Why expert performance is special and cannot be extrapolated from studies of performance in the general population: A response to criticisms

Highlights

• Current performance, “what is”, differs from performance after training, “what could be”.

• Among expert performers general cognitive ability does not correlate with performance.

• The heritability of expert performance is currently unknown.-Estimated hours of self-reported practice is a poor predictor of overall effects of training.

Abstract

Many misunderstandings about the expert-performance approach can be attributed to its unique methodology and theoretical concepts. This approach was established with case studies of the acquisition of expert memory with detailed experimental analysis of the mediating mechanisms. In contrast the traditional individual difference approach starts with the assumption of underlying general latent factors of cognitive ability and personality that correlate with performance across levels of acquired skill. My review rejects the assumption that data on large samples of beginners can be extrapolated to samples of elite and expert performers. Once we can agree on the criteria for reproducible objective expert performance and acceptable methodologies for collecting valid data. I believe that scientists will recognize the need for expert-performance approach to the study of expert performance, especially at the very highest levels of achievement.

Introduction

Contributors to the special issue on “Development of Expertise” criticized Ericsson, Krampe, and Tesch-Römer's (1993) theoretical framework for studying expert performance. In this response I will propose that many of these unflattering descriptions are due to the fact that the Expert-Performance framework is qualitatively different from the traditional theoretical frameworks for relating individual differences in ability. The fundamental controversy concerns the relation between normal and beginning levels of performance and the attainable expert levels of performance. The researchers working within the individual differences framework argue that the structure of expert performance can be extrapolated from the performance–ability relations observed in the general adult population. The structure of expert performance merely corresponds to extreme cases on the underlying ability distributions, and thus ability differences in general cognitive ability remain predictive of performance at the highest level. In contrast, the expert-performance framework hypothesizes that new cognitive mechanisms are gradually acquired during the extended period and they mediate the superior performance, thus leading to qualitative differences in structure compared to untrained performance.

During the last 20 years since our article in 1993 (Ericsson et al., 1993) peoples' conceptions about expertise and expert performance have changed. In the early 1990s, when our paper was written, the contemporary theories of skill acquisition (Anderson, 1982, Fitts and Posner, 1967) at that time proposed accounts for the acquisition of everyday skills such as driving a car, typing, and navigating in an unfamiliar environment. For these skills most individuals had the primary goal of reaching a level of proficiency that allowed them to perform these tasks adequately and effortlessly. At that time typical laboratory studies of skill acquisition were limited to 1 or 2 h per study and hardly ever lasted for more than 50 h total. A prototypical study involved choosing one of 4–16 alternatives. The task needed to be simple enough to be explained to college students in less than 15 min by reading instructions followed by a short period of warm-up. In their review of skill acquisition Fitts and Posner (1967) summarized the general structure of this type of skill acquisition and Ackerman (1987) proposed how individual differences in basic abilities would be predicted to correlate with performance at the different stages. During the first introductory phase of learning the skill (Fitts & Posner, 1967), beginners try to understand the requirements of the activity and focus on generating adequate actions while avoiding gross mistakes. During this phase individual differences in participants' performance were correlated with their general cognitive abilities (Ackerman, 1987). In the second phase of skill acquisition, once people had accumulated more task experience, salient mistakes become increasingly rare, sequences of actions are generated in a smoother fashion, and learners no longer need to concentrate to maintain an acceptable performance. Individual differences in participants' performance during this phase correlate with their tested spatial ability (Ackerman, 1987). During the third phase of learning individuals' performance skills become increasingly automated, and they are able to execute their skills smoothly and with minimal effort. As a consequence of automatization, performers lose their ability to control the execution of those skills, which makes intentional modifications and adjustments difficult. In the automated phase of learning, performance reaches a stable plateau, and no further improvements are typically seen. When the performance is triggered automatically, thus short-circuiting the cognitive and spatial representations, the individual differences in performance will primarily reflect the motor processes and the associated motor abilities (Ackerman, 1987). Drawing on Shiffrin and Schneider's (1977) theory for automation of consistent reactions Ackerman (1987) proposed that during acquisition of these types of skills the cognitive and spatial factors were eventually short-circuited and individual differences in these basic abilities would not correlate with resulting performance measured by the speed of the responses.

In contrast to the acquisition of everyday skills Ericsson et al. (1993, p. 363) proposed the study of objective reproducible performance of “exceptional individuals, whose performance in sports, the arts, and science is vastly superior to that of the rest of the population”. We found that acquiring these high levels of performance required years and even decades of demanding practice, a finding that has now been replicated in every type of expert performance that has been studied to date. We argued that with the extended periods of intense practice there is room for circumvention, adaptation and/or fundamental change in basic abilities: “the perceptual and motor systems show great adaptability in response to extended practice (a phenomenon discussed earlier in this article), it may be inappropriate to generalize the findings from relatively simple tasks involving 2–20 h of practice to expert performance acquired during a 10-year period of intense preparation” (p. 396). Expert performers in music, chess, and sports are constantly learning new things. Thus they maintain cognitive control over their performance and their performance does not tend to become fully automated. Ericsson et al. (1993) commented on how expert performers acquire domain-specific memory skills to recall past game situations, plan future actions, and evaluate their current performance. As is illustrated in Fig. 1, Ericsson et al. (1993) argued that expert performance requires the acquisition of new cognitive structures to enhance domain-specific performance and that “experts can acquire cognitive skills enabling them to circumvent the limits of short-term memory capacity and serial reaction time. This research rules out the hypothesis that individual differences in those functions will influence and constrain final adult performance” (p. 396). In contrast to many proposals that individual differences in working memory are limited by a basic general capacity for transient short term storage (Baddeley, 1986), Ericsson and Kintsch (1995) showed that expert performers develop long-term working memory (LTWM), where information is rapidly stored in long-term memory (LTM) associated with retrieval cues that allow the expert to access this information efficiently whenever the information is relevant for processing. Given that these skills are constructed based on available knowledge in LTM and choices about encoding methods, there will be substantial individual differences in the structure of the acquired skills, which have been demonstrated with experimental methods especially for memory experts (Ericsson and Kintsch, 1995, Hu and Ericsson, 2012). Investing a large proportion of one's lifetime in challenging deliberate practice activities causes profound qualitative changes in physiology and psychology.

Because expert performance is qualitatively different than other types of human performance, it requires the study of reproducibly superior performance on representative tasks that capture the essence of expertise in real world domains. Most commentators seem to accept our definition of expert performance and at least some of its implications. For example, Plomin, Shakeshaft, McMillan, and Trzaskowski (2014) acknowledge that: “It is important to understand the origins of expertise as it exists in the real world of sports, arts and skills” (p. 47). Ironically, however, Plomin et al. (2014) then go on to study precocious performance of 12-year-old children, who perform at a very high level on a reading test compared to their age-matched peers, that nearly everyone eventually masters within a few years. However, this level of performance is only precocious and six years later the majority of all of the other students are able to match their performance. In this response I will use the definition of Ericsson and Lehmann (1996, p. 277) and refer to expert performance “as consistently superior performance on a specified set of representative tasks for a domain” without any age conditions. Unfortunately other investigators use the term differently. For example, Wong, Palmeri, and Gauthier (2009) acknowledge that real-world expertise may differ but that they can produce performance showing “perceptual expertise” for classifying artificial figures after 2–10 h of training and call these participants “Ziggerin experts”. Other investigators refer to children with more knowledge about dinosaurs as experts (Gobbo & Chi, 1986). Finally, the requirement that experts display measureable superior performance in their domain is not met by many types of expertise. For example, Wai (2014) does not specify the nature of the reproducibly superior objective performance that his senators and billionaires are able to demonstrate in comparison to their peers. In many domains individuals referred to as experts by their peers are often unable to perform at a superior level to their peers on representative tasks from the domain (Ericsson, 2006a, Ericsson, 2009, Ericsson and Lehmann, 1996).

All papers in this issue criticize my earlier research for attributing too much emphasis to the effects due to deliberate practice. In our original paper, Ericsson et al. (1993) were very explicit that there might be other types of individual differences than those linked to innate talent: “It is quite plausible, however, that heritable individual differences might influence processes related to motivation and the original enjoyment of the activities in the domain and, even more important, affect the inevitable differences in the capacity to engage in hard work (deliberate practice)” (p. 399). The ability to engage in deliberate practice is an obvious requirement for improving performance through deliberate practice, but not all individuals may be able or willing to do so.

I find it disappointing that so many contributors seem to have misinterpreted quotes, where we argued that deliberate practice was necessary and based on our original study could explain “the major facts about the nature and scarcity of exceptional performance” (Ericsson et al., 1993, p. 392, italics added). They also seem to have overlooked many paragraphs where we tried to make our views on innate talent as clear as possible, such as this quote from our target paper in the High Ability Studies' issue on “Giftedness and Expertise”—a paper that most extensively addressed the issues of the special issue of Intelligence, but that Hambrick et al. (2014) seem unaware of and therefore did not cite:

“A common misconception of the expert performance framework is that this approach denies the possibility that differences in innate talent could ever be able to explain individual differences in attainable performance. The expert performance framework merely requires that valid evidence for innate talents must be presented and reviewed before it is accepted. This framework has long acknowledged the possibility that individual genetic differences might causally explain individual differences in elite achievement. However, according to recent reviews (Ericsson, 2007a, Ericsson, 2007b, Ericsson, 2007c) no evidence currently exists, with the exception of height and body size” (Ericsson, Roring, & Nandagopal, 2007, pp. 40–41).

In my response I will pursue this same argument as I examine the proposed empirical evidence in support of innate abilities and talent that would constrain, and thus correlate with the performance that an individual can attain in a domain of expertise. The central goal of Ericsson et al. (1993) and this response remains to identify immutable constraints on the acquisition of various types of expert performance that cannot be overcome or circumvented by the most effective forms of practice (deliberate practice) in order to provide “unique evidence on the potential and limits of extreme environmental adaptation and learning” (p. 363).

In the second section of the main body of my reply I will review the cited studies reporting significant relations between performance on tests of cognitive ability and domain-specific performance (Ackerman, 2014, de Bruin et al., 2014, Grabner, 2014, Hambrick et al., 2014). I will show that the findings are consistent with our proposal for the acquisition of expert performance, where acquired mechanisms gradually circumvent the role of any basic general cognitive capacities and thus reduce and even eliminate significant relations between general cognitive ability and domain-specific performance at the expert level of performance.

In the third section I will argue that there are no available estimates on heritability for attained expert performance in twins, when we define expert performance as reproducibly superior performance in a domain of expertise independent of age. I will also show that the other evidence cited by Plomin et al. (2014) for genes being necessary for attaining expert performance is currently lacking, with the long established exception of genetic effects on height and body size.

In the fourth section I will review the claim that precocious performance on Scholastic Achievement tests in Mathematics (SAT-M) in the 8th grade is predictive of expert adult performance in mathematics and science (described by Wai, 2014). I will show that these findings can be accounted for within the expert-performance framework without acknowledging that exceptional innate mental capacities are necessary to attain expert levels of performance.

In the fifth section I will discuss the relation between self-reported amounts of practice and current domain-specific performance. The expert-performance approach is consistent with Hambrick et al.'s (2014) first conclusion namely that the reported amount of practice is significantly correlated with current level of performance in music and chess. However, Hambrick et al.'s (2014) second conclusion concerns the relation of attained performance and the total amount of practice during the individuals' past life. Their analysis ignores the effects of forgetting, injuries, and accidents, along with the differential effects of different types of practice at different ages and levels of expert performance.

In a sixth and final section I will more briefly discuss a few remaining issues.