large (n=711) and demographically varied U.S. based sample via Amazon Mechanical Turk crowd work website. The results of the exploratory factor analysis can be interpreted either as a content-based verbal/visuospatial two-factor solution or as a three-factor solution with verbal/visuospatial factors and an n- back factor. These results indicate that the content-based division is a prominent feature of working memory. The n-back factor could reflect a function-based division, possibly related to use of speeded recognition rather than free recall in n-back tasks. Furthermore, the two-factor model suggested that the n-back tasks are processed spatially irrespective of their content. As regards online cognitive testing, the present results indicate that measurement of working memory is feasible online. The reliability of online-based working memory tasks was assessed by comparing means, correlations and standard deviations of working memory scores obtained by lab and online testing. Mean performances in the two testing conditions were similar and the correlations between the WM tasks concurred with earlier findings in the literature.