Capturing the Naturally Occurring Superior Performance of Experts in the Laboratory Toward a Science of Expert and Exceptional Performance

Expertise researchers have traditionally shied away from studying the highest levels of achievement in favor of studying basic cognitive processes, such as memory and categorization. In this article, we present a different approach that is focused on capturing superior (expert) performance on representative tasks that reveal the essential characteristics of expertise in a given domain. In domains where expert performance is measurable, acquisition is gradual and the highest levels are only attained after 10 years of intense preparation—even for the most ‘‘talented.’’ Analyses of reproducibly superior performance show that it is mediated by physiological adaptations and cognitive skills acquired as a result of the cumulative effects of special practice activities (deliberate practice). It appears that the genes necessary to attain such adaptations and expert skills can be activated in healthy children—the only clear exceptions to date being genes that control body size and height. Our knowledge of how experts acquire their superior skills provides insights into the potential for human adaptation and skill acquisition and has important implications for theories of the structure of general and expert cognition, as well as for training interventions in applied psychology and education. KEYWORDS—expertise; skill acquisition; experience; deliberate practice; physiological adaptation The achievements of experts and eminent individuals are among the most admired and complex phenomena in our culture. They offer serious challenges to scientists attempting to study them. For example, how can a researcher fully understand the cognitive processes in which a chess master engages while generating moves when the researcher is unable defeat the master? Those who assume that eminent individuals are endowed with innately superior capacities will also question whether lesser-endowed scientists could fully understand the master’s associated thought processes. As a consequence, there is a common misconception that the only legitimate judges of experts are their peers. When the study of expertise flourished in the 1970s and 1980s, the research focus was not on the cognitive processes that mediate experts’ outstanding achievements but on more basic memory performance in the laboratory. In the last review of expertise for this journal over a decade ago, Bedard and Chi (1992) described studies that primarily examined differences between experts and novices in memory for representative stimuli or categorization of problems from introductory academic courses, such as physics. Bedard and Chi (1992) even questioned the relationship between such laboratory studies and real-word expertise. They observed that some experts performed poorly ‘‘on some real-world problems’’ (p. 138) and even described situations in which ‘‘novices not only can perform as well as experts, but actually surpass the experts’’ (p. 138). Since their review, researchers have become increasingly dissatisfied with studies that define experts by social criteria (e.g., peer nominations), by extended domain experience, or by (high) education. When scientists measured the level of real-world performance of so-called experts they found that many were not reliably better than their less-experienced colleagues. For example, reviews (Ericsson, 2006) have described studies showing that a psychotherapist’s level of education and clinical experience is unrelated to their treatment outcomes. Recent reviews of performance in health-related fields show that extended experience of doctors and nurses Address correspondence to K. Anders Ericsson, Department of Psychology, Florida State University, 1107 West Call St., P.O. Box 3064301, Tallahassee, Florida 32306-4301; e-mail: [email protected]. CURRENT DIRECTIONS IN PSYCHOLOGICAL SCIENCE 346 Volume 16—Number 6 Copyright r 2007 Association for Psychological Science (beyond the first couple of years of practice) is not associated with continued improvements, as most people had thought. In fact, performance has frequently been found to decrease without continued training (Choudhry, Fletcher, & Soumerai, 2005; Ericsson 2004). There are numerous domains of expertise in which some performers are reproducibly superior to most others engaging in the domain. For example, chess masters, bridge experts, and experts in other games consistently beat less-skilled individuals. More generally, in the arts and the sciences some individuals reliably generate superior products that are selected for presentation in journals, conferences, and competitions. Some individuals are able to perform at a reproducibly higher level in professional domains too. Our framework focuses on understanding the mechanisms that mediate consistently superior performance, and to distinguish this approach from the traditional study of expertise we will refer to it as the expert-performance approach (see Feltovich, Prietula, & Ericsson, 2006, for a description of the historical development of several approaches to the study of expertise using psychological methods). CAPTURING REPRODUCIBLY SUPERIOR PERFORMANCE UNDER STANDARDIZED CONDITIONS In everyday life, each expert encounters unique challenges that make it nearly impossible to compare levels of performance between different experts. For example, some doctors treat clients with more complex and difficult problems, and some sales people may be assigned better territories. The key challenge for anyone interested in individual differences in expert performance is to measure highly representative performance under controlled conditions that are standardized for everyone tested. In a groundbreaking series of studies, de Groot (1978) was able to capture the superior performance of world-class chess players using standardized tasks. The ultimate criterion for eminence in chess is consistent tournament success when competing against similarly skilled chess players. To reliably estimate a chess player’s rating on an interval scale, one would need outcomes from 20 to 50 tournament matches (roughly 40–100 hours of play). In contrast, de Groot (1978) identified critical chess positions from actual matches played between chess masters and recreated these situations under controlled conditions by presenting them to chess players during an individual testing session. In this pioneering approach, de Groot (1978) also asked chess players to ‘‘think aloud’’ while they generated the best possible next move. Differences in the quality of selected moves distinguished world-class from skilled club players. Subsequent research with large groups of chess players has shown that the ability to select the next best move is closely related to tournament ratings (van der Maas & Wagenmakers, 2005). More interestingly, de Groot (1978) was able to pinpoint the thought processes that led to the experts’ superior move selection by analyzing their think-aloud protocols. In the early 1990s, the expert-performance approach, based on de Groot’s paradigm, was initially proposed and later elaborated (Ericsson, 2006). In this approach, experts’ real-world performance is scrutinized to identify naturally occurring situations that require immediate action and that capture the experts’superior selection or execution of actions in the associated domain, as is illustrated in Figure 1. These situations are then standardized by recreating them in the laboratory, and experts and less skilled individuals are instructed to respond in real time by acting appropriately. By instructing participants to think aloud and collecting other observable evidence on their thought processes, such as eye Fig. 1. Model for capturing expert performance by representing naturally occurring tasks in standardized form in the laboratory. The top of the figure shows the stream of naturally occurring activities of experts, within which it is possible to identify a challenging situation requiring immediate action—such as making moves for chess positions during tournaments or returning a serve in a tennis match. As shown in the bottom part of the figure, one can present a chess position from an actual game and ask for the best next move in a laboratory setting. Alternatively, one can project a film of a player serving a tennis ball and ask the participants to attempt to return the ‘‘imaged ball’’ with the best shot using their racket. These representative tasks can then be presented to participants differing in skill, and reaction times, accuracy, and process-tracing data—such as eye-fixations and concurrent or retrospective verbal reports—can be recorded to permit explication of mechanisms that can account for the reproducible differences in performance. Volume 16—Number 6 347 K. Anders Ericsson and Paul Ward