PergamoD 0021-9630(95)00169^-7 J. Child Psyehol. Psyehiat. Vol. 37, No. 6, pp. 673-686, 19% €> 1996 Association for Child Psychology and Psydiiatry Published by Elsevier Sdence Ltd Printed in Great Britain. All rights reserved 0021-%30/96 $15.00 + 0.00 Working Memory in Children with Autism and with Moderate Learning Difficulties James Russell and Christopher Jarrold Department of Experimental Psychology, Cambridge University, U.K. Lucy Henry Department of Psychology, Reading University, U.K. We asked whether children with autism are specifically impaired on tests of working memory. Experiment 1 showed that children with autism were at least as likely as normal children to employ articulatory rehearsal (criterion: evincing the "word length effect") and that they had superior spans to that of children with moderate learning difficulties. In Experiment 2, participants were given "capacity tasks" in order to examine group differences in the capacity of the central executive of working memory. The performance of the children with autism was inferior to that of the normally developing group and similar to that of the children with moderate learning difficulties. Copyright © 1996 Association for Child Psychology and Psychiatry. Keywords: Autism, executive functions, working memory, mental handicap Abbreviations: ID/ED, the Intra-dimensional/Extradimensional task; TOH, the Tower of Hanoi task; TOL, the Tower of London task; VMA, verbal mental age; WCST, the Wisconsin Card Sorting Task It is now well established that children with autism are specifically impaired on tests of executive functioning. Such tests require on-line planning, shifting of atten- tional set and the inhibition of incorrect responses that are "prepotent" because salient features of the environ- ment call them out or because past learning has made them difficult to abandon. In different ways, such tests require top-down, goal-directed processing; and the mental operations necessary for completing them are assumed to be those carried forward by the pre-frontal cortex (Shallice, 1988). Individuals with autism perform poorly compared to controls matched for verbal mental age on the Tower of Hanoi (TOH) planning task, a task that requires the generation and the holding in mind of future moves (Ozonoff, Pennington & Rogers, 1991) and on its (derivative the Tower of London (TOL) task (Hughes, Russell & Robbins, 1994). Their difficulties with second-order "theory of mind" tasks (e.g. "A thinks that B thinks that X"; Happe, 1994; Ozonoff et al, 1991) may also be executive in nature insofar as they require the on-line, strategic manipulation of data (see Tager- Flusberg & Sullivan, 1994 for a study in support of this view). They perform poorly on the Wisconsin Card Sorting Task (WCST) that requires a shifting of response category (Ozonoff et al, 1991) and on the intra- dimensional/extra-dimensional (ID/ED) shift task requir- Requests for reprints to: James Russell, Department of Experimental Psychology, Cambridge University, Downing Street, Cambridge CB2 3EB, U.K. ing a similar kind of shift (Hughes et al, 1994). Their difficulties with inhibiting prepotent responses are evident in their performance on tasks requiring subjects to ignore a goal object and to refer instead to an adjacent, empty location (the "windows task": Russell, Mauthner, Sharpe & Tidswell, 1991; Hughes & Russell, 1993) or to ignore a goal object and make a seemingly arbitrary response (the "box task": Hughes & Russell, 1993). Their difficulties with appearance-reality judgements about trick objects (Baron-Cohen, 1989) also yield to an explanation in terms of prepotency. The reason why specific impairments in "mental- ising" (e.g. Baron-Cohen, Tager-Flusberg & Cohen, 1993) and in executive control co-exist in autism is not clear. Some have argued that "theory of mind" impairments are sui generis and can be regarded separately from executive problems (Leslie & Thaiss, 1992); while others have argued that social (Pennington, 1994) and mentalising (Russell, 1995, 1996) impair- ments may be caused by early executive impairments. But whatever the relationship between these two impairments, it is important to determine which components of the executive system are impaired in the disorder. The principal aim of the present investigation was to ask whether working memory is one of the components of the executive system impaired in autism. The reason for doing this is that all of the executive tasks described above depend upon the adequate functioning of the working memory system. In every case, an arbitrary rule has to be held in mind (e.g., sort by colour; point to the empty box) while the influence of a prepotent response 673