Direct Effects of Binge Drinker Stereotypes on Drinking ... - PsycNET

Health Psychology 2013, Vol. 32, No. 5, 571–580

© 2013 American Psychological Association 0278-6133/13/$12.00 http://dx.doi.org/10.1037/a0029859

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.

Automatic Risk Behavior: Direct Effects of Binge Drinker Stereotypes on Drinking Behavior Amanda Rivis

Paschal Sheeran

University of Nottingham

University of Sheffield

Objective: We investigated whether research on stereotype priming effects can increase understanding of binge drinking among young people. Four studies tested whether evaluations of the binge drinker stereotype have a direct and automatic effect on binge drinking behavior, and whether a self-regulation intervention can overcome stereotype effects on drinking behavior. Methods: Studies 1a and 1b were prospective surveys conducted over two weeks. Study 2 was a field experiment with a 1-week behavioral follow-up. Study 3 was an exploratory RCT that tested whether implementation intentions can override binge drinker stereotype effects on behavior over one month. Participants were college students (Studies 1a, 1b, and 2; Ns ⫽ 226, 72, and 255, respectively) and school-age youth (Study 3, N ⫽ 202). Results: Correlational (Studies 1a and 1b) and experimental (Study 2) evidence indicated that favorable binge drinker stereotypes had direct effects on binge drinking among experienced drinkers. These effects were automatic in the sense that participants were neither aware of, nor intended, this influence. Study 3 showed that implementation intentions geared at increasing self-focused attention overcame stereotype effects on binge drinking behavior. Conclusion: These findings suggest that stereotype priming research can offer valuable insights into binge drinking behavior. Implications for health behavior theories and behavior change interventions are outlined. Keywords: priming, stereotypes, implementation intentions, binge drinking, alcohol

pants to walk more slowly leaving the laboratory (Bargh, Chen, & Burrows, 1996, Experiment 2). Priming effects are automatic in the sense that participants were not aware of, and did not intend, the impact of the prime on their behavior. In the present research, we investigate whether stereotype priming can aid understanding of a key health-risk behavior— binge drinking among young people. To date, only four studies—that all concerned snack consumption in the immediate context—appear to have examined priming effects on health behaviors (Albarracin, Leeper, & Wang, 2009; Harris, Bargh, & Brownell, 2009; Papies & Hamstra, 2010). One reason for the paucity of studies may be a concern about the durability of behavior priming effects. In most priming studies, stereotypes or goals are activated (in a supposed “first” study) and immediately thereafter participants have the opportunity to act on the primed construct (in a putative “second, unrelated” study). That is, priming studies generally assess behavioral effects over very short durations (a few minutes) and thus may seem to offer little traction on the risky behaviors of interest to health psychologists that are performed repeatedly over extended periods (e.g., alcohol consumption). However, two lines of research offer clues about how priming effects might be studied in field settings and over longer follow-up periods than have been examined heretofore. First, Cesario, Plak, and Higgins’ (2006) analysis of stereotype priming effects—the Motivated Preparation Account (MPA)—indicates that stereotype evaluation, or attitude toward the relevant social category, is crucial in determining the behavioral impact of stereotype activation. For instance, Cesario et al. (2006) found that if participants had favorable attitudes toward elderly people, then priming the

Understanding the determinants of health-risk behavior is a key goal for health psychologists. To this end, several health behavior theories have been developed such as the health belief model, protection motivation theory, the theory of planned behavior, and social– cognitive theory. These models all share the assumption that conscious thoughts (e.g., beliefs, attitudes, and intentions) are the key determinants of behavior. More recently, dual-process models of health behavior have drawn attention to implicit (nonconscious) processes such as attentional bias and implicit attitudes that are also important for understanding health actions (reviews by Friese, Hofmann & Wiers, 2011; Hofmann, Friese, & Wiers, 2008; Wiers et al., 2010). However, the most influential social psychological research on nonconscious processes—work on behavior priming— has, so far, made few inroads in health psychology (Sheeran, Gollwitzer, & Bargh, in press). Numerous studies have shown that unobtrusively activating (i.e., priming) traits, stereotypes, or goals leads to behavior in line with the primed construct (for reviews, see, e.g., Aarts, 2007; Bargh, 2006). For instance, priming the stereotype of elderly people causes partici-

Amanda Rivis, Institute of Work, Health, & Organisations, University of Nottingham, Nottingham, United Kingdom; Paschal Sheeran, Department of Psychology, University of Sheffield, Sheffield, United Kingdom. This research was supported by a grant from the ESRC (RES-000-220260). We thank Mathew Blackman for collecting the data for Study 3, and Tom Webb and Paul Norman for feedback on the paper. Correspondence concerning this article should be addressed to Paschal Sheeran, Department of Psychology, University of Sheffield, Sheffield S10 2TP, United Kingdom. E-mail: [email protected] 571


572

RIVIS AND SHEERAN

category “elderly” reduced walking speed (similar to Bargh et al., 1996). If, however, participants negatively evaluated elderly people, then elderly priming actually increased walking speed. According to the MPA, these findings come about because stored evaluations of social groups provide useful information about how to interact with members of those groups—“one slows down to interact effectively with elderly people” but “one speeds up to avoid interacting with the stereotypically slow disliked outgroup member” (Cesario et al., 2006, p. 899). That is, activating the social category causes people to engage in motivated preparation to interact with a member of that category that, in turn, dictates the nature of the automatic behavior (fast vs. slow walking). The nature of automatic behavior also depends upon experience, however. Dijksterhuis, Aarts, Bargh, and van Knippenberg (2000) showed that elderly priming only affected behavior when participants had experience with people who are elderly. Participants with little experience did not have strong associations between the category “elderly” and stereotypic traits (e.g., slow), so priming did not engender automatic behavior. These two lines of research lead to the following prediction in relation to binge drinking: Positive evaluations of the binge drinker stereotype will automatically increase binge drinking behavior among participants who have experience of binge drinking. The MPA indicates that evaluations of binge drinkers dictate the impact of encountering a member of that category on individuals’ own drinking behavior. As the present studies all concerned participants’ naturalistic alcohol consumption in group settings (e.g., bars, clubs), it was safe to assume that participants would encounter “binge drinkers”. Such encounters should prime the binge drinker stereotype in situ for our participants, and so lead to greater drinking among participants who like the category. Thus, the present research focuses on the automatic effects of stereotype evaluation (and not merely on stereotype activation) as evaluation determines the behavioral impact of the activated stereotype. The impact of evaluations of binge drinkers can be deemed “automatic” if liking the stereotype engenders greater binge drinking behavior (a) independent of participants’ drinking intentions and other conscious beliefs (i.e., directly), and (b) even when participants do not believe that the stereotype influences their behavior (cf. Bargh & Chartrand, 2000). Evaluations of the binge drinker stereotype should have greater impact on binge drinking behavior among participants who have experience of binge drinking because these are the participants who possess strong associations between (mental representations of) the stereotype and the behavior (cf. Dijksterhuis et al., 2000).

Rationale for the Research and Overview of Studies Binge drinking behavior is defined as consuming at least one half of one’s recommended weekly intake of alcohol in a single session, and is associated with numerous serious health problems (Centers for Disease Control and Prevention; CDCP, 2011). Young people aged 18 to 20 years comprise the largest group of binge drinkers in the US, and about 90% of all alcohol consumed by this age group takes the form of binge drinking (CDCP, 2011). Young people in the United Kingdom engage in more binge drinking compared to other EU countries (Anderson et al., 2009), and 54% percent of 15- and 16-year olds in the United Kingdom binge drink as compared to 22% in the US. (Hibell et al., 2004).

Binge drinking is therefore an important context in which to test the behavioral impact of stereotype evaluations, and to examine whether stereotype influence can be overcome. Four studies were undertaken. Studies 1a and 1b tested whether stereotype evaluations directly predicted binge drinking behavior even (a) after controlling for cognitions about binge drinking and past behavior, and (b) when participants did not believe that stereotypes influenced their behavior. Study 2 manipulated the favorability of the binge drinker stereotype and investigated whether this intervention automatically influenced binge drinking. Finally, Study 3 tested whether a self-regulation intervention can overcome the impact of favorable evaluations of the binge drinker stereotype on participants’ binge drinking.

Studies 1a and 1b: Do Evaluations of Drinker Stereotypes Predict Binge Drinking Behavior? Studies 1a and 1b were prospective surveys. The baseline questionnaires contained measures of the variables specified by Ajzen’s (1991) theory of planned behavior (TPB) and the prototype/willingness model (PWM; Gibbons, Gerrard, Blanton, & Russell, 1998). In fact, the PWM specifies stereotype evaluation and similarity to the stereotype as important predictors of healthrisk behaviors (such stereotypes are termed prototypes in the model). However, according to the PWM, the influence of stereotype evaluation and stereotype similarity should be mediated by behavioral willingness, that is, participants’ inclination to act in a risky fashion given conducive circumstances (Gibbons, Gerrard, & Lane, 2003). The TPB, on the other hand, suggests that any effects of stereotypes should be mediated by behavioral intention (Ajzen, 1991). The present studies envisage a direct and automatic effect of evaluations of the binge drinker stereotype on behavior (not mediated by willingness or intention) for experienced drinkers. Thus, the experience ⫻ stereotype evaluation interaction should predict behavior after TPB and PWM variables are taken into account, even for participants who do not believe that drinker stereotypes influence their behavior. In particular, more favorable evaluation of the binge drinker stereotype should be associated with greater binge drinking among experienced drinkers.

Method Participants and Procedure Undergraduates at the University of Sheffield (Study 1a) and University of Derby (Study 1b) volunteered to take part in a survey concerning their “views about binge drinking”. At Time 1, participants (N ⫽ 333 and 91, respectively) completed measures of TPB variables, past behavior, willingness, descriptive norm, and their evaluations of, and similarity to, the binge drinker stereotype. Two weeks later, participants reported their binge drinking behavior and completed a measure of awareness of stereotype influence. Response rates were 68% (N ⫽ 226) and 79% (N ⫽ 71) for Studies 1a and 1b, respectively. Representative checks revealed no differences on any of the Time 1 variables between participants who completed the first wave of data collection only and participants who completed both waves, in either study.

AUTOMATIC RISK BEHAVIOR


Questionnaire Measures TPB and PWM variables were measured with respect to “engaging in a binge drinking session at least twice in the next two weeks.” A binge drinking session was defined as “drinking at least one half of one’s recommended weekly intake of alcohol at one session.” Unless otherwise stated, responses to all items were on 7-point bipolar scales. Attitude was measured by responses to the stem, “For me, engaging in a binge drinking session would be. . .” on six bipolar scales (pleasant–unpleasant, wise–foolish, healthy– unhealthy, beneficial–harmful, enjoyable–unenjoyable, exciting– boring; ⬀ ⫽ .81 and 80 for Studies 1a and 1b, respectively). Intention was measured by three items, for example, “I intend to engage in a binge drinking session at least twice in the next two weeks” (definitely do–definitely don’t; ⬀ ⫽ .92 and .91). Subjective norm was assessed by two items (e.g., “Most people who are important to me think I should engage in a binge drinking session at least twice in the next two weeks”; likely–unlikely; ⬀ ⫽ .93 and .92). Factor analysis indicated that perceived behavioral control comprised two factors (Trafimow, Sheeran, Conner, & Finlay, 2002); two items measured perceived difficulty (e.g., “I feel confident that I will be able to engage in a binge drinking session at least twice in the next two weeks”; strongly agree–strongly disagree) and proved reliable (⬀ ⫽ .64; there were three perceived difficulty items in Study 1b, ⬀ ⫽ .92). The two perceived control items (e.g., “I feel in complete control of whether or not I engage in a binge drinking session at least twice in the next two weeks,” strongly agree–strongly disagree) were reliable in Study 1a (⬀ ⫽ .78) but not Study 1b, and were treated as separate items. Descriptive norm was measured by two items (e.g., “Of the people you know, how many engage in a binge drinking session at least twice a week?”; 5-point scale, none to all; ⬀ ⫽ .78 and .92). Willingness was assessed by 3 items in Study 1a and by 8 items in Study 1b (e.g., “Suppose you were at a friend’s house party. One of your friends suggests that you all play a ‘drinking game’ which involves ‘downing’ as many shorts as you can in one minute. The person who drinks the greatest number of shorts in the time limit is the winner. How likely is it that you would . . . take part in the game?”; not at all likely–very likely; ⬀ ⫽ .62 and .86). Stereotype evaluation was assessed by two items. The first item was the “evaluation thermometer” (e.g., Haddock & Zanna, 1994). Participants were asked to indicate their overall evaluation of the type of person who engages in a binge drinking session at least once a week on a 1 to 100 scale. The second item asked “How likable or dislikeable do you think the type of person who engages in a binge drinking session at least once a week is?” (very likable– very dislikeable). Scores were standardized and proved reliable (⬀ ⫽ .79 and .71). Stereotype similarity was assessed by two items (e.g., “In general, how similar are you to the type of person who engages in a binge drinking session at least once a week?”; not at all–very) (⬀ ⫽ .82 and .87). Experience or past behavior was measured by the number of times they had engaged in a binge drinking session in the past month in Study 1a, and by the item, “How often have you engaged in a binge drinking session in the last month?” (5-point scale; never–every day) in Study 1b. Binge drinking behavior at Time 2 was measured by participants’ reports of the number of times they had engaged in a binge drinking session during the previous two weeks (Study 1a). In Study 1b, behavior was measured by two items (a) self-reports of

573

the number of times participants had engaged in a binge drinking session in the past two weeks, and (b) “How often have you engaged in a binge drinking session in the past two weeks?” (5-point scale; never–every day). These items proved reliable (⬀ ⫽ .91), and were standardized and averaged to form a scale. Awareness of stereotype influence was measured by the item, “Did your image or opinion of the typical binge drinker influence your own binge drinking in the past two weeks?” (4-point scale; definitely did not–definitely did) in Study 1b.

Results and Discussion Study 1a Hierarchical linear regression was used to analyze the data. Behavior was regressed on TPB and PWM variables (step 1) and the experience ⫻ stereotype evaluation interaction (step 2). Variables were mean-centered to reduce potential multicollinearity prior to computing the interaction term (Aiken & West, 1991). Stereotype evaluation and past behavior were the only significant predictors of behavior at the first step (Table 1, left panel). The experience ⫻ stereotype evaluation interaction term was significant (␤ ⫽ .17, p ⬍ .05) and contributed a significant increment in the variance at the second step (⌬R2 ⫽ .01, ⌬F ⫽ 7.92, p ⬍ .01). Decomposition of the interaction via simple slopes analyses indicated that when participants had little experience of binge drinking (M ⫺ 1 SD for past behavior), then stereotype evaluation did not predict behavior (B ⫽ .35, ns). However, among experienced binge drinkers (M ⫹ 1 SD), stereotype evaluation was a highly significant predictor of subsequent behavior (B ⫽ .68, p ⬍ .001). The fact that positive evaluations of the typical binge drinking predicted behavior over and above the effects of intentions, willingness, and other conscious considerations is consistent with idea that stereotype evaluation effects are automatic.1

Study 1b The awareness item in Study 1b enabled us to examine whether or not participants believe that the binge drinker stereotype influences their binge drinking, and to test whether the experience ⫻ stereotype evaluation interaction still predicts behavior even among participants who disavow influence by the stereotype. Almost one half of participants believed that prototypes “definitely did not” influence their behavior (47%, n ⫽ 34), and a further 29% (n ⫽ 21) thought prototypes “probably did not” influence them. 1 Evidence that people are not aware of the influence of stereotype evaluation on their behavior (unlike, for instance, intentions and willingness from the TPB and PWM) would offer more convincing support for automatic influence. To test this idea, we adapted a paradigm developed by Gilbert, Pinel, Wilson, Blumberg, and Wheatley (1998). Participants equivalent to those who took part in Studies 1a and 1b (N ⫽ 55) received definitions of the constructs and sample items from the study questionnaires and rated “how big an influence would each of the factors described above have on your binge drinking over a two week period” (7-point scale; has no influence–has a very big influence). Findings showed that willingness, intention, PBC, attitude, and past behavior were all perceived as influential (i.e., means were significantly larger than the scale midpoint; Ms ⫽ 5.55, 5.27, 4.98, 4.69, and 4.43, respectively, all ts ⬎ 2.08, ps ⬍ .05), and were seen as significantly more influential than stereotype evaluation (M ⫽ 3.79), all ts ⬎ 2.44, ps ⬍ .02.

RIVIS AND SHEERAN

574

Table 1 Regressions of Behavior on TPB and PWM Variables, the Past Behavior–Prototype Evaluation Interaction Term (Studies 1a and 1b) Study 1a Model


Step 1

Step 2 R2 Model F a ⴱ

Study 1b

Variable

Beta

Beta

Beta

Beta

Intention Perceived Ease Perceived Control Perceived Controla Attitude Subjective Norm Willingness Prototype Evaluation Prototype Similarity Past Behavior Descriptive Norms Past ⫻ Evaluation

.16 .05 .03

.17 .08 .01

.02 ⫺.04 ⫺.06 .15ⴱ ⫺.01 .47ⴱⴱⴱ .08 — .53 23.76ⴱⴱⴱ

.03 ⫺.05 ⫺.00 .10 .00 .36ⴱⴱⴱ .11 .17ⴱⴱ .54 23.11ⴱⴱⴱ

⫺.13 ⫺.02 ⫺.04 .04 .17 .07 .25 .18 ⫺.04 .56ⴱⴱ ⫺.09 — .59 5.55ⴱⴱⴱ

⫺.00 ⫺.05 .01 .01 .12 .10 .20 .27ⴱ ⫺.10 .45ⴱⴱ ⫺.03 .28ⴱⴱ .64 6.43ⴱⴱⴱ

The perceived control items did not scale in Study 1b and were therefore treated as separate items. p ⬍ .05. ⴱⴱ p ⬍ .01. ⴱⴱⴱ p ⬍ .001.

Seventeen participants (21%) thought that the binge drinker stereotype “probably did” influence their behavior whereas only 2 participants (3%) reported that the prototype “definitely did” influence their behavior. To test the key hypothesis, participants who believed prototypes “probably” or “definitely” affected their behavior (n ⫽ 19) were eliminated from subsequent analyses. (Findings were the same when these participants were included.) Again, a two-step hierarchical linear regression was used to test whether the experience ⫻ stereotype evaluation interaction predicted behavior after controlling for the other variables (see Table 1, right panel). Findings showed that past behavior was the only significant predictor of behavior at Step 1. Entry of the experience ⫻ stereotype evaluation interaction term on the second step produced a significant increment in the variance explained in behavior (⌬R2 ⫽ .05, ⌬F ⫽ 7.21, p ⬍ .01). Past behavior, stereotype evaluation, and their interaction were the only significant predictors of behavior in the final equation. Simple slopes analyses indicated that stereotype evaluation did not predict behavior when experience was low (M ⫺ 1 SD; B ⫽ ⫺.03, ns). However, at high levels of experience (M ⫹1 SD), stereotype evaluation strongly predicted behavior (B ⫽ .37, p ⬍ .004; see Figure 1). Thus, the findings indicate that for experienced drinkers, positive evaluations of the typical binge drinker are directly related to binge drinking behavior. This effect emerges even when (a) participants’ conscious intentions are taken into account (Studies 1a and 1b), and (b) participants do not believe that drinker stereotypes influence their behavior (Study 1b). The behavioral impact of evaluation of the binge drinker stereotype can therefore be deemed automatic.

behavior, and to provide further evidence that the stereotype evaluation-behavior effect occurs outside of participants’ awareness. Study 2 therefore manipulated the favorability of the binge drinker stereotype by allocating different numeric values to the 11-point scales used to evaluate the stereotype: a ⫺5 to ⫹5 (bipolar) scale and a 0 to 10 (unipolar) scale. Research on life satisfaction by Schwarz, Knauper, Hippler, Noelle-Neumann, and Clark (1991) showed that the bipolar scale engendered higher life satisfaction scores compared to the unipolar scale. This is because the ⫺5 to ⫹5 scale leads participants to believe that the dimension being investigated is bipolar (failure vs. success) whereas the 0 to 10 scale leads participants to believe that the dimension is unipolar (less vs. more success). People are likely reluctant to characterize their life as a failure and so rate their life satisfaction as higher in the ⫺5 to ⫹5 condition.

Study 2: Does Changing Evaluations of the Binge Drinker Stereotype Change Behavior? Although the correlational findings from Studies 1a and 1b support the idea that drinker stereotypes automatically influence behavior, an experimental test is needed to confirm that evaluations of the binge drinker stereotype have a causal impact on

Figure 1. Simple slopes for stereotype evaluation in predicting binge drinking over two weeks by experience (Study 1b).


We predicted that a manipulation that reduced the favorability of binge drinker stereotypes would result in less binge drinking among experienced drinkers. In particular, participants in the unipolar (0 to 10) condition should have less positive evaluations of binge drinkers and subsequently engage in less binge drinking behavior compared to participants in the bipolar (⫺5 to ⫹5) condition.

Method


Participants and Procedure Participants were 379 undergraduates at the University of Sheffield who voluntarily completed three questionnaires about binge drinking. At Time 1, participants completed a TPB questionnaire that also contained a stereotype evaluation manipulation. One week later (Time 2), 255 participants reported their binge drinking behavior, for a response rate of 67%. Representativeness checks showed no difference on Time 1 questionnaire variables between participants who completed both waves of data collection and participants who completed the first questionnaire only. Two weeks after distributing the Time 2 questionnaire (Time 3), participants were asked to indicate how much they thought their binge drinking behavior had been influenced by their evaluations of the binge drinker stereotype. There were 58 responses, for a response rate of 23%. There were no differences on Time 1 variables between participants who completed the Time 1 questionnaire only and participants who completed both the Time 1 and Time 3 questionnaires. Stereotype evaluation manipulation. Participants were asked to evaluate the “type of person your age who engages in a binge drinking session at least once a week” on four dimensions: outgoing, fun-loving, cheerful, and friendly. These dimensions were the most frequently reported positive attributes of the “typical binge drinker” in a study of 333 participants (Rivis, 2002). Participants were randomly assigned to either the unipolar or the bipolar stereotype evaluation conditions. In both conditions, the verbal endpoints of the rating scales were “not at all” and “extremely.” In the bipolar condition, the numeric endpoints associated with these verbal labels were ⫺5 and ⫹5, whereas in the unipolar condition they were 0 and 10. Cognition and behavior measures. After completing the stereotype evaluation items, participants in both conditions reported their intention to binge drink (“Do you intend to engage in a binge drinking session at least once in the next week?”; definitely no– definitely yes), and the perceived ease of binge drinking (“How easy or difficult will it be for you to engage in a binge drinking session at least once in the next week?”; very difficult–very easy). Responses were on 11-point scales; numeric values were not provided for responses to these items. Experience or past behavior was measured by participants’ reports of the number of times they had engaged in a binge drinking session during the previous two weeks. Two questions were used to measure behavior at Time 2: “How many times have you engaged in a binge drinking session in the last week?” and “How often did you engage in a binge drinking session in the last week?” (5-point scale; not at all–every day; ⬀ ⫽ .91). At Time 3, awareness of stereotype influence was measured using the item described in Study 1 (modified to the relevant time period).

575 Results and Discussion

Findings for the Full Sample (N ⴝ 255) Manipulation check. The four items used to measure prototype evaluation were reliable (␣ ⫽ .84) and were averaged to form a scale. A t test was used to assess whether the numeric values manipulation affected evaluations of the “typical binge drinker.” As anticipated, findings showed that the prototype was rated less favorably (M ⫽ 6.72, SD ⫽ 1.46) in the unipolar (0 to 10) condition as compared to the bipolar (⫺5 to ⫹5) condition (M ⫽ 7.46, SD ⫽ 1.48), t ⫽ 4.86, p ⬍ .001. There was no difference between the conditions on intentions, perceived ease, or past behavior, Fs ⬍ 1. Effects on behavior. Hierarchical linear regression analysis was used to examine whether the interaction between past behavior and the stereotype evaluation manipulation affected future drinking behavior, over and above the TPB variables and past behavior. Intention, perceived ease, past behavior, and condition (unipolar scale ⫽ 0, bipolar scale ⫽ 1) were entered on the first step and the interaction term was added on the second step. Scores were mean-centered prior to computing the interaction term. Table 2 (left panel) shows that, as predicted, the interaction between condition and past behavior contributed a significant increment in the variance in the final step of the equation (⌬R2 ⫽ .01, ⌬F ⫽ 6.70, p ⬍ .01). The interaction term, intention, and past behavior accounted for 52% of the variance in behavior. Simple slopes analyses showed that the manipulation significantly affected future binge drinking behavior at high levels of past behavior (M ⫹1 SD; B ⫽ .57, p ⬍ .01) but not at low levels of past behavior (M – 1 SD; B ⫽ ⫺.30, ns; see Figure 2). Reducing the favorability of evaluations of the typical binge drinker resulted in significantly less binge drinking among experienced drinkers.

Findings for the Unaware Sample (N ⴝ 52) Participants generally did not believe that their evaluation of the stereotype had influenced their binge drinking behavior. None of the participants reported that the stereotype “definitely did” influence their behavior, and only 6 participants (10%) thought that the stereotype “probably did” influence their behavior. (These participants were excluded from subsequent analyses.) Thirty-four participants (59%) reported that stereotypes “probably did not” influence their drinking behavior and the remainder (n ⫽ 18, 31%) indicated that the stereotype “definitely did not” influence their drinking behavior. Thus, the “unaware” sample comprised 52 participants. Manipulation check. The binge drinker stereotype was again rated less favorably (M ⫽ 6.31, SD ⫽ 1.16) in the 0 to 10 condition compared to the ⫺5 to ⫹5 condition (M ⫽ 7.37, SD ⫽ 1.11), t(50) ⫽ 3.39, p ⬍ .001. There were no differences between conditions on intentions or perceived ease, Fs(1, 50) ⬍ 1. However, participants in the unipolar condition reported binge drinking less often in the past than did participants in the bipolar condition (Ms ⫽ 1.73 and 2.92, SDs ⫽ 1.28 and 2.43, respectively), F(1, 50) ⫽ 4.89, p ⬍ .05; past behavior was therefore controlled in subsequent analyses. Effects on behavior. Hierarchical linear regression analysis again showed that the interaction between past behavior and con-

RIVIS AND SHEERAN

576

Table 2 Hierarchical Regression of Behavior on TPB Variables, Past Behavior, Condition, and the Condition by Past Behavior Interaction Term for the Full and Unaware Samples (Study 2) Full sample (N ⫽ 255) Model

Variables entered


Step 1

Intention Perceived ease Past behavior Condition Condition ⫻ Past Behavior

Step 2 R2 Model F ⴱ

p ⬍ .05.

ⴱⴱ

p ⬍ .01.

ⴱⴱⴱ

Beta ⴱⴱⴱ

.27 .03 .50ⴱⴱⴱ .04 — .51 66.49ⴱⴱⴱ

Unaware sample (N ⫽ 52)

Beta ⴱⴱⴱ

.27 .02 .39ⴱⴱⴱ .04 .16ⴱⴱ .52 55.75ⴱⴱⴱ

Beta ⴱ

.27 .04 .59ⴱⴱⴱ ⫺.03 — .67 23.62ⴱⴱⴱ

Beta .36ⴱⴱ .05 .10 .14 .46ⴱⴱⴱ .71 22.35ⴱⴱⴱ

p ⬍ .001.

dition contributed a significant increment in the variance at the second step (⌬R2 ⫽ .04, ⌬F ⫽ 6.41, p ⬍ .02). Simple slopes analyses showed that the manipulation significantly affected future binge drinking behavior at both high (B ⫽ 2.89, p ⬍ .02) and low levels of drinking experience (B ⫽ 1.68, p ⬍ .05). However, the significant interaction indicates that altering participants’ evaluations of drinker stereotypes engendered a significantly greater reduction in behavior among more experienced binge drinkers compared to less experienced binge drinkers. Moreover, the effect of the stereotype evaluation manipulation was automatic as participants’ conscious intentions were taken into account and participants were not aware that the binge drinker stereotype was influencing their behavior.

Study 3: Can a Self-Regulation Intervention Undermine the Impact of Binge Drinker Stereotypes on Drinking Behavior? Studies 1a, 1b, and 2 suggest that binge drinker stereotypes have direct and automatic effects on binge drinking behavior. These findings beg the question, how can young people self-regulate stereotype priming effects? One promising self-regulation inter-

Figure 2. Simple slopes for condition in predicting binge drinking over one week by experience (Study 2).

vention is forming implementation intentions (Gollwitzer & Sheeran, 2006; Gollwitzer, 1999). Implementation intentions are if–then plans that specify when and how one will act toward a goal. For example, “If x happens, then I’ll do y (to reach my goal)!”. Neurological evidence indicates that implementation intentions are effective because action control is switched from top-down control by goals to bottom-up control by specified cues (Gilbert, Gollwitzer, Cohen, Oettingen, & Burgess, 2009). That is, behavior comes under the cues that were specified in one’s plan rather than goals that are activated by the immediate context. Consistent with this idea, Gollwitzer, Sheeran, Trötschel, and Webb (2011) showed that planning out a behavior in an if-then format protected its performance from unwanted priming effects (e.g., planning one’s driving behavior at critical junctures prevented the characteristic increases in driving speed and errors goal that accrued from priming the goal of being fast; Study 3). Whereas Gollwitzer et al. (2011) focused on planning out the focal behavior, the implementation intention in the present study was geared at alleviating the influence of the binge drinker stereotype. To this end, the “if” part of the plan specified a critical juncture (when participants’ alcohol consumption had reached the threshold for a binge). We drew on two lines of research in designing the “then” part of the plan. First, the plan should prevent participants from mentally elaborating upon any urge to drink— such elaboration consumes working memory resources and increases cravings for alcohol (Achtziger, Gollwitzer, & Sheeran, 2008; Kavanagh, Andrade, & May, 2005). Second, the plan should enlist a process that is known to disrupt behavior priming effects. Self-focused attention (e.g., Duval & Wicklund, 1972), that is, metacognitive awareness of what one is doing, is one such process. For instance, Dijksterhuis and van Knippenberg (2000) showed that stereotyping priming no longer influenced behavior when participants were seated in front of mirror during the experiment— shifting participants’ focus of attention to their current functioning was sufficient to override the automatic behavior. The implementation intention attempted to instill self-focused attention by having participants take a third-person perspective. Evidence indicates that taking a third-person perspective brings people’s attention to bear on what they are doing and engenders greater dispositional, and less situational, control of behavior (e.g., Libby, Shaeffer, Eibach, & Slemmer, 2007). We predicted than an implementation intention that specified a critical situation (“If I am


at the limit for a binge . . .”), and an effective means of acting in that situation (“. . . then I ignore the urge to drink and will look at the situation as if I were someone else!”) would reduce the impact of favorable binge drinker stereotypes on subsequent binge drinking behavior.

Method


Participants and Procedure All sixth form school pupils (N ⫽ 326) at a school in the east of England were invited to complete a questionnaire at Time 1 that measured TPB and PWM variables, past behavior, and evaluations of the binge drinker stereotype. One month later (Time 2), 202 participants (55.4% female; M age ⫽ 16.62, SD ⫽ 0.68) reported their binge drinking behavior, for a response rate of 62%. Comparison of the questionnaire responses for participants who were retained at both waves and individuals who took part in the first wave only showed no significant differences on any of the variables (ps ⬎ .21) except similarity to the stereotype. Participants who left the study showed marginally stronger identification with binge drinkers compared to participants who were retained (Ms ⫽ 2.96 and 2.68, SDs ⫽ 1.46 and 1.40, respectively), F(1, 324) ⫽ 3.01, p ⫽ .084. The study adopted a 2 (Implementation intention: formed vs. not formed) ⫻ 2 (Stereotype evaluation: low vs. high) factorial design. Implementation intention was manipulated and stereotype evaluation was measured (groups were generated using a median split, Mdn ⫽ 3.50). Participants were randomly assigned to implementation intention conditions. The questionnaire began by defining binge drinking. Unless otherwise stated, all items were measured using 7-point scales. Attitude was measured using six items from the same stem— “Engaging in a binge-drinking session, at least four times in the next month, would be . . .”; (unpleasant–pleasant, wise–foolish, healthy–unhealthy, beneficial–harmful, exciting–boring, enjoyable–not enjoyable; ⬀ ⫽ 0.82). Three items were used to measure intention (e.g., “I intend to engage in a binge-drinking session at least four times in the next month”; definitely yes– definitely no; ⬀ ⫽ 0.90). PBC was measured with three items (e.g., “I feel confident that I will be able to engage in a binge-drinking session four times in the next month”; strongly agree–strongly disagree; ⬀ ⫽ 0.81). Subjective norm was assessed by a single term, “Most people who are important to me think that I should engage in a binge-drinking session at least four times in the next month.” (unlikely–likely). The descriptive norm scale (⬀ ⫽ 0.67) comprised two items (e.g., “How many of your friends at school engage in a binge-drinking session at least once a week?”). Willingness was measured using three items (e.g., “If the circumstances were very attractive, how willing would you be to engage in a binge-drinking session at least four times in the next month?”; extremely willing–extremely unwilling; ⬀ ⫽ 0.78). Stereotype evaluation was measured by two items (e.g., “How likable or dislikeable do you think the type of person who engages in a binge-drinking session is?”; very likable–very dislikeable; ⬀ ⫽ 0.67). Similarity to the drinker stereotype was measured by two items (e.g., “Do the characteristics that you think describe a binge-drinker of your age describe you?”; definitely no–definitely yes; ⬀ ⫽ 0.74). Past behavior was measured by two items (e.g.,

577

how often participants engaged in a binge-drinking session; 5-point scale, never–very often; ⬀ ⫽ 0.82). The behavior measure at Time 2 was participants’ reports of the number of binge drinking sessions they had engaged in since completing the first questionnaire.

Implementation Intention Induction The implementation intention induction came at the end of the Time 1 questionnaire for relevant participants. Participants were told “Sometimes people do things they don’t want to do when they are binge drinking. To help you do what you want in these circumstances, please read the statement below three times and commit yourself to acting on it: If I am at the limit for a binge, then I ignore the urge to drink and will look at the situation as if I were someone else!” The questionnaire for the control condition was identical except for the omission of this induction.

Results and Discussion Randomization Check A MANOVA was conducted on the cognition measures and past behavior to check randomization to implementation intention conditions. The multivariate test was not significant, F(9, 192) ⫽ 0.77, ns. However, the univariate test for PBC was significant, F(1, 200) ⫽ 4.91, p ⬍ .05. Participants who did not form implementation intentions had higher PBC scores than participants who formed plans (Ms ⫽ 4.25 vs. 3.69, SDs ⫽ 1.82 and 1.76). As this difference favors the control group and covarying PBC in subsequent analyses did not alter the key findings, randomization could be considered largely successful.

Moderation of the Stereotype Evaluation-Binge Drinking Relation by Implementation Intentions A 2 (Implementation intention) ⫻ 2 (Stereotype evaluation) ANOVA was used to assess effects on binge drinking at Time 2 (see Figure 3). The main effect of implementation intention was

Figure 3. Binge drinking at one-month follow-up by stereotype evaluation and implementation intention (Study 3).

RIVIS AND SHEERAN


578

significant, F(1, 198) ⫽ 4.94, p ⬍ .03. Participants who formed an if–then plan engaged in less binge drinking (M ⫽ 1.31, SD ⫽ 1.78) compared to controls (M ⫽ 1.70, SD ⫽ 2.15). Stereotype evaluation also had a significant main effect, F(1, 198) ⫽ 26.45, p ⬍ .001. Participants who evaluated the binge drinker stereotype more positively (M ⫽ 2.65, SD ⫽ 2.56) engaged in more binge drinking than did participants with less positive evaluations (M ⫽ 1.08, SD ⫽ 1.51). The predicted implementation intention ⫻ stereotype evaluation interaction also proved significant, F(1, 198) ⫽ 7.99, p ⬍ .01. Simple effects analyses indicated that, for participants who did not plan, holding a positive drinker stereotype was associated with engaging in more than three times as many binge drinking sessions as compared to holding a less positive stereotype (Ms ⫽ 3.30 and 0.99, respectively), F(1, 198) ⫽ 33.64, p ⬍ .001. However, when participants had formed an implementation intention, stereotype evaluation no longer influenced subsequent binge drinking (Ms ⫽ 1.16 and 1.86), F(1, 198) ⫽ 2.54, p ⬎ .11. Thus, forming an implementation intention geared at instilling self-focused attention at the point when participants were at the threshold for a drinking binge was effective in reducing the impact of favorable binge drinking stereotypes on binge drinking behavior.2

General Discussion The goal of the present research was to discover whether social psychological research on stereotype priming (e.g., Bargh et al., 1996; Cesario et al., 2006) could help to understand an important health-risk behavior— binge drinking among young people. Findings from four studies suggested that this was the case. Two prospective surveys (Studies 1a and 1b) found that favorable evaluations of the binge drinker stereotype were directly related to binge drinking behavior among experienced drinkers. A field experiment that reduced the favorability of evaluations of the binge drinker stereotype (Study 2) obtained the same findings. In Studies 1b and 2, binge drinker stereotypes could be deemed to have influenced drinking behavior automatically— because the behavioral impact of the stereotype bypassed participants’ explicit cognitions (e.g., intentions) and their conviction that the stereotype had not influenced their behavior. Finally, an exploratory RCT among school-age youth (Study 3) found that a self-regulation intervention geared at heightening participants’ self-focused attention could undermine the link between binge drinker stereotypes and subsequent binge drinking behavior. These findings suggest that, for young people’s binge drinking at least, implicit or automatic routes to behavior may warrant greater consideration alongside the more intentional or reflective route as specified in traditional health behavior theories (Sheeran et al., in press). In addition, the findings speak to the particular utility of research on behavior priming for understanding health actions—a consideration that has not so far been advocated by dual-process models of health behavior (Friese et al., 2011; Hofmann et al., 2008; Wiers et al., 2010). The present research drew upon the motivated preparation account (MPA) of stereotype priming effects (Cesario et al., 2006) which points to critical role of evaluations of the stereotype in determining the nature of the automatic behavior in the wake of category activation: Liking the social category engenders a response that is geared at promoting a smooth interaction with a

category member whereas disliking generates the opposite response. Studies 1a, 1b, and 2 assumed that binge drinker stereotypes would be activated in drinking situations, and thus focused on the automatic effects of stereotype evaluation rather than stereotype activation. This shift in focus—from category priming to category liking—is important as it becomes possible to explore the automatic effects of a wide variety of health-risk (and healthprotective) stereotypes on behavior in field settings, over extended periods, and deploy methods traditionally used in health psychology research (e.g., surveys, randomized experiments using manipulations embedded in questionnaires). Our findings showing that health-risk stereotypes have direct and automatic effects on behavior also have implications for the prototype/willingness model (PWM; Gibbons et al., 1998, 2003). The PWM pioneered research on the role of stereotypes (or prototypes) in predicting young people’s risk behavior, and has proved valuable in drawing attention to behavioral willingness as a less intentional route to action. Whereas the PWM assumes that the influence of stereotypes on behavior is mediated by willingness, the present findings indicate that there is another route— one that is direct and automatic. These findings suggest that the PWM might usefully be extended to reflect the consideration that behavioral willingness may not always mediate the stereotype– behavior relation. As with any new program of research, the present studies have several limitations that should be acknowledged. First, the samples sizes for unaware participants were relatively small in Studies 1b and 2. Although the key findings were replicated among larger samples (Ns ⫽ 226 and 255), further tests using larger samples of participants who disavow influence by stereotypes would be valuable. Second, behavioral follow-ups were relatively short (one, two, two, and four weeks in Studies 2, 1a, 1b, and 4, respectively); further tests over longer periods are warranted. One potential consequence of short follow-up periods is a floor effect for frequency of binge drinking behavior. Although the mean frequency of binge drinking was approximately 1 session even for less experienced drinkers in the present studies, studies over longer periods will permit tests using a greater range of behavior scores. It is also the case that the implementation intention intervention in Study 3 was compared to a passive control condition, and so tests using active control groups or alternative interventions would be valuable. Finally, the present research drew upon the MPA to explain the effects of binge drinker evaluations. There are several other theoretical models of behavior priming effects such as the Indirect Prime-to-Behavior Model (Smeesters, Wheeler, & Kay, 2010) and the Situated Inference Model (Loersch & Payne, 2011) that could plausibly account for the present findings, however. We acknowledge that the present studies were not designed to test these competing theories. 2 We undertook several additional analyses to double check these results. Findings showed that the critical implementation intention ⫻ stereotype evaluation interaction remained significant even when (a) similarity to the stereotype (where there was a marginal difference in the representativeness check) and PBC (where there was a significant difference in the randomization check) were covaried in the ANOVA, (b) stereotype evaluation was treated as a continuous variable and linear regression was used to analyze the data, and (c) the behavior measure was dichotomized (0 ⫽ no binge drinking, 1 ⫽ any binge drinking) and logistic regression was used to analyze the data.



Another apparent limitation of the present research is that stereotype evaluation effects were observed only among experienced drinkers. Although the present findings do not speak to inexperienced drinkers, the identification of a potentially modifiable determinant of experienced drinkers’ behavior—stereotype evaluation—is important. Evidence indicates that interventions based on traditional health behavior theories (e.g., that aim to change behavioral intentions) are most effective for samples that have less experience with the behavior, and are much less effective among experienced samples where habits have formed (e.g., Webb & Sheeran, 2006). Thus, for less experienced drinkers, there are a variety of intervention approaches based on well-established theories that should prove effective. For experienced drinkers, however, targeting cognitions about drinking may not be sufficient because these cognitions do not predict their drinking behavior. For this group, interventions that target binge drinker stereotypes by (a) preventing exposure to positive stereotypes, (b) modifying the valence of stereotypes, or (c) facilitating effective selfregulation of stereotype influence should prove worthwhile.

References Aarts, H. (2007). Health behavior and the implicit motivation and regulation of goals. Health Psychology Review, 1, 53– 82. Achtziger, A., Gollwitzer, P. M., & Sheeran, P. (2008). Implementation intentions and shielding goal striving from unwanted thoughts and feelings. Personality and Social Psychology Bulletin, 34, 381–393. Aiken, L. S., & West, R. R. (1991). Multiple regression: Testing and interpreting interactions. Newbury Park, CA: Sage. Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50, 179 –211. Albarracin, D., Leeper, J., & Wang, W. (2009). Immediate increases in eating after exercise promotion messages. Obesity, 17, 1451–1452. Anderson, P., de Bruijn, K., Angus, K., Gordon, R., & Hastings, G. (2009). Impact of alcohol advertising and media exposure on adolescent alcohol use: A systematic review of longitudinal studies. Alcohol & Alcoholism, 44, 229 –243. Bargh, J. A. (Ed.) (2006). Social psychology and the unconscious: The automaticity of higher mental processes. New York, NY: Psychology Press. Bargh, J. A., & Chartrand, T. L. (2000). The mind in the middle: A practical guide to priming and automaticity research. In H. T. Reis & C. M. Judd (Eds.), Handbook of research methods in social and personality psychology. New York, NY: Cambridge University Press. Bargh, J. A., Chen, M., & Burrows, L. (1996). Automaticity of social behavior: Direct effects of trait construct and stereotype activation on action. Journal of Personality and Social Psychology, 71, 230 –244. Centers for Disease Control and Prevention. (2011). Fact sheets: Binge drinking. Retrieved from http://www.cdc.gov/alcohol/fact-sheets/bingedrinking.htm Cesario, J., Plaks, J. E., & Higgins, E. T. (2006). Automatic social behavior as motivated preparation to act. Journal of Personality and Social Psychology, 90, 893–910. Dijksterhuis, A., Aarts, H., Bargh, J. A., & van Knippenberg, A. (2000). On the relation between associative strength and automatic behavior. Journal of Experimental Social Psychology, 36, 531–544. Dijksterhuis, A., & van Knippenberg, A. (2000). Behavioral indecision: Effects of self-focus on automatic behavior. Social Cognition, 18, 55–74. Duval, S., & Wicklund, R. A. (1972). A theory of objective self-awareness. London, United Kingdom: Academic Press. Friese, M., Hofmann, W., & Wiers, R. W. (2011). On taming horses and strengthening riders: Recent developments in research on interventions

579

to improve self-control in health behaviors. Self and Identity, 10, 336 – 351. Gibbons, F. X., Gerrard, M., Blanton, H., & Russell, D. W. (1998). Reasoned action and social reaction: Willingness and intention as independent predictors of health risk. Journal of Personality and Social Psychology, 74, 1164 –1180. Gibbons, F. X., Gerrard, M., & Lane, D. J. (2003). A social-reaction model of adolescent health risk. In J. J. Suls & K. A. Wallston (Eds.), Social psychological foundations of health and illness (pp. 107–136). Oxford, United Kingdom: Blackwell. Gilbert, D. T., Pinel, E. C., Wilson, T. D., Blumberg, S. J., & Wheatley, T. P. (1998). Immune neglect: A source of durability bias in affective forecasting. Journal of Personality and Social Psychology, 75, 617– 638. doi:10.1037/0022-3514.75.3.617 Gilbert, S., Gollwitzer, P. M., Cohen, A-L., Oettingen, G., & Burgess, P. W. (2009). Separable brain systems supporting cued versus selfinitiated realization of delayed intentions. Journal of Experimental Psychology: Learning, Memory, and Cognition, 35, 905–915. Gollwitzer, P. M. (1999). Implementation intentions: Strong effects of simple plans. American Psychologist, 54, 493–503. Gollwitzer, P. M., & Sheeran, P. (2006). Implementation intentions and goal achievement: A meta-analysis of effects and processes. Advances in Experimental Social Psychology, 38, 69 –119. Gollwitzer, P. M., Sheeran, P., Trötschel, R., & Webb, T. L. (2011). Self-regulation of behavior priming effects. Psychological Science, 22, 901–907. Haddock, G., & Zanna, M. P. (1994). Preferring “housewives” to “feminists”: Categorization and the favorability of attitudes towards women. Psychology of Women Quarterly, 18, 25–52. Harris, J. L., Bargh, J. A., & Brownell, K. D. (2009). Priming effects of television food advertising on eating behavior. Health Psychology, 28, 404 – 413. Hibell, B., Andersson, B., Bjarnason, T., Ahlstrom, S., Balakireva, O., Kokkevi, A., & Morgan, M. (2004). The ESPAD report 2003: Alcohol and other drug use among students in 35 European countries. Stockholm, Sweden: Swedish Council for Information on Alcohol and Other Drugs. Hofmann, W., Friese, M., & Wiers, R. W. (2008). Impulsive versus reflective influences on health behavior: A theoretical framework and empirical review. Health Psychology Review, 2, 111–137. Kavanagh, D. J., Andrade, J., & May, J. (2005). Imaginary relish and exquisite torture: The Elaborated Intrusion theory of desire, Psychological Review, 112, 446 – 467. Libby, L. K., Shaeffer, E. M., Eibach, R. P., & Slemmer, J. A. (2007). Picture yourself at the polls: Visual perspective in mental imagery affects self-perception and behavior. Psychological Science, 18, 199 – 203. Loersch, C., & Payne, B. K. (2011). The situated inference model: An integrative account of the effects of primes on perception, behavior, and motivation. Perspectives on Psychological Science, 6, 234 –252. Papies, E. K., & Hamstra, P. (2010). Goal priming and eating behavior: Enhancing self-regulation by environmental cues. Health Psychology, 29, 384 –388. Rivis, A. (2002). The role of social influences in young people’s healthrelated behaviors: Descriptive norms and social images (Unpublished doctoral thesis). University of Sheffield, United Kingdom. Schwarz, N., Knauper, B., Hippler, H. J., Noelle-Neumann, E., & Clark, L. (1991). Rating scales: Numeric scales may change the meaning of scale labels. International Journal of Public Opinion Research, 7, 72–74. Sheeran, P., Gollwitzer, P. M., & Bargh, J. A. (in press). Nonconscious processes and health. Health Psychology. Smeesters, D. H. R. V., Wheeler, S. C., & Kay, A. C. (2010). Indirect prime-to-behavior effects: The role of perceptions of the self, others, and

580

RIVIS AND SHEERAN


situations in connecting primed constructs to social behavior. Advances in Experimental Social Psychology, 42, 259 –317. Trafimow, D., Sheeran, P., Conner, M., & Finlay, K. A. (2002). Is perceived behavioural control a multidimensional construct? Perceived difficulty and perceived control. British Journal of Social Psychology, 41, 101–122. Webb, T. L., & Sheeran, P. (2006). Does changing behavioral intentions engender behavior change? A meta-analysis of the experimental evidence. Psychological Bulletin, 132, 249 –268. Wiers, R. W., Houben, K., Roefs, A., De Jong, P., Hofmann, W., & Stacy, A. W. (2010). Implicit cognition in health psychology: Why common

sense goes out of the window. In B. Gawronski & K. Payne (Eds.), Handbook of implicit social cognition (pp. 463– 488). New York, NY: Guilford Press.

Received May 31, 2011 Revision received February 3, 2012 Accepted February 10, 2012 䡲