Elsevier

Addictive Behaviors

Volume 33, Issue 10, October 2008, Pages 1314-1328
Addictive Behaviors

Implicit cognition and substance use: A meta-analysis

https://doi.org/10.1016/j.addbeh.2008.06.009Get rights and content

Abstract

A meta-analysis of 89 effect sizes based on the responses of 19,930 participants was conducted to estimate the magnitude of the relationship between substance-related implicit cognitions and the use of legal and illegal substances. The analysis produced a weighted average effect size of r = .31. Moderation analyses revealed significant heterogeneity in effect sizes related to facet of implicit cognition, measurement strategy, sample composition, and substance type. The largest effect sizes were found in studies that assessed implicit semantic associations, employed word association measures, and focused on marijuana use. The findings suggest that implicit cognition is a reliable predictor of substance use, although effect sizes vary as a function of several methodological factors.

Introduction

Considerable research has been conducted to identify the cognitive factors underlying decisions that lead to substance use and addiction. Until recently, much of this literature has been based on the premise that humans are rational decision makers who systematically evaluate possible positive and negative consequences of substance use, and make decisions using something similar to a cost-benefit analysis. Prominent theories based on this rational perspective include the health belief model (Becker, 1974), protection motivation theory (Rogers, 1983), the theory of reasoned action (Fishbein & Ajzen, 1975), and the theory of planned behavior (Ajzen, 1988).

More recent research has brought these traditional models into question — in particular the premise that decision-making is governed by conscious, rational processes. As Wiers and Stacy (2006a, p. 292) note “…the typical problem in addiction is not that drug abusers do not realise that the disadvantages of continued drug use outweigh the advantages. The central paradox in addictive behaviors is that people continue to use drugs even though they know the harm.”

Dual process models of cognition help resolve this apparent paradox by postulating that human behavior is guided by two distinct information-processing systems. Although several variations on the dual process theme have been proposed, most models distinguish between (1) an experiential system that is passive, effortless, rapid and intimately tied to intuition and affect, and (2) a rational-analytic system that is intentional, effortful, logic-based, and largely affect free. The rational system is under the conscious control of the individual and operates in a manner similar to that proposed by traditional decision models. The experiential system, on the other hand, functions automatically with little conscious input from the decision maker.

To the extent that output from the experiential system encourages substance use, and is not over-ruled by contrary advice from the rational system, dual process models predict that substance use will occur. For example, based on output from the rational system, an individual may conclude that the costs associated with continued substance use outweigh the benefits, and hence may develop a conscious intention to avoid using the substance. Output from the experiential system may undermine these cessation intentions by automatically, and perhaps pre-consciously, guiding the individual toward social situations and locations in which the substance is likely to be present.

Consistent with this view, many addiction researchers have drawn attention to the apparent irrationality of persisting with drug use despite the negative and sometimes devastating consequences. For example, Loewenstein (1996) suggested that drug use is perpetuated by uncontrolled, visceral urges, which override the user's rational knowledge of the negative consequences she or he will experience in the future. Bechara, Noel, and Crone's (2006) dual process model of addiction indicates that the impulsive present-oriented cognitive system often dominates the reflective, future-oriented system. Similarly, Yin and Knowlton (2006) argued that once drug use behavior is learned, the behavior is guided by automatic cognition, which bypasses explicit cognitive goals.

The distinction between rational and experiential information processing overlaps considerably with the concepts of explicit and implicit cognition. Explicit cognitions, like output from the rational system, are available to conscious introspection, and are typically assessed using self-report measures in which participants are asked to describe their attitudes, beliefs, or expectancies. In contrast, implicit cognitions, like the experiential system, are assumed to be automatic, and thus less available to conscious awareness. These cognitions are typically assessed using indirect measures involving reaction times, attentional bias, arousal, and memory associations.

Wiers and Stacy (2006b) note that there are several important benefits associated with assessing implicit cognition in the context of addiction. First, it enables researchers to move beyond traditional conceptualizations of judgment and decision making by assessing cognitive processes that are not available to introspection. Second, relative to explicit measures, implicit measures are less sensitive to social desirability effects, and thus may provide a more accurate measure of individuals' true attitudes or beliefs about substance use — particularly in situations where self-presentation biases are likely to be operating (Hofmann, Gawronski, Gschwendner, Huy, & Schmitt, 2005). Third, as noted earlier, implicit cognitions can help explain situations in which behavior appears to be at odds with consciously held beliefs and attitudes. Fourth, acknowledging the important distinction between implicit and explicit cognition helps bridge the gap between addiction research and other disciplines including social psychology (Epstein, 1994, Fazio, 1990), cognitive psychology (Slovic et al., 2005, Stanovich and West, 2000) and cognitive neuroscience (Bechara et al., 1997, Damasio, 1994, Lieberman, 2000) in which dual process models of cognition are already firmly established.

There are several distinct facets of implicit cognition that may influence substance use decisions and behaviors. These include: implicit attitudes, attentional bias, implicit arousal, and memory associations. In the sections that follow, we describe each of these facets and how they influence decisions and behavior. We also provide a summary of the measurement strategies that have been developed to assess each facet.

Implicit attitudes are spontaneous evaluative responses that have been shown to influence behavior in a variety of contexts (Perugini, 2005). A fundamental characteristic of implicit attitudes is that they involve an automatic and often pre-conscious evaluation of an attitude object. Positive evaluations of substances are assumed to increase the probability of approach behavior and use, whereas negative evaluations increase the likelihood that the substance will be avoided. Implicit approach tendencies may also reflect positive implicit evaluations (⁎Mogg et al., 2003, Waters and Sayette, 2006) or they may be involved in wanting in the absence of liking (see Houben, Wiers, & Roefs, 2006).

Implicit attitudes are frequently assessed using the Implicit Association Test (IAT; Greenwald, McGhee, & Schwartz, 1998). The IAT employs a dual categorization task to assess the relative associative strength of a target stimulus with evaluative extremes. Target and contrast stimuli (e.g., alcohol and soft drink) are presented on opposite sides of a computer screen. Typically, one stimulus appears with a positive attribute word (e.g., good) and the other with a negative attribute word (e.g., bad), with stimulus-attribute type pairings reversed for half of the presentations during the test. Participants are required to categorize stimulus or attribute exemplars appearing in the center of the screen as quickly as possible by pressing one key if the item belongs on the left side of the computer screen and another key if it belongs on the right. If two concepts are strongly associated in memory, response latencies will be shorter when the correct categorization response is congruent with that association. For example, individuals who prefer alcohol to soft drinks will classify positive attribute words more quickly when these words share a response key with alcohol words than when they share a response key with soft drink words.

The original version of the IAT has been criticized on several grounds, leading to the creation of several variants on the procedure. One important criticism is that the original IAT assumes a bipolar model of implicit attitudes, which implies that attitudes can be assessed on a single continuum ranging from positive to negative. However, several researchers (e.g., Jajodia & Earleywine, 2003) have argued that individuals frequently possess both positive and negative implicit associations with a target stimulus. This point is especially pertinent in the area of substance abuse because abusers often have ambivalent feelings toward the substance (Houben et al., 2006). Recent studies of implicit drug associations have addressed this problem by employing unipolar IATs that assess positive and negative attitudes separately (e.g., McCarthy & Thompson, 2006).

A second potential problem with the original IAT is that implicit attitudes can only be assessed in relation to a contrast category (Houben et al., 2006). This can be especially problematic in substance use studies in which the target stimulus (e.g., cocaine) does not have an obvious contrast category. This problem has led to the development of several new IAT-type measures that include a single target, but no contrast category. These include the single target IAT (stIAT; Wigboldus, Holland, & van Knippenberg, submitted for publication) and the Go/No-Go Association Task (GNAT; Nosek & Banaji, 2001).

Several other implicit attitude measures rely on response time. For example, the Extrinsic Affective Simon Task (EAST; De Houwer, 2003) requires participants to classify white adjectives based on their valence, and colored target and contrast stimulus words (e.g., words relating to alcohol and soft drink) based only on their color. Two response keys are used, one for positive words and words of one color, and the other for negative words and words of another color. The EAST assumes that performance is facilitated when items that are associated in memory share a response key. For example, when green words and negative words share a response key, an individual who dislikes alcohol will respond more quickly when alcohol words are green.

Another response time measure, affective priming, involves presenting a target stimulus followed by a word with an evaluative connotation. Participants must respond as quickly as possible by indicating whether the word is positive or negative. Evaluation of the target stimulus primes evaluative memory associations that can facilitate or inhibit subsequent responses, depending on whether the correct response has the same evaluation as the target stimulus. Thus, when individuals who possess positive attitudes toward smoking are primed with a smoking-related stimulus, they should respond more quickly to a subsequently presented positive attribute word than to a negative attribute word.

Palfai and colleagues (e.g., ⁎Ostafin et al., 2003, ⁎Palfai and Wood, 2001) have used an expectancy accessibility measure to assess the relative ease with which participants can access expectancies relating to substance use. Participants are presented with a series of outcome phrases such as less anxious that are preceded by a sentence stem relating to substance use such as alcohol makes me… A control sentence stem such as watching TV makes me… and related outcome phrases are also used. The relative speed with which participants can endorse substance-related outcomes provides an indication of the strength of these outcome expectancies within the memory network.

The stimulus–response compatibility task (SRC; De Houwer, Crombez, Baeyens, & Hermans, 2001) involves multiple presentations of drug-related and drug-neutral images on a computer screen. Participants are given two assignments in counter-balanced order. One assignment is to move a cursor toward the image appearing on the screen as quickly as possible if it is a drug-related image, and to move the cursor away from the image if the image is drug-neutral. In the other assignment instructions are reversed, i.e., the participant must move the cursor away from drug images and toward non-drug images. Studies employing the measure have shown that drug users perform faster than non-users on the approach drug images task, while non-users perform faster than users on the avoid drug images task (e.g., Field, Eastwood, Bradley, & Mogg, 2006).

As an alternative to response time measures, implicit attitudes toward substance use can be assessed using word association techniques. One example of this approach is Peters and Slovic's (1996)affect pool measure. Participants are asked to list the first thoughts or images that come to mind following the presentation of a word or phrase representing the stimulus of interest (e.g., drinking alcohol or smoking cigarettes). They are then required to rate each thought or image on a five-point scale with response options ranging from very negative to very positive.

Attentional bias represents a second facet of implicit cognition that may guide substance use behavior at a preconscious level (Cox, Fadardi, & Pothos, 2006). Due to natural predispositions and/or past learning experiences, individuals may be more or less likely to have their attentional focus automatically captured by drug cues in the environment. Once attention is captured, the cues are assumed to exert a greater influence on behavior.

The most frequently used measure of attentional bias is the Addiction-Stroop test. This test assesses the degree to which individuals are distracted by drug cues. It does this by measuring the speed and accuracy with which participants can name the colors of neutral versus drug-related words. The level of performance impairment in the presence of drug-related distractors provides an indicator of attentional bias for drug cues. A recent meta-analysis of the Addiction-Stroop test by Cox et al. (2006) found that studies employing substance-related manipulations, such as deprivation, produced larger effect sizes. This suggests implicit cognition may be tied to physiological aspects of substance use such as craving.

Visual focus localization measures represent a second common approach for assessing attentional bias in the substance-abuse literature. These measures include the dot probe task, visual probe task, and flicker paradigm. All three tasks are computer-based, and assess the speed with which participants the appearance of a dot (dot probe task), symbol (visual probe task), or change (flicker paradigm) presented in the location of a drug stimulus on the screen. The measures work by identifying the location of the participant's visual attention. If a participant detects more quickly a stimulus or change appearing in the location of a drug stimulus, relative to the location of a neutral stimulus, she or he has exhibited an attentional bias toward the drug stimulus. Studies employing visual focus localization methods have demonstrated that attentional bias for drug-related stimuli correlates with self-reported drug use (e.g., ⁎Mogg et al., 2003, ⁎Townshend and Duka, 2001, ⁎Yaxley and Zwaan, 2005).

Implicit arousal may involve automatically activated arousal-related cognitions in response to drug cues, or actual physiological arousal in response to drug cues. Implicit arousal to substance-related cues has been linked to heavier alcohol and cocaine use, and implicit sedation with lighter use of these substances (e.g., Dunn and Earleywine, 2001, Goldman et al., 1999, ⁎Wiers et al., 2007). The association between implicit arousal and substance use may be explained by incentive sensitization theory (Robinson & Berridge, 1993), which suggests that repeated drug use leads to an increased arousal response associated with sensitization of the mesolimbic dopamine system.

Most research on implicit arousal has employed one of two approaches. The first involves assessing arousal directly via psychophysiological measures, such as those assessing electrocardiographic and electrodermal activity, and galvanic skin response. In a meta-analysis of cue exposure and substance use, Carter and Tiffany (1999) found small to medium effect sizes for indicators of physiological arousal among smokers, alcoholics, opiate addicts and cocaine addicts.

Implicit arousal may also be measured indirectly by assessing automatic arousal-related cognition. This form of arousal assessment is often conducted using the arousal IAT. The test differs from the attitude IAT in that words associated with arousal and sedation are used instead words associated with positive and negative valence. As with the attitude IAT, unipolar (e.g., Houben & Wiers, 2006a) and single target (e.g., Thush & Wiers, 2007) versions of the arousal IAT have been used in substance use studies.

Extensive work by Szalay and colleagues established word association techniques within substance use research (Doherty and Szalay, 1996, Szalay et al., 1993). A word association technique developed and used primarily by Stacy and colleagues (e.g., ⁎Ames and Stacy, 1998, ⁎Leigh and Stacy, 1998, ⁎Stacy, 1995) has assessed the strength of drug-related interpretations of ambiguous cues potentially related to drugs and drug use outcomes. In Stacy's memory association measure, participants are instructed to free associate with ambiguous cue words, which could be interpreted as either substance-related or substance-neutral (e.g., draft). In addition, the measure typically includes an outcome/behavior association task, where participants are provided with a list of outcome phrases, some of which could be associated with drug use (e.g., feeling relaxed). Participants write the first behavior that comes to mind after reading each cue phrase. Heavier substance users consistently provide more substance-related associations in Stacy's test and related word association tests (e.g., Green & Galbraith, 1986). The number of drug associations a participant provides indicates the strength, and likely the influence, of the drug-related memory network (Stacy, 1997).

Some researchers have questioned whether word association measures should be considered implicit given that they involve the conscious retrieval of memory content. A contrary view is that these measures, when applied correctly, involve spontaneous responses that require no analytic deliberation (e.g., Nelson, McEvoy, & Dennis, 2000). Lieberman (2006) makes a similar point in a recent article in which he contrasts between the automaticity of free associations with the conscious effort involved in controlled responses. It is also worth noting that Stacy's word association approach, which contributes a substantial number of effect sizes to the current meta-analysis, was designed to ensure that respondents remain unaware of what is being assessed. This reduces the possibility of consciously distorted responding, an important potential problem with explicit measures of cognition. Thus, for the current review, we included word association measures, with the intention of assessing measurement approach for moderator effects.

It is clear that adolescents and adults behave differently from each other in many situations. Poor decision-making is commonly associated with adolescence. Krank and Goldstein (2006) suggested that implicit cognition may be of particular importance to adolescents, who tend to be more impulsive (Pechmann, Levine, Loughlin, & Leslie, 2005) and more inclined towards delay discounting of rewards than are adults (Green, Fry, & Myerson, 1994). Older adolescents have been found to engage in more analytic thinking than younger adolescents (Klaczynski, 2001). Additionally, a study by Halpern-Felsher and Cauffman (2001) showed that, when making decisions involving risk, the decision making processes of adults more closely reflected explicit decision making models than did those of adolescents in that adults tended to weigh the costs and benefits associated with the available options.

There is evidence that biology is partly responsible for the differences in adolescent and adult functioning. A vast amount of research implicates the frontal lobes in problem solving, rationality, and analytic cognition (e.g., Coolidge and Griego, 1995, Damasio, 1985, Goldberg and Podell, 2000, Mutsuo, 2004). Importantly, this region of the brain is also linked with impulse control (Hill, 2004). Adolescents' frontal lobes are in a process of maturation, which continues until the mid 20s (Geid, 2004). One study comparing adolescents with adults showed that adults rely on the frontal lobes for processing information about facial expressions, while adolescents rely on the amygdala, a region of the brain associated with impulse and emotion (Yurgelun-Todd, 1998).

It is plausible that implicit cognition plays a greater role in adolescent substance use than it does in adult substance use. The evidence discussed above suggests adolescents may rely more strongly on implicit cognition for making decisions than do adults. In addition to this, the nature of adolescent substance use suggests a pervasive use of non-rational judgment on the part of adolescents. Therefore, implicit cognition could be stronger predictor of substance use for adolescents than for adults. This possibility will be investigated as part of the meta-analysis.

Past reviews of implicit cognition and addiction have provided an indication of how well implicit cognition measures can predict substance use (e.g., Wiers & de Jong, 2006). However, these reviews have been qualitative in nature, and have not explicitly compared the predictive validity of different implicit cognition measures. When such comparisons have been made, they have been based on single studies (e.g., ⁎Sherman et al., 2003, ⁎Wiers et al., 2003). Although two previous meta-analyses relating to implicit cognition and addiction have been conducted (Carter and Tiffany, 1999, Cox et al., 2006), these were more narrowly focused than the present meta-analysis. Carter and Tiffany's study focused on physiological responses to substance-related cues, and Cox et al.'s (2006) review focused exclusively on studies employing the Addiction-Stroop test.

The present meta-analysis reviewed studies investigating the relationship between implicit cognition and substance use. Our primary objectives were to: (1) determine whether there is a reliable relationship between implicit cognition and substance use and quantify the magnitude of this relationship, and (2) determine whether four methodological factors – facet of implicit cognition, measurement approach, participant age, and substance type – operate as moderators.

The results of these analyses have potentially important theoretical, methodological and applied implications. From a theoretical perspective, they will provide insights into whether all facets of implicit cognition are equally robust predictors of substance use behavior, and the specific boundary conditions under which each cognition type predicts or fails to predict substance use. From a methodological perspective, the results will aid future researchers and practitioners by identifying implicit measures that exhibit the best concurrent and predictive validity. Finally, from an applied perspective, the analysis will provide valuable information to practitioners about how to develop more effective interventions to prevent and treat substance use problems. For example, if the analysis revealed that attentional bias measures of implicit cognition are much stronger predictors of substance use than semantic memory association measures, or vice versa, this would provide useful advice about which implicit system to target. Similarly, if the results suggest that implicit cognition is a stronger predictor of substance use for adolescents, but not adults, this would imply that different types of interventions may be required for these two groups.

Section snippets

Literature search

We searched the PsycINFO and PubMed databases for two categories of keywords/phrases. The first category related to measures of implicit cognition. Search terms included: implicit, priming, IAT, word association, free association, Extrinsic Affective Simon Task, Go/No-Go Association Task, attentional bias, flicker paradigm, dot probe, visual probe, dual task, artificial grammar-learning, expectancy accessibility, and indirect measure(s). The second set of search terms related to substance use,

Results

Table 1 summarizes the key characteristics of studies included in the meta-analysis. As shown in Table 2, the overall weighted r was .31. According to Cohen's (1988) recommendations, the correlations for small, medium, and large effect sizes are .10, .30, and .50, respectively. Thus, the overall effect size was medium by these standards. The fail-safe N statistic is used to establish a level of confidence that a significant meta-analytic effect size is not the result of the “file draw” problem (

Discussion

This meta-analysis investigated the relationship between implicit cognition and substance use using 89 effect sizes. The overall weighted average correlation was .31, equating to an overlap of 9.6% between the two constructs. Thus, implicit cognition was reliably associated with substance use, and this association was moderate in magnitude.

We hypothesized that the relationship between implicit associations with substance use and self-reported substance use would be stronger for studies

Conclusions

This meta-analysis showed an overall significant relationship between substance-related implicit cognition and self-reported substance use. The findings reinforce the importance of considering the role of implicit cognition in the initiation and maintenance of drug use, and of developing drug use interventions specifically targeting affect and various types of automatic cognitive responses.

References (142)

  • GadonL. et al.

    Negative alcohol consumption outcome associations in young and mature adult social drinkers: A route to drinking restraint?

    Addictive Behaviors

    (2004)
  • Halpern-FelsherB.L. et al.

    Costs and benefits of a decision. Decision making competence in adolescents and adults

    Journal of Applied Developmental Psychology

    (2001)
  • HillA.B. et al.

    Alcohol dependence and semantic priming of alcohol related words

    Personality and Individual Differences

    (1992)
  • HillE.L.

    Evaluating the theory of executive dysfunction in autism

    Developmental Review

    (2004)
  • HoubenK. et al.

    Assessing implicit alcohol associations with the implicit association test: Fact or artifact?

    Addictive Behaviors

    (2006)
  • HuijdingJ. et al.

    Automatic associations with the sensory aspects of smoking: Positive in habitual smokers but negative in non-smokers

    Addictive Behaviors

    (2006)
  • HuijdingJ. et al.

    Implicit and explicit attitudes toward smoking in a smoking and a nonsmoking setting

    Addictive Behaviors

    (2005)
  • LoewensteinG.

    Out of control: Visceral influences on behavior

    Organizational Behavior and Human Decision Processes

    (1996)
  • PalfaiT.P. et al.

    Alcohol-related motivational tendencies in hazardous drinkers: Assessing implicit response tendencies using the modified-IAT

    Behaviour Research and Therapy

    (2003)
  • PothosE.M. et al.

    Cognitive bias for alcohol-related information in inferential processes

    Drug and Alcohol Dependence

    (2002)
  • AjzenI.

    Attitudes, personality, and behavior

    (1988)
  • AmesS.L. et al.

    Implicit cognition in the prediction of substance use among drug offenders

    Psychology of Addictive Behaviors

    (1998)
  • AmesS.L. et al.

    Implicit cognition and dissociative experiences as predictors of adolescent substance use

    The American Journal of Drug and Alcohol Abuse

    (2005)
  • AmesS.L. et al.

    Comparison of indirect assessments of association as predictors of marijuana use among at-risk adolescents

    Experimental and Clinical Psychopharmacology

    (2007)
  • BassettJ.F. et al.

    A portable version of the go/no-go association task (GNAT)

    Behavior Research Methods

    (2005)
  • BecharaA. et al.

    Deciding advantageously before knowing the advantageous strategy

    Science

    (1997)
  • BecharaA. et al.

    Loss of willpower: Abnormal neural mechanisms of impulse control and decision making in addiction

  • BeckerM.H.

    The health belief model and personal health behavior

    (1974)
  • BradleyB.P. et al.

    Attentional and evaluative biases for smoking cues in nicotine dependence: Component processes of biases in visual orienting

    Behavioral Pharmacology

    (2004)
  • CarterB.L. et al.

    Meta-analysis of cue-reactivity in addiction research

    Addiction

    (1999)
  • ChassinL. et al.

    Parental smoking cessation and adolescent smoking

    Journal of Pediatric Psychology

    (2002)
  • CohenJ.

    Statistical power analysis for the behavioral sciences

    (1988)
  • CoolidgeF.L. et al.

    Executive functions of the frontal lobes: Psychometric properties and a self-rating scale

    Psychological Reports

    (1995)
  • CoxW.M. et al.

    The Addiction-Stroop test: Theoretical considerations and procedural recommendations

    Psychological Bulletin

    (2006)
  • CummingS. et al.

    Semantic priming of expectancies among high- and low-restraint non-problem drinkers

    Australian Journal of Psychology

    (2001)
  • DamasioA.R.

    Descartes' error

    (1994)
  • DamasioA.R.

    The frontal lobes

  • De HouwerJ.

    The extrinsic affective Simon task

    Experimental Psychology

    (2003)
  • De HouwerJ.

    What are implicit measures and why are we using them?

  • De Houwer, J., & Custers, R. (2004). Implicit approach-avoidance and valence associations with smoking and alcohol....
  • De HouwerJ. et al.

    On the generality of the affective Simon effect

    Cognition and Emotion

    (2001)
  • De HouwerJ. et al.

    Do smokers have a negative implicit attitude toward smoking?

    Cognition and Emotion

    (2006)
  • De JongP. et al.

    Relevance of research on experimental psychopathology to substance misuse

  • DeutschR. et al.

    Reflective and impulsive determinants of addictive behavior

  • DohertyK.T. et al.

    Statistical risk versus psychological vulnerability: Why are men at greater risk for substance abuse than women?

    Journal of Alcohol and Drug Education

    (1996)
  • DunnM.E. et al.

    Activation of alcohol expectancies in memory in relation to limb of the blood alcohol curve

    Psychology of Addictive Behaviors

    (2001)
  • EpsteinS.

    Integration of the cognitive psychodynamic and unconscious

    American Psychologist

    (1994)
  • FallonB.M.

    To smoke or not to smoke: The role of schematic information processing

    Cognitive Therapy and Research

    (1998)
  • Feldman-BarrettL. et al.

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

    Psychological Bulletin

    (2004)
  • FieldM. et al.

    Craving and cognitive biases for alcohol cues in social drinkers

    Alcohol and Alcoholism

    (2005)
  • Cited by (220)

    • Implicit beliefs and automatic associations in smoking

      2024, Journal of Behavior Therapy and Experimental Psychiatry
    View all citing articles on Scopus

    References marked with an asterisk indicate studies included in the meta-analysis.

    View full text