Elsevier

Cognition

Volume 116, Issue 3, September 2010, Pages 321-340
Cognition

Implicit learning as an ability

https://doi.org/10.1016/j.cognition.2010.05.011Get rights and content

Abstract

The ability to automatically and implicitly detect complex and noisy regularities in the environment is a fundamental aspect of human cognition. Despite considerable interest in implicit processes, few researchers have conceptualized implicit learning as an ability with meaningful individual differences. Instead, various researchers (e.g., Reber, 1993; Stanovich, 2009) have suggested that individual differences in implicit learning are minimal relative to individual differences in explicit learning. In the current study of English 16–17 year old students, we investigated the association of individual differences in implicit learning with a variety of cognitive and personality variables. Consistent with prior research and theorizing, implicit learning, as measured by a probabilistic sequence learning task, was more weakly related to psychometric intelligence than was explicit associative learning, and was unrelated to working memory. Structural equation modeling revealed that implicit learning was independently related to two components of psychometric intelligence: verbal analogical reasoning and processing speed. Implicit learning was also independently related to academic performance on two foreign language exams (French, German). Further, implicit learning was significantly associated with aspects of self-reported personality, including intuition, Openness to Experience, and impulsivity. We discuss the implications of implicit learning as an ability for dual-process theories of cognition, intelligence, personality, skill learning, complex cognition, and language acquisition.

Introduction

The ability to automatically and implicitly detect complex and noisy regularities in our environment is a fundamental aspect of human cognition. Much of this learning takes place on a daily basis without our intent or conscious awareness, and plays a significant role in structuring our skills, perceptions, and behavior (Hassin et al., 2005, Kihlstrom, 1987, Lewicki et al., 1987, Lewicki and Hill, 1987, Reber, 1967, Reber, 1993, Stadler and Frensch, 1997). This type of learning is often referred to as implicit learning (Reber, 1967, Reber, 1993, Stadler and Frensch, 1997) and is typically characterized by a set of automatic, associative, nonconscious, and unintentional learning processes, as distinguished from the conscious, deliberate, and reflective learning processes that are thought to be associated with executive functioning and working memory (e.g., Barrett, Tugade, & Engle, 2004).

Despite considerable interest in implicit processes, few researchers have conceptualized implicit learning as an ability. While researchers of the cognitive unconscious have investigated the nature of the unconscious using the experimental approach, they have tended to treat individual differences as “noise” (error or otherwise unexplained variance), or have posited that whatever individual differences in implicit cognition do exist are minimal relative to individual differences in explicit cognition (Reber, 1993; Stanovich, 2009). For example, in distinguishing between the “algorithmic mind” and the “autonomous mind”, Stanovich (2009) states that “…continuous individual differences in the autonomous mind are few. The individual differences that do exist largely reflect damage to cognitive modules that result in very discontinuous cognitive dysfunction such as autism or the agnosias and alexias (p.59).” As a consequence of these long-held assumptions, little research has investigated whether there exist meaningful individual differences in implicit learning or the correlates of such individual differences. In the current study we investigated the association of implicit learning ability with a variety of cognitive and personality variables, building on previous research examining the relation of implicit learning to psychometric intelligence, basic cognitive mechanisms, and personality traits. We take up discussion of each association in turn.

In investigating the relation between implicit learning and intelligence, researchers have relied on measures of psychometric intelligence, defined as Spearman’s general intelligence, or g, the common variance across disparate tests of cognitive ability (Spearman, 1904). What is the link between implicit learning and g? According to Reber, 1989, Reber, 1993, Reber and Allen, 2000, individual differences in implicit learning should be expected to be largely independent of individual differences in psychometric intelligence. The argument is based on the assumption that implicit learning is evolutionarily older than explicit cognition, with the latter developing only with the rise of Homo sapiens. The older mechanisms of implicit learning are believed to have been unaffected by the arrival of explicit cognition, which includes hypothesis-guided learning and deduction, and they continue to function independently of one another today. These thoughts converge with arguments advanced by Mackintosh and colleagues (Mackintosh, 1998, McLaren et al., 1994) that the processes underlying performance on implicit learning tasks may be automatically associative rather than cognitive in nature, and are consistent with various other dual-process theories of human cognition (Chaiken and Trope, 1999, Epstein et al., 1996, Evans and Frankish, 2009, Sloman, 1996, Stanovich and West, 2000).

Thus far, the evidence suggests that performance on implicit learning tasks is independent of differences in IQ, or at most only weakly related. Some paradigms have never shown an association with psychometric intelligence (e.g., artificial grammar learning; Gebauer and Mackintosh, 2007, McGeorge et al., 1997, Reber et al., 1991), whereas for other paradigms the majority of studies have not found a significant association (e.g., serial reaction time learning; Feldman et al., 1995, Unsworth et al., 2005; but see Salthouse, McGuthry, & Hambrick, 1999). The relation between IQ and one other implicit learning paradigm, which involves incidental exposure to pictures, has been investigated only once but was significant (Fletcher, Maybery, & Bennett, 2000). A possible explanation for the occasional significant association between IQ and implicit learning is that different implicit learning paradigms are only weakly correlated with one other (Gebauer and Mackintosh, 2007, Gebauer and Mackintosh, in preparation, Pretz et al., 2010, Salthouse et al., 1999) and may differ in the extent to which they are measuring implicit learning without relying on explicit processes (e.g., Seger, 1994).

Direct comparisons of implicit and explicit versions of specific tasks may further help to explain contradictory results. In some studies, researchers administered the same implicit learning task under two conditions: in one condition, participants were explicitly instructed to detect the underlying covariation, and in the other condition participants did not receive such an instruction, thereby making learning ‘incidental’ to the task requirements. In these studies, psychometric intelligence was more highly correlated with the task under explicit instructions than under incidental conditions (Unsworth and Engle, 2005a, Gebauer and Mackintosh, 2007). Similarly, in study of 455 adolescents, Feldman et al. (1995) separated an intentional declarative component of an implicit learning task from the procedural component and found that, although the declarative learning component significantly correlated with psychometric intelligence, the procedural component did not. Overall it appears that individual differences in psychometric intelligence, which are clearly associated with variation in explicit cognition, are either weakly or not at all associated with variation in implicit learning (e.g. McGeorge et al., 1997, Reber et al., 1991).

While implicit learning is only weakly related to psychometric intelligence, recent research suggests that individual differences in implicit learning may make an independent contribution to complex cognition. Gebauer and Mackintosh (in preparation) administered a battery of 15 traditional implicit learning tasks and nine traditional psychometric intelligence tests to 195 German school pupils. Factor analyses revealed a low correlation between two second-order principal components, the first corresponding to psychometric intelligence and the second corresponding to implicit learning. In addition, their second-order factor of implicit learning correlated significantly with school grades in Math and English (a foreign language for the German participants in the study). Controlling for psychometric intelligence, the correlation between the implicit learning factor and English grades remained, while the relation to Math was no longer significant. Similarly, Pretz et al. (2010) found a significant relation between a measure of serial reaction time (SRT) probabilistic learning and Math and English achievement scores. These results suggest there may be variance in implicit learning ability that is independent of psychometric intelligence but nevertheless related to some aspects of school learning.

A number of basic cognitive mechanisms, including working memory, explicit associative learning, and processing speed, have been posited as contributors to intelligence (e.g., Kaufman, DeYoung, Gray, Brown, & Mackintosh, 2009). Examining their relations to implicit learning may help to clarify the relation of implicit learning to other aspects of cognition. Below we review the available evidence on the relation of these cognitive mechanisms to implicit learning.

Working memory is defined as the ability to maintain, update, and manipulate information in an active state, over short delays. It depends heavily on executive attention; those with a high working memory are better able to control their attention, maintaining task goals in the presence of interference (Conway et al., 2001, Kane et al., 2001, Unsworth et al., 2004). Over the last two decades, considerable debate has arisen over the question of whether implicit learning, like working memory, depends on the executive functions of attention, or whether it arises automatically as a by-product of processing a set of correlated events (Jiménez, 2003, Shanks, 2003). Much experimental work (e.g., Baker et al., 2004, Frensch and Miner, 1995, Jiang and Chun, 2001, Jiménez and Mendez, 1999, Turke-Browne et al., 2005) appears to be converging on the conclusion that, for implicit learning to occur, selective attention to the relevant stimuli is required. However, learning about the stimuli that are selectively attended to then occurs automatically, regardless of an intention to learn, and without necessitating any further executive processing resources.

One implication of this conclusion is that central executive resources should be engaged under explicit learning instructions, to guide the focus of attention, whereas only selection processes should be required for incidental learning (Cowan, 1988, Cowan, 1995, Frensch and Miner, 1995, Johnson and Hirst, 1993). Working within this framework, Unsworth and Engle, 2005a, Unsworth and Engle, 2005b demonstrated that working memory differences emerge in an implicit learning task under explicit instructions to detect the covariation, but not under incidental conditions where no such instructions were given. Similarly, Feldman et al. (1995) found nonsignificant correlations between implicit learning and measures of working memory.

In sum, the available data suggest that implicit learning operates in an automatic fashion once relatively low-level perceptual attention is selectively allocated to the appropriate stimuli, without necessarily requiring executive attention. This leads us to hypothesize that individual differences in implicit learning are not associated with individual differences in working memory.

Associative learning, as conceived in the implicit learning literature, differs from the type of associative learning typically discussed in the intelligence literature (e.g., Underwood et al., 1978, Williams and Pearlberg, 2006). In the implicit learning literature, learning is often termed as associative (as opposed to cognitive), when learning proceeds incidentally, because it describes the incidental formation of associations. Connectionist modeling based on this assumption has successfully modeled many aspects of implicit learning (e.g., Cleeremans, 1993, Cleeremans and Dienes, 2008). By contrast, in the intelligence literature, associative learning is often used to describe the learning of associations acquired consciously and intentionally according to explicit instruction and feedback. To date, no study has investigated the relationship between implicit learning and explicit associative learning. Although prior studies have ostensibly compared explicit and implicit learning (e.g., Reber et al., 1991), measures of “explicit learning” in these studies have typically been measures of explicit reasoning, such as series completion, that do not, in fact, require learning over the course of the experiment. Despite the fact that both implicit learning and explicit associative learning must involve the formation of associations, we hypothesized that they are unrelated as abilities, for the same reasons that working memory seems likely to be unrelated to implicit learning: executive attention should be required only when learning is intentional.

Processing speed involves the speed at which very simple operations can be performed. Differences in intelligence may partly reflect the overall efficiency and speed of the nervous system (Anderson, 1992, Jensen, 1998), in addition to more specific capabilities like working memory (Kaufman et al., 2009). Given the primitive and broad nature of processing speed as a parameter, one might expect it to be related to individual differences in implicit learning, even in the absence of implicit learning’s association with more complex cognitive mechanisms. Accordingly, Salthouse et al. (1999) found a significant relation between two processing speed measures and implicit learning. One of these measures was the Digit-Symbol Coding test, part of the standard WAIS battery for IQ. Although the factor structure of the WAIS indicates a processing speed factor as one of four second-level factors below g, the processing speed tests load on g more weakly than other types of test (Deary, 2001). We therefore expected that, although implicit learning may be only weakly or not at all related to g, it may show a significant relation to processing speed.

Research on the personality correlates of implicit learning is limited. However, theoretical links between implicit learning and intuition allowed us, in conjunction with the available evidence, to make predictions regarding personality traits reflecting an intuitive cognitive style, especially those related to the Big Five trait domain of Openness/Intellect and to impulsivity.

Implicit learning and intuition are closely related constructs. Indeed, it has been argued that intuition is the subjective experience associated with the accumulated knowledge gained through an implicit learning experience (Dienes, 2008, Lieberman, 2000, Reber, 1989). Reber (1989) further explains how implicit learning and intuition can be related:

To have an intuitive sense of what is right and proper, to have a vague feeling of the goal of an extended process of thought, to “get the point” without really being able to verbalise what it is that one has gotten, is to have gone through an implicit learning experience and have built up the requisite representative knowledge base to allow for such a judgement (p. 233).

Woolhouse and Bayne (2000) looked at the relation between personality as measured by the Myers-Briggs Type Indicator (MBTI) (Myers, McCaulley, Quenk, & Hammer, 1998), and performance on a hidden covariance detection task (Lewicki, Hill, & Sasaki, 1989), in which participants implicitly learned to judge the job suitability of job applicant personality profiles based on the covariance between personality profiles and information about job suitability in the training phase. A test phase with new profiles showed that participants learned the covariation regardless of whether they were explicitly aware of the rules. Individual differences emerged, however, when considering task performance along the MBTI dimension of intuition/sensation, which was designed to measure the extent to which people prefer to make decisions using factual, simple, and conventional methods (sensation) vs. a preference for the possible, complex, and original (intuition) (McCrae, 1994). Sensation types were more likely to be consciously aware of the covariation and apply it effectively. Among those who lacked awareness of the underlying rule, however, there was a tendency for participants with a more intuitive personality to make greater and more accurate use of their intuition on the implicit learning task. These authors concluded that personality influences whether people will use intuition based on implicit knowledge to help them arrive at a correct answer in the absence of explicit knowledge.

The five factor model or Big Five is the most widely used and best validated taxonomy of personality traits (Goldberg, 1990, Markon et al., 2005). Within the Big Five, the MBTI dimension of sensing-intuition falls within the domain of Openness/Intellect (McCrae, 1994). The compound label for this dimension reflects an old debate about how best to characterize this personality factor, with some researchers favoring the label “Intellect” (e.g., Goldberg, 1990) and others favoring “Openness to Experience” (e.g., Costa & McCrae, 1992). This debate has been largely resolved by the recognition that Openness and Intellect reflect separable but related aspects of the larger domain (Johnson, 1994, Saucier, 1992). This distinction was recently given more empirical support by the finding of two correlated factors within 15 scales measuring different lower-level facets of Openness/Intellect (DeYoung, Quilty, & Peterson, 2007). The two factors were clearly recognizable as Intellect and Openness, with Intellect reflecting a combination of perceived cognitive ability and tendency toward intellectual engagement, and Openness reflecting artistic and contemplative qualities and engagement with sensory and perceptual information. The analysis of DeYoung et al. (2007) generated new scales to measure Openness and Intellect separately and also demonstrated that different subscales of the NEO PI-R Openness to Experience scale (Costa & McCrae, 1992) could be used as markers of Openness and Intellect. McCrae (1994) found that the MBTI intuition scale was more strongly related to Openness than to Intellect, at the facet level.

Based on the link between Openness and intuition, we hypothesized that scales loading on Openness would be positively associated with implicit learning. Scales related to Intellect, in contrast, appear to be more closely linked to intelligence, working memory, and explicit associative learning (DeYoung et al., 2005, DeYoung et al., 2009). We hypothesized that they would be associated with these other cognitive abilities, but not with implicit learning.

In recent years, dual-process theories of reasoning have become increasingly required for explaining cognitive, personality, and social processes (see Evans & Frankish, 2009). Although the precise specifications of the theories differ, most have in common the idea that humans possess both automatic and controlled processes that jointly contribute to behavior. This idea has recently been elaborated on by Strack and Deutsch (2004) who argue that behavior is multiply determined by both impulsive and reflective processes.

Prior research shows that impulsivity is negatively related to both g and working memory (Kuntsi et al., 2004, Shamosh and Gray, 2007, Shamosh et al., 2008). Here we investigate the relation between implicit learning and impulsivity. According to Strack and Deutsch (2004), the impulsive system involves an associative network that is automatically activated through learning and experience. They argue that “structures emerge in the impulsive system that bind together frequently co-occurring features and form associative clusters (p. 223).” They further state that “the impulsive system has low flexibility but is fast and needs no attentional resources” (p. 224). This characterization strongly suggests that implicit learning ability might be positively associated with trait impulsivity.

Whiteside and Lynam (2001) identified four major dimensions of variance pertaining to impulsivity: urgency, lack of premeditation, lack of perseverance, and sensation seeking. We hypothesized that the most relevant form of impulsivity for implicit learning is lack of premeditation, in that individuals who deliberate extensively may do so in part because they are poor at detecting incidental covariances and therefore have reduced access to quick and intuitive decisions.

Section snippets

The present study

To investigate the cognitive and personality correlates of individual differences in implicit learning we used what we believe to be the best measure of implicit learning currently available (see Section 3). In line with the prior literature just reviewed, our hypotheses regarding the pattern of relations to implicit learning are as follows:

Hypothesis 1. Psychometric intelligence is correlated more strongly with explicit associative learning than with implicit learning.

Hypothesis 2. Implicit

Participants

The 153 participants (47 males and 106 females) included in the analysis were aged 16–18 years (Mean = 16.9, SD = .65), and attended a selective sixth form college (which takes high-achieving students who are in their last 2 years of secondary education) in Cambridge, England. Data were collected for 27 other participants, but 24 of these were removed from the analysis because they were missing implicit learning scores, 2 were removed because their Raven Advanced Progressive Matrices scores were

Validation

We first validated that implicit learning took place on the probabilistic serial reaction time (SRT) learning task. Fig. 2 shows learning on each block at the group level of analysis, comparing mean RT for trials that followed the most probable (85%) sequence with the mean RT for trials that do not follow the most probable sequence (15%).

A repeated-measures analysis of variance (ANOVA) with block (8) and type of trial (2, training vs. control) was conducted on the measures of RT. The results

Discussion

The current study conceptualized implicit learning as an ability and assessed the relation of individual differences in implicit learning to psychometric intelligence, elementary cognitive tasks commonly associated with psychometric intelligence, and personality. Contrary to the long-held assumption that individual differences in implicit cognition are minimal relative to individual differences in explicit cognition (e.g., Reber, 1993; Stanovich, 2009), meaningful individual differences in

Conclusion

Implicit learning can be assessed as an ability with individual differences that are meaningfully related to other important variables in individual differences research. Implicit learning ability was related to Openness to Experience and the associated construct, intuition, and to the tendency to make decisions without premeditation. Implicit learning ability was not related to psychometric intelligence, working memory, explicit associative learning, or self-rated Intellect. The pattern of

Acknowledgements

The authors would like to thank Sheila Bennett for her kind assistance in recruiting participants, Jim Blair and Nikhil Srivastava for computer support, and the administration at Hills Road Sixth Form College for the use of their facilities.

References (130)

  • B. Marshalek et al.

    The complexity continuum in the radex and hierarchical models of intelligence

    Intelligence

    (1983)
  • E. Norman et al.

    Fringe consciousness in sequence learning: The influence of individual differences

    Consciousness and Cognition

    (2006)
  • E. Norman et al.

    Gradations of awareness in a modified sequence learning task

    Consciousness and Cognition

    (2007)
  • J.E. Pretz et al.

    Measuring individual differences in affective, heuristic, and holistic intuition

    Personality and Individual Differences

    (2007)
  • J.E. Pretz et al.

    The effects of mood, cognitive style, and cognitive ability on implicit learning

    Learning and Individual Differences

    (2010)
  • A.S. Reber

    Implicit learning of artificial grammars

    Journal of Verbal Learning and Verbal Behavior

    (1967)
  • P.L. Ackerman

    Determinants of individual differences during skill acquisition: Cognitive abilities and information processing

    Journal of experimental psychology: General

    (1988)
  • J.R. Anderson

    Rules of the mind

    (1993)
  • M. Anderson

    Intelligence and development: A cognitive theory

    (1992)
  • Arbuckle, J. (2006). Amos 7.0 (Build 1140): Amos Development...
  • C.I. Baker et al.

    Role of attention and perceptual grouping in visual statistical learning

    Psychological Science

    (2004)
  • L.F. Barrett et al.

    Individual differences in working memory capacity and dual-process theories of the mind

    Psychological Bulletin

    (2004)
  • J.T. Cacioppo et al.

    The need for cognition

    Journal of Personality and Social Psychology

    (1982)
  • J.T. Cacioppo et al.

    Dispositional differences in cognitive motivation: The life and times of individuals varying in need for cognition

    Psychological Bulletin

    (1996)
  • J.B. Carroll

    Human cognitive abilities: A survey of factor-analytic studies

    (1993)
  • C.F. Chabris

    Cognitive and neurobiological mechanisms of the law of general intelligence

  • S. Chaiken et al.

    Dual-process theories in social psychology

    (1999)
  • F. Chang

    Implicit learning as a mechanism of language change

    Theoretical Linguistics

    (2008)
  • A. Cleeremans

    Mechanisms of implicit learning: Connectionist models of sequence processing

    (1993)
  • A. Cleeremans et al.

    Computational models of implicit learning

  • A.R.A. Conway et al.

    The cocktail party phenomenon revisited: The importance of working memory capacity

    Psychonomic Bulletin & Review

    (2001)
  • T.P. Corporation

    DAT for selection-verbal reasoning (European adaptation)

    (1995)
  • N. Cowan

    Evolving conceptions of memory storage, selective attention, and their mutual constraints within the human information-processing system

    Psychological Bulletin

    (1988)
  • N. Cowan

    Attention and memory: An integrated framework

    (1995)
  • A. Destrebecqz et al.

    Can sequence learning be implicit? New evidence with the process dissociation procedure

    Psychonomic Bulletin & Review

    (2001)
  • Destrebecqz, A., & Cleeremans, A. (2008). How do we find words in implicit artificial language learning? Paper...
  • C.G. DeYoung et al.

    Sources of openness/intellect: Cognitive and neuropsychological correlates of the fifth factor of personality

    Journal of Personality

    (2005)
  • C.G. DeYoung et al.

    Between facets and domains: 10 aspects of the Big Five

    Journal of Personality and Social Psychology

    (2007)
  • C.G. DeYoung et al.

    Intellect as distinct from Openness: Differences revealed by fMRI of working memory

    Journal of Personality and Social Psychology

    (2009)
  • Z. Dienes

    Subjective measures of unconscious knowledge

    Progress in Brain Research

    (2008)
  • C.D. Elliot

    British ability scales II

    (1996)
  • S. Epstein et al.

    Individual differences in intuitive-experiential and analytical–rational thinking styles

    Journal of Personality and Social Psychology

    (1996)
  • Epstein, S., Pacini, R., & Norris, P. (1998). The rational–experiential inventory: Long form. University of...
  • J.S.B.T. Evans et al.

    In two minds: Dual processes and beyond

    (2009)
  • J. Fletcher et al.

    Implicit learning differences: A question of developmental level?

    Journal of Experimental Psychology: Learning, Memory, and Cognition

    (2000)
  • P.A. Frensch et al.

    Zur Rolle des Arbeitsgedächtnisses beim impliziten Sequenzlernen [The role of working memory in implicit sequence learning]

    Zeitschrift für Experimentelle Psychologie

    (1995)
  • H. Gardner

    Frames of mind: The theory of multiple intelligences

    (1993)
  • G.F. Gebauer et al.

    Psychometric intelligence dissociates implicit and explicit learning

    Journal of Experimental Psychology: Learning, Memory, and Cognition

    (2007)
  • Gebauer, G. F., & Mackintosh, N. J. (in preparation). Implicit learning and intelligence: A principal component...
  • Cited by (327)

    • Unraveling sequence learning in obsessive–compulsive disorder

      2023, Current Opinion in Behavioral Sciences
    View all citing articles on Scopus
    1

    Present address: Center Leo Apostel for Interdisciplinary Studies, Free University of Brussels and the Department of Psychology, New York University, USA.

    View full text