Memory or attention?, understanding working memory in children

Abstract

This dissertation explores the construct of working memory (WM) in children, defined as the ability to concurrently remember and process information over brief periods of time. The research presented here had several goals with respect to children's working memory: 1) to develop tests of working memory that have adequate psychometric properties; 2) to ascertain whether working memory is distinct from short-term memory; and 3) to investigate the relative contributions of processing speed (PS), controlled attention (CA), and short-term memory (STM) in accounting for individual differences in working memory capacity.

To address these questions, tests thought to measure WM, STM, CA, and PS were administered to 119 normally functioning children between the ages of nine and thirteen. Two working memory tasks were modeled after the work of Daneman and Carpenter (1980), Engle, Carullo and Collins (1991), and Salthouse, Mitchell, Skovronek and Babcock (1989), that involved concurrent storage and semantic/computational processing of orally presented sentences/arithmetic calculations. The new WM measures were shown to have adequate internal consistency but inadequate test-retest reliability. CA was operationalized using the Stroop Colour and Word Test, the Trail-making Test, and commission errors on the Continuous Performance Test. STM was measured using the California Verbal Learning Test and the Semantic Categorization subtest from the Swanson Cognitive Processing Test. PS was assessed using the Visual Matching subtest from the Woodcock-Johnson tests of Cognitive ability, and the Symbol Search subtest from the Wechsler Intelligence Scale for Children, Third Edition.

Structural equation modeling techniques were used to investigate the relations between working memory and other cognitive abilities. The results indicated that WM is distinct from, though strongly correlated with, STM. Path models indicated that this correlation is largely a function of individual differences in controlled attention, which accounts for about half of the variance in the latent WM factor. Tests of PS and CA were found to best fit a one-factor solution. Because PS and CA were very highly correlated (.96, and therefore, indistinguishable) in the present sample, it was not possible to test predictions about how they would interact with each other in the prediction of WM. The implications of this result with respect to understanding individual differences in WM capacity are discussed. Overall, the results of the present study are consistent with Engle, Tuholski, Laughlin and Conway's (1999) model in suggesting that CA is a significant predictor of WM capacity. Indeed, when one accounts for CA, STM appears to add little to the prediction of WM capacity.