Journal of Experimental Psychology: Learning, Memory, and Cognition 1990, Vol. 16, No. 2, 291-304 Copyright 1990 by the American Psychological Association, Inc. O278-7393/9O/SOO.75 An Analysis of Memory-Based Theories of Automaticity David L. Strayer and Arthur F. Kramer University of Illinois at Urbana-Champaign Memory-based theories of automaticity predict that performance in a memory search task which is automatic will not require a representation of the memory set in working memory. The information contained in working memory was manipulated by inserting an interference task between the presentation of a memory set and a probe stimulus in a memory search task. The interference task prevented rehearsal, necessitating the retrieval of the memory set from long- term memory in variably mapped (VM) conditions. Performance in consistently mapped (CM) conditions provided strong support for memory-based theories of automaticity. With CM practice, both the effects of memory load and the effect of the interference task were eliminated. Furthermore, there was a temporal coupling in the reduction of these two effects with consistent practice. Monte Carlo simulations of memory-based automaticity predict such a temporal coupling. Automaticity is viewed as a continuum reflecting the relative contribution of the direct memory access of past solutions from long-term memory on performance. The changes in information processing accompanying the acquisition of cognitive skill are of fundamental importance to the understanding of human behavior. A large part of skill is thought to be the development, integration, and coordina- tion of a number of component automatic processes (e.g., Kramer & Strayer, 1988; LaBerge & Samuels, 1974; Logan, 1985; Myers & Fisk, 1987). In recent years a number of theories have been proposed to account for the quantitative and qualitative changes that occur during the development of automatic processing (e.g., LaBerge & Samuels, 1974; Logan, 1988b; Neumann, 1984; Posner & Snyder, 1975; Shiffrin & Schneider, 1977). In the theory developed by Shiffrin and Schneider (1977; Schneider & Shiffrin, 1977) two modes of information proc- essing were characterized. The first, controlled processing, represents a temporary sequence of operations that are under the control of the subject, require active attention, and are capacity limited. Controlled processing is used in novel situ- ations and in situations in which stimulus-response relations are varied over time. Shiffrin and Schneider referred to this as varied mapping (VM). The second mode of processing described by Shiffrin and Schneider was automatic processing. Automatic processing is fast, often insensitive to capacity limits, and is difficult to modify once initiated. Automatic processing develops as a result of practice with consistent stimulus-response relations. Shiffrin and Schneider referred This article is based on the doctoral dissertation research of David L. Strayer. The research was supported by NASA (National Aeronau- tics and Space Administration) Ames Research Center under Contract NASA NAG-2-308, with Sandra Hart as the technical monitor and by the Office of Naval Research under Contract N0O014-89-J-1493, with Harold Hawkins as contract monitor. We would like to thank Ritske De Jong, Gordon Logan, Colin MacLeod and Delos Wickens for comments on a previous version of this article. We are also indebted to Jean Buckley, William Garison, and Karen Seidler for their assistance in testing subjects. Correspondence concerning this article should be addressed to David L. Strayer, Department of Psychology, 603 E. Daniel St., Champaign, Illinois 61820. to this as consistent mapping (CM). Consistency of practice is a necessary condition for the development of automaticity (Logan, 1979; Schneider & Fisk, 1982). An important issue for theories of automaticity, and one that is the focus of the present research, concerns the changes in information processing that underlie the development of automaticity. A recent approach to this issue views automa- ticity as a memory phenomenon (Logan, 1988a, 1988b; Schneider, 1985; Schneider & Detweiler, 1987, 1988). Ac- cording to memory-based theories, automaticity involves re- trieving information from a knowledge base established through practice. Initial performance is limited by a lack of knowledge, and the development of automaticity is the result of an increasing knowledge base which facilitates memory retrieval. The memory-based theory proposed by Logan assumes that "each encounter with a stimulus is encoded, stored, and retrieved separately" (Logan, 1988b, p. 493). According to Logan's instance theory, the development of automaticity reflects a transition from the use of a generic set of cognitive procedures (i.e., an algorithm) to the reliance on a direct memory access of past solutions. Instance theory postulates that performance is determined by the outcome of a race between a computational algorithm and the direct memory access. Initially, the algorithm is more rapid and reliable and dominates performance. However, after a number of trials (which form multiple instances in memory), the direct mem- ory access finishes prior to the algorithm and dominates performance. The direct memory access speeds up because the minimum retrieval time decreases as the number of instances in memory increases. Schneider (1985; Schneider & Detweiler, 1987, 1988) also proposed a memory-based theory of automaticity. Unlike Logan's instance theory, Schneider's is a strength theory. According to this view, consistent practice serves to strengthen the input-output relations. Following consistent practice, ac- cessing the trace is more rapid, accurate, and reliable than the algorithm. In contrast, because the input-output relations vary under VM practice, the trace is so weak that the algo- rithmfinishesprior to the direct memory access. Thus regard- 291 This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.