10
Health Psychology 2001, Vol. 20, No. 6, 424-433 Copyright 2001 by the American Psychological Association, Inc. 0278-6133/01/S5.00 DOI: 10.1037//0278-6133.20.6.424 Defensive Evaluation of Antismoking Messages Among College-Age Smokers: The Role of Possible Selves Mark A. Freeman, Emily V. Hennessy, and Denise M. Marzullo Drew University This study hypothesized that individuals respond to antisraoking messages in a biased or defensive manner to the degree that smoking is a personally relevant activity for them. The authors operationalized the personal relevance of smoking variously as smoking behavior (smoking status, rate, duration, and recent attempts to quit), endorsement of the smoker stereotype, and importance of smoking behavior as an identity within the self-concept (current self and possible selves). In the experiment, smokers (n = 82) and nonsmokers (n = 105) privately viewed several antismoking video segments. Smoking status, current smoking identity, and long-term future smoking identity were significantly associated with a defensive evaluation of antismoking messages. The study concludes that the concept of possible selves (H. Markus & P. Nurius, 1986) is critical in understanding college-age smoking and in the design of effective antismoking campaigns. Key words: tobacco smoking, self-concept, antismoking campaigns, health behavior, attitude change, possible selves Smoking rates among high school youth have been stable or have even declined slightly in the past few years since peaking during the middle of the 1990s (Johnston, O'Malley, & Bachman, 2000). However, a recent trend study among students (N > 14,000) at 116 four-year U.S. colleges found that smoking rates increased sharply among this group overall (27.8%) from 1993 to 1997, and 28% of college smokers reported that they began smoking regularly after reaching college age (Wechsler, Rigotti, Gledhill-Hoyt, & Lee, 1998; see also Johnston, O'Malley, & Bachman, 1999). Thus, although most long-term smoking be- havior does begin in adolescence (U.S. Department of Health and Human Services [DHHS], 1994), to prevent occasional adolescent smokers from becoming regular college-age smokers and to en- courage new smokers to quit before their habit becomes long term, the college-age population represents an important target for anti- smoking interventions (Wechsler et al., 1998). The 1998 settlement between 46 state Attorneys General and several tobacco companies yielded billions of dollars .for use in antismoking media campaigns, so research on the effectiveness of such campaigns is especially timely. To date, two kinds of data support the conclusion that antismoking campaigns are at least moderately effective: qualitative, focus-group research used in the development of antismoking media (Goldman & Glantz, 1998; McKenna & Williams, 1993); and outcome data showing reduc- Maric A. Freeman, Emily V. Hennessy, and Denise M. Marzullo, De- partment of Psychology, Drew University. This study was funded by a research grant from Drew University. We thank Joseph M. Kinast for his contribution to the initial phase of this project. Correspondence concerning this article should be addressed to Mark A. Freeman, Department of Psychology, Drew University, 36 Madison Ave- nue, Madison, New Jersey 07940. Electronic mail may be sent to [email protected]. tions in tobacco consumption in states that aired antismoking messages extensively (Pechmann, 1997; Worden et al., 1996). Antismoking media campaigns as a whole thus appear to have positive effects, but the dearth of theoretically informed empirical research exploring individual differences in responses to antismok- ing media (Pechmann, 1997) is a significant lacuna in the general understanding of these differences. Antismoking media are pre- mised on a wide variety of different persuasive strategies, and it is critical that media planners understand how well and with what target groups (nonsmokers, heavy smokers, occasional smokers, long-term smokers, etc.) specific antismoking messages are most effective. The present study begins to address this issue by closely investigating how one important individual-difference variable— level of personal involvement with smoking—mediates individual responses to specific antismoking messages (Chaiken, Liberman, & Eagly, 1989; Johnson & Eagly, 1989; Liberman & Chaiken, 1992; Petty & Cacioppo, 1979). It is generally hypothesized that individuals respond to anti- smoking messages in a biased or defensive manner and rate the messages as less effective to the degree that smoking is a person- ally relevant activity for them. We have operationalized personal involvement with smoking in a variety of ways, using several different measures of smoking behavior and of smoking-related attitudes. In particular, we argue here that having a smoker self- identity—both in the present (Falomir & Invernizzi, 1999; Shadel & Mermelstein, 1996; Stryker, 1987) but especially as an expected "possible self' in the future (Markus & Nurius, 1986)—is a par- ticularly potent expression of personal involvement with smoking in that it is strongly associated with a defensive response to antismoking messages. Personal Relevance and Attitude Change Ever since the adverse health consequences of cigarette smoking became more widely known, cognitive dissonance theory (Fest- 424

Defensive evaluation of antismoking messages among college-age smokers: The role of possible selves

Embed Size (px)

Citation preview

Page 1: Defensive evaluation of antismoking messages among college-age smokers: The role of possible selves

Health Psychology2001, Vol. 20, No. 6, 424-433

Copyright 2001 by the American Psychological Association, Inc.0278-6133/01/S5.00 DOI: 10.1037//0278-6133.20.6.424

Defensive Evaluation of Antismoking Messages Among College-AgeSmokers: The Role of Possible Selves

Mark A. Freeman, Emily V. Hennessy, and Denise M. MarzulloDrew University

This study hypothesized that individuals respond to antisraoking messages in a biased or defensivemanner to the degree that smoking is a personally relevant activity for them. The authors operationalizedthe personal relevance of smoking variously as smoking behavior (smoking status, rate, duration, andrecent attempts to quit), endorsement of the smoker stereotype, and importance of smoking behavior asan identity within the self-concept (current self and possible selves). In the experiment, smokers (n = 82)and nonsmokers (n = 105) privately viewed several antismoking video segments. Smoking status,current smoking identity, and long-term future smoking identity were significantly associated with adefensive evaluation of antismoking messages. The study concludes that the concept of possible selves(H. Markus & P. Nurius, 1986) is critical in understanding college-age smoking and in the design ofeffective antismoking campaigns.

Key words: tobacco smoking, self-concept, antismoking campaigns, health behavior, attitude change,possible selves

Smoking rates among high school youth have been stable orhave even declined slightly in the past few years since peakingduring the middle of the 1990s (Johnston, O'Malley, & Bachman,2000). However, a recent trend study among students(N > 14,000) at 116 four-year U.S. colleges found that smokingrates increased sharply among this group overall (27.8%) from1993 to 1997, and 28% of college smokers reported that theybegan smoking regularly after reaching college age (Wechsler,Rigotti, Gledhill-Hoyt, & Lee, 1998; see also Johnston, O'Malley,& Bachman, 1999). Thus, although most long-term smoking be-havior does begin in adolescence (U.S. Department of Health andHuman Services [DHHS], 1994), to prevent occasional adolescentsmokers from becoming regular college-age smokers and to en-courage new smokers to quit before their habit becomes long term,the college-age population represents an important target for anti-smoking interventions (Wechsler et al., 1998).

The 1998 settlement between 46 state Attorneys General andseveral tobacco companies yielded billions of dollars .for use inantismoking media campaigns, so research on the effectiveness ofsuch campaigns is especially timely. To date, two kinds of datasupport the conclusion that antismoking campaigns are at leastmoderately effective: qualitative, focus-group research used in thedevelopment of antismoking media (Goldman & Glantz, 1998;McKenna & Williams, 1993); and outcome data showing reduc-

Maric A. Freeman, Emily V. Hennessy, and Denise M. Marzullo, De-partment of Psychology, Drew University.

This study was funded by a research grant from Drew University. Wethank Joseph M. Kinast for his contribution to the initial phase of thisproject.

Correspondence concerning this article should be addressed to Mark A.Freeman, Department of Psychology, Drew University, 36 Madison Ave-nue, Madison, New Jersey 07940. Electronic mail may be sent [email protected].

tions in tobacco consumption in states that aired antismokingmessages extensively (Pechmann, 1997; Worden et al., 1996).

Antismoking media campaigns as a whole thus appear to havepositive effects, but the dearth of theoretically informed empiricalresearch exploring individual differences in responses to antismok-ing media (Pechmann, 1997) is a significant lacuna in the generalunderstanding of these differences. Antismoking media are pre-mised on a wide variety of different persuasive strategies, and it iscritical that media planners understand how well and with whattarget groups (nonsmokers, heavy smokers, occasional smokers,long-term smokers, etc.) specific antismoking messages are mosteffective. The present study begins to address this issue by closelyinvestigating how one important individual-difference variable—level of personal involvement with smoking—mediates individualresponses to specific antismoking messages (Chaiken, Liberman,& Eagly, 1989; Johnson & Eagly, 1989; Liberman & Chaiken,1992; Petty & Cacioppo, 1979).

It is generally hypothesized that individuals respond to anti-smoking messages in a biased or defensive manner and rate themessages as less effective to the degree that smoking is a person-ally relevant activity for them. We have operationalized personalinvolvement with smoking in a variety of ways, using severaldifferent measures of smoking behavior and of smoking-relatedattitudes. In particular, we argue here that having a smoker self-identity—both in the present (Falomir & Invernizzi, 1999; Shadel& Mermelstein, 1996; Stryker, 1987) but especially as an expected"possible self' in the future (Markus & Nurius, 1986)—is a par-ticularly potent expression of personal involvement with smokingin that it is strongly associated with a defensive response toantismoking messages.

Personal Relevance and Attitude Change

Ever since the adverse health consequences of cigarette smokingbecame more widely known, cognitive dissonance theory (Fest-

424

Page 2: Defensive evaluation of antismoking messages among college-age smokers: The role of possible selves

SMOKING AND POSSIBLE SELVES 425

inger, 1957) has provided psychologists with a useful frameworkfor understanding why people continue to smoke in the face ofsuch knowledge. Dissonance theory proposes that people strive tomaintain consistency between all cognitive and behavioral ele-ments in their consciousness and that inconsistencies produce astate of psychological discomfort known as dissonance. Individu-als experiencing dissonance are motivated to reduce it by changingone of the implicated cognitive or behavioral elements. Festinger(1957) suggested that because behavioral change is often difficult,cognitive change, which may involve avoiding, misperceiving,invalidating, or forgetting dissonance-arousing information, ismore likely. Applied to the present study, then, dissonance theorypredicts that smokers involuntarily exposed to dissonance-arousing antismoking messages will be motivated to misperceive,invalidate, or otherwise distort such messages.

Festinger (1957) also highlighted the role of the importance ofthe dissonant cognitive elements to the self, positing that, "If twoelements are dissonant with one another, the magnitude of thedissonance will be a function of the importance of the elements"(p. 16). Following this line of argument, antismoking messagesshould arouse more cognitive dissonance—and consequently moremotivation to misperceive, invalidate, or otherwise distort suchmessages—in a smoker for whom smoking-related attitudes andbehaviors are rated as important or personally relevant as com-pared with a smoker for whom smoking is less personally relevant.

This theoretical assertion, that the importance of a cognition orbehavior produces greater levels of dissonance, naturally begs thequestion of how this concept is defined. Since Festinger's (1957)original insight, this definitional issue has been dealt with mostextensively in the literature on persuasion and attitude change,which has conceptualized personal importance in a variety of waysunder the broad terms involvement or ego involvement (for re-views, see Johnson & Eagly, 1989; Levin, Nichols, & Johnson,2000). Involvement has been variously defined as the integration ofan attitude into the values that define the self (C. W. Sherif, Sherif,& Nebergall, 1965), as the degree to which attitude endorsementproduces desired consequences for the self (Petty & Cacioppo,1979), as membership in groups known to endorse particularattitudinal positions (M. Sherif & Hovland, 1961), or as the degreeto which public endorsement of an attitude raises self-presentational concerns (Cooper & Fazio, 1984). Because thesedifferent varieties of involvement are known to have somewhatdifferent effects on attitude change (Johnson & Eagly, 1989; Levinet al., 2000), the present study incorporates several different op-erational definitions of the personal relevance of or involvement inthis behavior.

Defensive Processing of Threatening Messages

Irrespective of the effects of personal involvement on attitudechange per se, most research supports the conclusion that personalinvolvement (regardless of how it is defined) increases systematicand objective processing of persuasive messages (Chaiken, 1987;Petty & Cacioppo, 1979). However, of particular relevance to thepresent study, and more in line with Festinger's (1957) originaltheory, Chaiken and colleagues (Chaiken, 1987; Chaiken et al.,1989) posit that when persuasive messages are perceived as threat-ening to self-concept, personal involvement can prompt individu-als to engage in more biased, defensive processing.

For example, Falomir and Invernizzi (1999) found that, forweakly identified smokers (smokers for whom smoking was nothighly relevant to their self-concept), exposure to an antismokingmessage increased intention to quit smoking relative to a controlgroup not exposed to this message; for highly identified smokers,however, such exposure did not affect their intention to quitsmoking. Falomir and Invernizzi suggested that increased personalrelevance of smoking caused highly identified smokers to experi-ence the antismoking message as threatening and thus encouragedthem to defend their smoking identity by lowering their estimatesof their intention to quit. Weakly identified smokers, on the otherhand, were not threatened and so were more willing to acknowl-edge that their attitudes concerning smoking cessation were influ-enced by the antismoking message. In further support of thisinterpretation, the authors found that highly identified smokersexposed to the antismoking message also increased their estimatesof peer support and approval for their habit relative to the controlgroup, "to justify their lower rather than raised intention to quitsmoking" (p. 81). They concluded that strongly identified smokersseem to be thus "immunized" against persuasive attempts to getthem to quit (p. 81).

Analogously, Liberman and Chaiken (1992) found that heavycoffee drinkers were more critical than people who did not drinkcoffee of a study outlining the links between caffeine consumptionand fibrocystic disease. Liberman and Chaiken argued that "withnonthreatening messages, increased personal relevance will oftenincrease motivation to arrive at an accurate conclusion... but witha threatening message, increased personal relevance may insteadincrease motivation to arrive at or defend a preferred conclusion orreject an undesirable one" (p. 669). Other studies have similarlyfound that with threatening messages—whether the threat is in-creasing college tuition (Lundgren & Prislin, 1998), raising thedrinking age (Howard-Pitney, Borgida, & Omoto, 1986), or con-tracting physical illness (Kassarjian & Cohen, 1965; Kunda,1987)—increased personal relevance promotes increased defen-sive or self-protective processing of message content (for a review,see Kunda, 1990).

This finding is also consistent with research demonstrating thatsmokers significantly underestimate the health risks of cigarettesmoking (Hansen & Malotte, 1986; Lee, 1989; McCoy et al., 1992;McMaster & Lee, 1991; Strecher, Kreuter, & Kobrin, 1995).Although smokers do perceive their health risks to be greater thanthose of nonsmokers, their estimates of personal risk still fall shortof their estimates of the health risks incurred by a so-called"typical smoker." Smokers appear to maintain this optimistic biasor illusion of personal immunity not by avoiding or ignoringdissonance-arousing information about the health risks of smokingbut rather by processing such information defensively: selectivelyavoiding it, questioning its validity, being critical of it, or simplydisbelieving it (Frey & Stahlberg, 1986; Halpern, 1994; Kassarjian& Cohen, 1965; Kunda, 1990; Litz, Payne, & Colletti, 1987;Pervin & Yatko, 1965; Wicklund & Brehm, 1976).

Evidence also suggests that the relationship between degree ofthe personal relevance of smoking and defensive processing ofinformation about its attendant health risks is positive and linear.McCoy et al. (1992) found stronger optimistic bias among currentsmokers as compared with smokers who had recently attempted toquit and former smokers who had successfully quit Halpern(1994) similarly found that, among current smokers, heavy smok-

Page 3: Defensive evaluation of antismoking messages among college-age smokers: The role of possible selves

426 FREEMAN, HENNESSY, AND MARZULLO

ere evaluated their health risks as lower than did lighter smokers.Among former smokers, Halpern also found both that long-termcurrent smokers evaluated their health risk as less than did short-term smokers and that ex-smokers who had quit recently evaluatedtheir health risk as lower than did ex-smokers who had quit severalyears before (cf. Gibbons & Gerrard, 1991; Tagliacozzo, 1979).

Operationalizing Personal Involvement With Smoking

The present study exposed smokers and nonsmokers to a varietyof antismoking messages and asked them to rate the effectivenessof the messages. Consistent with the literature outlined above, wepredicted that participants for whom antismoking messages aremore personally relevant—that is, those who are more personallyinvolved with cigarette smoking—should be more critical of themessages, defensively rating them as less effective.

Within this broad prediction about individual differences in theratings of antismoking message effectiveness, the present study'sunique contribution lay in how it conceptualized personal involve-ment with smoking in four different ways. First, and most simply,personal involvement was assessed as smoking behavior, includingmeasures of smoking rate and duration. We predicted that smokerswould defensively rate the antismoking videos as less effectivethan would nonsmokers, and that among smokers, heavy andlong-term smokers would evaluate the messages as less effectivethan would light, short-term smokers.

Second, personal involvement with smoking was assessed as theextent to which one's self-concept was aligned with that of the"prototypical smoker" (Amos, Gray, Currie, & Elton, 1997; Gib-bons & Gerrard, 1995; Gibbons, Gerrard, Blanton, & Russell,1998; Lloyd & Lucas, 1998; Lloyd, Lucas, & Fembach, 1997). Inprevious research, this variable has traditionally been treated as arisk factor for smoking initiation (Aloise-Young, Hennigan, &Graham, 1996; Barton, Chassin, Presson, & Sherman, 1982). Incontrast, in the present study, consistency between the self-conceptand the prototypical smoker has been conceptualized as an indi-cator of personal involvement with smoking, following Gibbonsand others (Gibbons & Gerrard, 1991,1995), who found that suchconsistency closely tracks actual smoking behavior and recipro-cally reinforces smoking intentions. Gibbons and Gerrard (1995)found in a longitudinal study that the smoker prototype was ratedas more self-descriptive by those who were planning to smoke andactually wound up smoking at Time 2 but was rated as* lessself-relevant over time by current smokers who were trying to quit(Gibbons & Gerrard, 1991). We thus hypothesized here that in-creased correspondence of the self-concept with the smoker pro-totype, as an indicator of personal involvement with smoking,would be associated with decreased ratings of antismoking mes-sage effectiveness.

Third, the personal relevance of cigarette smoking was assessedas the degree to which cigarette smoking was perceived to be adefining characteristic of the current self-concept (Shadel & Mer-melstein, 1996). Previous research has demonstrated that posses-sion of a smoker self-identity is associated with increased smokingbehavior (Lloyd & Lucas, 1998). Moreover, a stronger smokerself-identity predicts decreased intentions to quit smoking in thefuture (Falomir & Invernizzi, 1999) as well as decreased success ofcessation efforts among smokers attempting to quit (Shadel &Mermelstein, 1996). In the present study, then, we hypothesized

that smokers for whom smoking was an important self-identitywould rate antismoking messages as less effective than wouldthose for whom it was not.

Fourth, the present study assessed personal involvement withsmoking as the degree to which cigarette smoking was a compo-nent of a current smoker's "possible self (Markus & Nurius,1986). Markus and Nurius (1986) were among the first to arguethat the self-concept contains specific representations of the self infuture states, so-called "possible selves" that represent what theself "would like to become ... could become ... or [is] afraid ofbecoming" (p. 954). Though related to one another, these repre-sentations correspond to separate aspects of the future self:"hoped-for" selves, "expected" selves, and "feared" selves (cf.Cross & Markus, 1991; Oyserman & Markus, 1990).

Possible selves serve two primary functions with respect to thecurrent or "now" self: They serve as motivators for behavior in thepresent, and, "by providing an evaluative and interpretive contextfor the current view of the self, possible selves are instrumental inthe affirmation and defense of the 'now' self... [they] serve asstandards for comparison or evaluation of the current self (Cross& Markus, 1991, p. 232). Thus, students with a high schoolgraduate possible self for the future will construe a D received onan assignment in the present quite differently than students with ahigh school dropout possible self (Oyserman, Gant, & Ager, 1995;Ruvolo & Markus, 1992), an elderly person with an importantpossible self in the domain of health has a higher motivation thanone without such a possible self to perform specific health-relatedbehaviors in the present (Hooker, 1994; Hooker & Kaus, 1994),and an adolescent with a risky or delinquent possible self will viewengagement in risky behaviors in the present differently thansomeone without such a possible self (Oyserman & Markus, 1990;Stein, Roeser, & Markus, 1998). By analogy, the possession of asmoking possible self can provide an important context for theconstrual of one's own smoking behavior in the present as well forthe evaluation of antismoking messages. Specifically, this studyhypothesized that college-age smokers who possess a long-termsmoking possible self should evaluate antismoking messages in thepresent as less effective than those who lack such a possible self.

This hypothesis about college-age smokers is based on twoobservations. First, tobacco exacts its heaviest toll on health onlywith long-term, chronic use. Appropriately, therefore, the contentof many antismoking messages is anchored squarely in the youngadult's future, raising awareness of the adverse events that mayoccur if smoking continues: life-threatening illness, addiction andinability to quit, harm to family and friends because of secondhandsmoke, and so on. Such antismoking messages essentially presentsmokers with alternative narratives of the future—that is, withhighly aversive smoking-related possible selves—and attempt toanchor them in their current smoking behavior. Smokers whoalready possess a (presumably positive or at least neutrally va-lenced) long-term smoking possible self should experience themost dissonance when presented with these highly aversive nar-ratives of alternative smoking possible selves and will defend theirown preferred possible self (and current smoking behavior) bycontesting these alternative narratives. In the context of the presentstudy, the most readily available means for doing this is to givelow ratings of the narratives' persuasiveness and effectiveness (cf.Falomir & Invernizzi, 1999).

Page 4: Defensive evaluation of antismoking messages among college-age smokers: The role of possible selves

SMOKING AND POSSIBLE SELVES 427

Second, and conversely, although some adolescent and college-age smokers may possess a long-term smoking possible self,surveys confirm that a majority do not because they report strongintentions of quitting in the future (Burt & Peterson, 1998; Mc-Kenna & Williams, 1993; Pechmann, 1997; Stanton, Lowe, &Gillespie, 1996). As one 19-year-old female college student par-ticipant in our study noted, "This is only something I'm doing fornow—it helps me with the stress [of college].... I plan to quitafter I graduate." Statements such as these on the nation's collegecampuses (Davis & Fowler, 1999) may well represent mere wish-ful thinking, but the important fact here is that they indicate thepresence of a smoke-free possible self in the average college-agesmoker's self-concept. If correct, the hypothesis concerning therole of smoking possible selves would suggest that these individ-uals—in contrast to those who do possess a smoking possibleself—would not be threatened by the narratives of aversive pos-sible selves that are embedded in many antismoking messagesbecause they would perceive these narratives not to be relevant tothem. As a result, such smokers should rate the messages asrelatively more persuasive and effective.

Method

Participants

Participants responded to university-wide telephone messages, e-mails,and announcements describing the study and offering cash prizes forparticipation, ranging from $ 10 to $250, on a lottery basis. The request alsostated that the study was especially interested in recruiting "regular smok-ers." The message thus initially went out to approximately 1,500 students.A total of 308 students responded initially to this campus-wide request,with 124 (40%) of these identifying themselves as regular smokers.

From this initial list of smokers, a total of 13 individuals were excludedfrom participation because follow-up questions revealed that they smokedless than once per week. Of the remaining 111 smokers, 82 (74% of theeligible smokers who initially responded to the campus-wide request) weresuccessfully recruited to participate. A total of 105 nonsmokers alsoparticipated (57% of the 184 nonsmokers who initially responded to thecampus-wide request).

The resulting sample ranged in age from 18 to 22 years (M = 20.09) andwas 65% female (the school's student population was 59% female). Therewas no significant relationship in the sample between age and smokingstatus, x*(4, N = 187) = 0.22, p = .64, or between sex and smoking status,X*(l, N = 187) = 3.70, p = .45. The sample contained 160 (86%)Europeans/North Americans, 12 (6%) Asian Americans, 8 (4%) Latinos orLatin Americans, and 1 (1%) African American, with 6 (3%) reportingtheir ethnicity as "other." There were roughly equal numbers of freshman,sophomore, junior, and senior students in the sample. No data werecollected on the characteristics of nonrespondents, but the obtained sampleis demographically quite representative of the undergraduate college pop-ulation from which it was drawn.

Selection of Video Materials

To examine the possibility that different persuasive strategies would bemore or less effective with different kinds of smokers, video segmentsfeaturing a variety of antismoking themes were used. The antismokingvideos were chosen from those contained in Waves I, n, and HI of theCenters for Disease Control's Media Campaign Resource Center videocatalog (DHHS Centers for Disease Control and Prevention, 1998, 1999).Mark A. Freeman and a research assistant independently viewed the entirevideo collection and developed coding schemes according to the nature ofthe persuasive strategies used in the videos. Goldman and Glantz's (1998)

coding scheme provided a starting point for this informal content analysis.The final coding scheme, which slightly modifies the latter, consisted ofthe eight categories described in Table 1.

Two 30-s videos were then selected to represent each category. Thevideos were chosen according to the following three subjective criteria: (a)the video had to be a relatively pure example of the theme in question, andthus videos using multiple themes were excluded; (b) it had to be appro-priate for a college-age audience; and (c) it could not use well-knownpublic figures. The selected videos (see Table 1) were converted intocomputer files for viewing through a Web browser as part of a computer-administered questionnaire.

To validate the reliability of this coding scheme, a separate sample of 20judges (10 male, 10 female undergraduate students) were familiarized withits eight categories, viewed all 16 selected videos, and then viewed themagain one at a time, classifying each in one of the eight categories as theydid so. The percentage of interobserver agreement averaged 88% acrossthe 16 videos (ranging from 80% to 100% across each of the videos), andin all cases the most frequently selected category for a video was the onedisplayed in Table 1. The eight-category coding scheme was thereforesufficiently reliable, and the videos were adequately clear representationsof the eight themes.

Independent Variables: Smoking-Related Behaviors andAttitudes

Smoking behavior. After providing some basic demographic informa-tion, those who had identified themselves as smokers during the recruit-ment phase (see above) were directed to the detailed questions concerningsraoking-related attitudes and behavior. Nonsmokers skipped this section.To assess smoking rate, smoking participants were asked, "How much doyou currently smoke?" (1 = less than one cigarette per day, 7 = morethan 2 packs per day). The median response was one-half pack (about 10cigarettes) per day. To assess smoking duration, smokers were asked,"When did you first start smoking on a regular basis?" (1 = less than 3months ago, 1 = more than 5 years ago). Here the median response was"3 years ago." An index of smoking-cessation behavior, also included,asked respondents how many times they had "tried to quit in the past year"(1 = never, 2 = not in the past year, 7 = more than 5 times in the pastyear). Here the median response was "twice." For all three smokingbehavior measures, responses spanned the full range of the scales, and thedistribution of responses did not deviate significantly from normality.

Conformity to smoker prototype. All 187 participants described them-selves in terms of a series of traits that are known to be aspects of astereotype associated with smokers (Amos et al., 1997; Barton et al., 1982;Gibbons & Gerrard, 1995; Lloyd & Lucas, 1998; Lloyd et al., 1997). Thetraits were presented as a series of semantic differential items: unhealthy-healthy, beautiful-not beautiful, likes to be with a group-likes to be alone,tough-gentle, brave-not brave, wise-not wise, sexy-not sexy, popular-notpopular, sociable-not sociable, smart-not smart, outgoing-shy, exciting-not exciting. The first item in each pair was part of the smoker stereotype(half of the pairs were reversed on the questionnaire itself). Participants'answers on a 16-point scale were scored so that high scores indicatedconformity of the participant's self-concept with the smoker stereotype.The items unhealthy and tough were dropped because of low item-totalcorrelations, and a single index (Cronbach's a = .83) was computed fromthe average of the remaining 10 items.

Current smoking identity. The last set of items assessed the degree towhich smoking was a part of the current self-concept using Shadel andMermelstein's (1996) five-item Smoker Self-Concept Scale. The scale asksparticipants to agree or disagree (6 =strongly agree, 1 = strongly dis-agree) with statements that assert the centrality of smoking to respondents'self-concept, personality, daily life, and so on. Factor analysis (varimaxrotation) identified the one-factor structure of the scale, as was the case inShadel and Mermelstein's original validation of the measure. The itemswere averaged to yield a single index of current smoking identity (Cron-

Page 5: Defensive evaluation of antismoking messages among college-age smokers: The role of possible selves

428 FREEMAN, HENNESSY, AND MARZULLO

Table 1Video Coding Scheme

Video category Two videos representing category (title and description)

1. Long-term health

2. Secondhand smoke

3. Addiction

4. Industry manipulation

5. Cessation

6. Smoking is disgusting

7. Romantic rejection

8. Smoking is not cool

Lung: Animation of smoke entering lung, causing decay.Voice Box: Woman states she cannot stop smoking, ends by smoking through

a hole in her throat.Restaurant: Shows waitresses in smoky restaurant, with voiceover "Obey the

new laws against secondhand smoke."Silence: Slow-motion text: "30 seconds of silence for the 14 Californians who

will die today of second-hand smoke."Breath: Young man underwater must come up for air, voiceover explains,

"That's what addiction feels like."Cold: Young man lighting up outdoors on a bitterly cold day—cigarettes are

the only reason he'd be outside on such a day.The Truth: Medical scientist discusses big tobacco's repression of his internal

research showing addictive properties of nicotine.Industry Spokesman: Tobacco executives in a smoky room declare need to

recruit new, young smokers.Quitting Takes Practice: Animated character learns that quitting takes

practice.Ask Your Doctor: Doctors "surgically" remove cigarette from smoker's grasp.In Your Mouth: Reverse time-lapse sequence of cigarette on city sidewalk

"returning" to smoker's mouth.Cow: Flatulent cow, with voiceover warning that methane gas is also

contained in cigarettes.Me Strikes Out: Animated male smoker fails to get a date.Personals: Personal columns that show even "desperate" people specifying

nonsmokers as their desired mates.Toilet Paper: Young man strutting with cigarette in front of young women

who are not impressed.Lonely People: Depicts smokers lighting up in dingy places; text at end states,

"Smoking: It's not as glamorous as it looks."

bach's a = .89). On a 7-point scale, the mean average response was 4.16(SD = 1.32).

Future smoking identity. To assess the degree to which smoking be-havior was incorporated into participants' possible selves, participantswere asked to "try to actually visualize yourself' at a number of future timeperiods, and then to "indicate how likely it is that you will be smoking ateach of these future time periods" (1 = definitely not, 0% chance; 7 =definitely, 100% chance). The 10 time periods in this index ranged from"one month" to "20 years" and included the specific events "graduate fromcollege," "have a full-time job," "live with a romantic partner," "when/ifyou get married," and "when/if you have children."

Responses were factor analyzed using varimax rotation, and a cleartwo-factor structure emerged: Four items ("20 years," "10 years,".,"getmarried," and "have children") loaded highly on a long-term factor, threeitems ("1 month," "1 year," and "graduate") loaded highly on a short-termfactor, and the three items referencing time periods of intermediate length("5 years," "have a full-time job," and "live with a romantic partner")loaded on both factors. These three items were dropped from the scale, andtwo factor scores were generated, one based on the four long-term items(Factor 1: 53% variance) and another based on the three short-term items(Factor 2: 23% variance), yielding indexes that assessed how much smok-ing was a part of participants' long-term (Cronbach's a = .86) andshort-term (Cronbach's a = .84) possible selves.

The mean response to a composite measure of the three (unnormalized)indicators of short-term smoking identity was 2.61 (SD = 1.12) on the7-point scale, which corresponds to about a 70% chance estimate, aboutmidway between the qualitative responses "very likely (60%)" and "prob-ably (80%)." The mean response to the average of the four (unnormalized)indicators of long-term smoking identity was much lower, though stillwidely distributed, on the same 7-point scale (M = 5.59, SD = 1.17),corresponding to about a 15% chance estimate about midway between the

qualitative responses "very unlikely (20%)" and "extremely unlikely(10%)." Thus, the average smoker in this sample, as in the college-agesmoking population as a whole, stated that he or she was very likely tocontinue smoking in the short term but possessed a long-term possible selfthat was very unlikely to include cigarette smoking.

Intercorrelations among independent variables. The intercorrelationsamong the independent variables for the smokers in the sample are dis-played in Table 2. Heavy smokers were more likely to identify thecharacteristics of the smoker stereotype as self-relevant and were muchmore likely to claim that smoking was an important part of their currentself-concept. Of special interest is the lack of any strong relationshipbetween current smoking self-identity and having a smoking identity as along-term (r = .20, ns) or a short-term (r = .19, ns) possible self. Thisrinding again supports the idea that college-age smokers with an estab-lished smoking habit and those with a strong current smoking identity donot necessarily imagine themselves smoking in the future.

Dependent Measures: Video Evaluations

In the third section of the questionnaire, each participant viewed one ofsix video sequences containing one video from each of the eight videocategories. Each video sequence presented the video categories in randomorder, to control for ordering effects. After viewing each video, the par-ticipants evaluated it on five 16-point semantic differential items:persuasive-not persuasive, good-bad, clever-stupid, effective-not effec-tive, makes me want to quit smoking-does not make me want to quitsmoking (nonsmokers completing the questionnaire were asked to answerthe last item "as if' they smoked currently). Factor analysis of the five-itemrating scales for all 16 videos revealed that they each defined single factors,explaining from 70% to 89% of the matrix variance (Cronbach's as rangedfrom .89 to .97).

Page 6: Defensive evaluation of antismoking messages among college-age smokers: The role of possible selves

SMOKING AND POSSIBLE SELVES 429

Table 2Intercorrelation Matrix for Independent Variables (Smokers Only, n = 82)

Variable 1

Smoking behavior measures

1. Smoking rate2. Smoking duration3. Smoking cessation behavior4. Smoker stereotype

5. Current6. Short-term future7. Long-term future

.46**

.23* .23*

.29** .19

Smoking identity

.60** .31**-.09 -.15

.13 .01

-.10

measures

.23*-.04-.28**

.04-.19

.04.19.20 .40**

*p < .05 (two-tailed). ** p < .01 (two-tailed).

Factor scores generated for each video yielded an index of how effectivean individual rated that video relative to the rest of the sample that viewedthat video. The standardized effectiveness ratings for the two videos withineach category were combined to yield eight indexes, one for each videocategory (see Table 1). Thus, for example, a new variable, addiction, wascreated, which for about half of the respondents was the standardizedeffectiveness index for the video Breath and for the other half was thestandardized effectiveness index for the video Cold.1 Finally, the averageeffectiveness rating for all eight video categories was computed as aseparate variable (Cronbach's a = .72).

Procedure

Participants, in groups of 6 at most, completed the computer-administered questionnaire in a private classroom computer facility oncampus that was reserved for the times of data collection. Participants weredirected to sit at computer terminals along the room's outer walls, facingaway from each other and wearing headphones to ensure that there was novideo or audio interference from neighboring terminals. The questionnaireand display of all video materials were automated, so participants were ableto view the antismoking video segments and provide their feedback with-out any intervention from the researcher.

Results

Nonsmokers consistently rated the videos as significantly moreeffective than did smokers (see Table 3). This finding is consistentwith the hypothesis that personal involvement with smoking in-creases the extent to which individuals respond defensively toantismoking messages. A breakdown of this result by video cate-gory revealed that this finding held true for all video categories butwas not statistically significant for two video categories, long-termhealth and cessation.

As can be seen in the first three rows of Table 4, the threemeasures of current smoking behavior calculated for the 82 smok-ers in the sample were not substantially related to the averagerating of video effectiveness. The smoker stereotype variable,assessed for the entire sample, was strongly associated with actualsmoking status, rpfc(185) = .37, p<. 001, and among smokers withsmoking rate, K80) = .29, p = .008, and smoking duration, r =.27, p = .014 (all two-tailed). However, the smoker stereotypevariable was not significantly correlated with any of the videoeffectiveness ratings or with the average effectiveness ratings, bothfor the entire sample and among smokers only (see Table 4).

The current smoking self-concept variable was not significantlyrelated to the average rating of video effectiveness, though theassociation represents a "small" (Cohen, 1977) effect in the pre-dicted direction. Though not significant if alpha is adjusted formultiple tests, the strength of current smoking identity exhibited asmall to moderate negative correlation with effectiveness ratingsfor the secondhand smoke and addiction video categories (seeTable 4).

The measure of smoking identity as a short-term possible selfwas also not significantly associated with the average rating ofvideo effectiveness, though its effects were again in the predicteddirection and represent a small effect This variable was alsoassociated with lower effectiveness ratings for the cessation videocategory. It stands to reason that smokers who foresee themselvessmoking for the next year or two would respond defensively toantismoking messages that highlight the benefits of quitting.

Of all seven measures of smoking behavior and attitudes, long-term smoking identity was the best predictor of participants'responses to antismoking messages. The negative relationship ofthis variable with the average effectiveness rating does not exceedthe Bonferroni-adjusted alpha for multiple planned comparisons,but the relationship is of small to moderate size (Cohen, 1977).Smokers with a long-term future smoking possible self appear to

1 Two videos were used to represent each video category to ensure thatthe data as a whole would reflect participants' responses to a type ofpersuasive message (e.g., "smoking is addicting," "smoking is disgusting")and not to any idiosyncratic features of the video used as an example of thattype of persuasive message. However, because each participant actuallyviewed only one of the two videos used as examples of each persuasivetheme, the possibility existed that the data, as analyzed by video category,might conceal differential reactions to each video in that category. For allreported analyses involving ratings of these video categories as dependentvariables, then, dummy variables representing the video viewed within acategory (as well as Video X Smoking Measure interactions, where ap-propriate) were created and run as independent variables. In no case didthese dummy variables have effects even approaching significance on theratings of video category, reinforcing the idea that participants were re-sponding to the theme common to the two videos representing it. For thesake of simplicity, then, the data reported below are univariate analyses ofthe effects of the smoking measures only, without controlling for theinsignificant effects of a video viewed within a category.

Page 7: Defensive evaluation of antismoking messages among college-age smokers: The role of possible selves

430 FREEMAN, HENNESSY, AND MARZULLO

Table3Ratings of Video Effectiveness for Smokers and Nonsmokers

Video t df p*

1. Long-term health2. Secondhand smoke3. Addiction4. Industry manipulationS. Cessation6. Smoking is disgusting7. Romantic rejection8. Smoking is not cool

Average video rating

1.092.90"3.482.350.743.423.452.25

4.20"

178133183184184184184185

154

.140

.002<.001

.006

.220<.001<.001

.013

<.001

.08

.24

.25

.17

.05

.24

.25

.16

.32

Note. Positive / values indicate that nonsmokers rated video type as moreeffective.• All p values are one-tailed. Bonfeironi-adjusted o^ = .0063 for eightplanned comparisons. b t test result calculated assuming unequalvariances.

respond most defensively to five of the eight video categories—long-term health, secondhand smoke, smoking is disgusting, ro-mantic rejection, and smoking is not cool. Thus, of the variablesexamined, long-term smoking identity represents the best predictorof individual differences in responses to antismoking messages.

Discussion

Conclusions

There are individual differences in college students' responsesto antismoking messages, and these differences are interpretablewithin a cognitive dissonance framework provided that an impor-tant mediating variable—the personal relevance of cigarette smok-ing—is operationalized appropriately. Among smokers, identity-based measures of the personal relevance of cigarette smokingwere the best predictors of a defensive response to antismokingmessages. Current smoking identity is the most obvious measure

of this construct, and this variable does a moderately good job ofpredicting negative ratings of two categories of antismoking mes-sages: those emphasizing the addictive properties of tobacco andthose highlighting the hazards of secondhand smoke. However, themost significant finding of the present study is that because mostantismoking messages contain a strong emphasis on the adversefuture consequences of smoking, they are most personally relevantto a smoker's future or possible self; thus, it is in this context thatthey provoke the most dissonance and consequent defensive eval-uation (cf. Falomir & Invernizzi, 1999).

This finding contributes to a developing approach in healthpsychology highlighting the role of the self, and particularly ofpossible selves, in mediating disease-related behavior in thepresent (Contrada & Ashmore, 1999; Hooker, 1992; Hooker &Kaus, 1994; Stein et al., 1998). Though this body of research hastended thus far to focus on adulthood and old age, it is consistentwith the present study's basic assertion that possible selves canprovide an important "evaluative and interpretive context" forcurrent behavior and experience (Markus & Nurius, 1986, p. 955).

Previous research has examined the role of future smokingintentions on current smoking-related attitudes and behaviors, usu-ally assessed as an "intention to quit" (e.g., Burt & Peterson, 1998;Falomir & Invernizzi, 1999; Stanton et al., W96; see also Gibbonset al., 1998). In contrast, the present study links such motivationsexplicitly with the expected or possible selves that populate theindividual smokers' self-concept. This represents an importanttheoretical shift of perspective for two reasons. First, examiningfuture smoking intentions through the theoretical lens of possibleselves serves to anchor this intention in a specific time period andconnects it with other important possible selves, such as collegegraduate, parent, or spouse. This approach arguably representsyoung adult smokers' own conceptions of their future smokingbehavior more accurately.

Second, this conceptualization of future smoking intentions ascomponents of possible selves provides a much more theoreticallysatisfying understanding of the mechanism by which such future

Table 4Smoldng-Related Behaviors and Attitudes and Video Effectiveness Ratings: Simple Correlations

Smoking-related behavior/attitude

Smoking behaviorsSmoking rateSmoking durationSmoking cessation behavior

Smoker stereotypeEntire sample (N = 187)Smokers only (n = 82)

Smoking identity measuresCurrentShort-term futureLong-term future

Averagevideorating*

-.11-.03

.05

-.02-.02

-.12-.11-.23*

Video category rating-j

1

-.02.22*.06

.09-.01

-.06-.03-.23*

2

-.14-.04

.04

-.09.02

-.23*-.10-.35**

3

-.08-.07

.03

-.03.12

-.25*-.17-.03

4

-.11-.10-.13

.08

.14

-.17.03

-.14

5

-.11-.06

.16

-.03-.05

.09-.20*-.03

6

-.02-.02

.13

-.06-.13

-.07-.16-.20*

7

-.08.01

-.09

-.08-.02

-.11-.09-.22*

8

.04-.02

.05

.01-.11

.03-.13-.22*

Note. 1 = long-term health, 2 = secondhand smoke, 3 = addiction, 4 = industry manipulation, 5 = cessation,6 = smoking is disgusting, 7 = romantic rejection, 8 = smoking is not cool. Because of missing values, validN ranged from 78 to 82.* Bonferroni-adjusted otpy, = .007 for seven planned comparisons involving the average video rating; withn = 82, r = .27 has a p value of .007 (one-tailed).*p < .05 (one-tailed). **p < .01 (one-tailed).

Page 8: Defensive evaluation of antismoking messages among college-age smokers: The role of possible selves

SMOKING AND POSSIBLE SELVES 431

intentions are involved in the defensive evaluation of antismokingmessages. It is a fact that many current college-age smokers simplydo not see themselves smoking in the long term, and this lack ofa smoking possible self would seem to represent perhaps the bestfoundation on which to construct continued antismoking cam-paigns. As such, it is critically important that these efforts beginwith a clear understanding of how this intention is psychologicallyrepresented—as a motivation? an affective state? a cognition? anattitude?—to transform this apparently widespread intention toquit into actual cessation attempts. Viewing such future motiva-tions as possible selves provides just such an understanding and,moreover, anchors them in an established literature on the role ofself-identity in the processing of persuasive messages (Bumkrant& Unnava, 1995; Falomir & Invernizzi, 1999; Levin et al., 2000;Liberman & Chaiken, 1992).

It is also interesting to speculate on the data concerning whichspecific types of antismoking messages seemed to produce themost dissonance among smokers with current, short-term, andlong-term smoking identities. Among smokers with a strong cur-rent smoking identity, messages that highlight the immediate aver-sive consequences of smoking might be most threatening andtherefore might be rated as least effective. Current smoking iden-tity was actually moderately negatively associated with the sec-ondhand smoke and addiction categories, which supports thisspeculation, though other categories emphasizing immediate con-sequences—industry manipulation, smoking is disgusting, roman-tic rejection, and smoking is not cool—do not.

The short-term smoking possible self variable was moderatelynegatively associated with ratings of the cessation video category.As noted above, this finding is consistent with the idea thatsmokers respond most defensively to messages that are mostpersonally relevant: Smokers who imagine themselves smoking hithe short-term future should be particularly motivated to discreditattempts to persuade them of the benefits of quitting in the sametime period. The long-term smoking possible self variable is in factmoderately negatively related to ratings of the antismoking mes-sage most directly relevant to long-term smoking: long-termhealth. Moreover, this is the only smoking-related variable that issubstantially associated with this video category (see Table 4). Thelong-term smoking possible self variable is also moderately neg-atively related to ratings of the secondhand smoke and romanticrejection themes, both of which have long-term implications aswell as immediate ones. However, this variable is also moderatelynegatively related to the ratings of two message categories that arenot clearly related to long-term smoking: the smoking is disgustingand smoking is not cool messages. Overall, then, these correla-tions, though not cleanly interpretable and based on a relativelylimited sample, are sufficiently intriguing to merit future researchinto the cognitive mechanisms that link smokers' responses tospecific antismoking messages with specific components of thesmoking self-concept.

Limitations of the Present Study and Directions forFuture Research

The findings of this study point to a need for longitudinalresearch on the effectiveness of antismoking messages, with actualsmoking behavior as the dependent variable, to answer the impor-tant questions it raises: How many college students actually adhere

to their professed plans to quit soon after graduation? What is therelationship between having a long-term smoking identity andlong-term smoking behavior? Does defensive evaluation of anti-smoking messages serve to resolve the dissonance aroused insmokers by antismoking messages, making subsequent behavioralchange less likely, or does it merely signal the presence of ahealthy dissonance that will only be resolved by eventual behav-ioral change? These questions can only be adequately addressedthrough a longitudinal analysis of actual smoking behavior.

The last of the questions above is probably the most crucial asfar as the future of antismoking media campaigns are concerned.The cognitive dissonance paradigm was developed with the as-sumption that when dissonance exists between a cognitive and abehavioral element in conscious awareness, resultant psychologi-cal discomfort is most readily resolved through change in thecognitive element (Festinger, 1957). In fact, early dissonanceresearch designs typically manipulated a behavioral element andassessed cognitive change as the dependent variable. In contrast,research on the effectiveness of antismoking campaigns reversesthis paradigm by introducing dissonant cognitive elements in thehopes of producing subsequent behavioral change.

The present study, as well as the literature on the processing ofthreatening messages, suggests that instead of changing behavior,the reaction to antismoking messages may instead be to processany new, dissonant messages defensively. In the worst possiblecase, then, exposing smokers to antismoking messages may ineffect be training them to discredit the very rationales that couldotherwise provide a foundation for subsequent behavioral change(Falomir & Invernizzi, 1999). Though this seems intuitively un-likely, the fact that college-age smoking saw its sharpest increasein decades during the period from 1993 to 1997, in spite ofunprecedented antismoking efforts in the media, demands moreattention than it has received (Wechsler et al., 1998). It is possiblethat such media campaigns may have the desired impact oncollege-age smokers overall, while having no impact or even areverse effect for the subpopulation of smokers who already pos-sess an established long-term smoking possible self. Such possi-bilities should be investigated through longitudinal research.

Longitudinal research would also be able to rule out any third-variable explanations of the association between the personalinvolvement with smoking—particularly when it is operational-ized as a long-term smoking possible self—and defensive evalu-ations of antismoking messages. Though reversal of the causalarrow in this case is unlikely—that is, that assessments of one'ssmoking possible self were made before the viewing of the anti-smoking videos in this study—it is possible that a third variablemay be mediating the relationship. For example, a "delinquent"(Oyserman & Markus, 1990) or "risky" (Gibbons et al., 1998;Stein et al., 1998) self-concept may well be associated with both along-term smoking possible self and a defensive response towardantismoking messages in the present. Carefully designed longitu-dinal research would be able to disentangle these possible rivalexplanations of the current results.

In addition, the present study focused primarily on a practicallyimportant, though psychologically speaking relatively coarse, de-pendent variable: overall effectiveness ratings of the antismokingmessages on a number of dimensions. The dissonance-personalinvolvement paradigm is consistent with these data, and in someversion is likely to be involved with the defensive response to

Page 9: Defensive evaluation of antismoking messages among college-age smokers: The role of possible selves

432 FREEMAN, HENNESSY, AND MARZULLO

antismoking messages of smokers in general and of those with along-term smoking possible self in particular. Increased personalinvolvement with smoking on these dimensions does in fact seemto promote defensive evaluation of antismoking messages. How-ever, the precise nature of the dominant dissonance-reducing cog-nitive mechanism remains an open question. Smokers with along-term smoking possible self may reduce the dissonancearoused by antismoking messages by simply avoiding such mes-sages (Frey & Stahlberg, 1986; Tagliacozzo, 1979; Wicklund &Brehm, 1976) or through more cognitively elaborate processes of"motivated reasoning" (Kunda, 1987, 1990), "schematic process-ing" (Litz et aL, 1987), or "defensive processing" (Liberman &Chaiken, 1992) of smoking's risks. These remain intriguing ques-tions for future research.

The study design did not permit an examination of nonrespon-dents, so it cannot eliminate the possibility of nonresponse bias inthese data. In particular, smoking nonrespondents may well differfrom smoking respondents: Such a study of smoking behavior mayhave attracted smokers who, if not necessarily more established intheir smoking habit (smoking participants in the study actually didpresent a full range of smoking rates and durations), were morecomfortable with a smoker social identity. Within the theoreticalframework of the present study, such participants should be ex-pected to experience greater cognitive dissonance with exposure toantismoking materials and thus exhibit greater defensive process-ing. The positive findings for the key measure of personal involve-ment with smoking in the present study, future smoking identity,would not seem to be significantly threatened by this possibility,but the null results for the other measures—smoking rate, duration,cessation behavior, and the smoker stereotype variable and currentsmoking identity—may well be an artifact of such a pattern ofnonresponse bias. Future research with more careful samplingprocedures should be conducted before the impact of theseindividual-differences variables on response to antismoking mes-sages is discounted.

Finally, it should be noted that the antismoking media effort todate has primarily focused on adolescents and has especially hadthe goal of reducing smoking initiation. From a cognitive disso-nance perspective, a focus on this age group makes good sense.Pro-smoking cognitions internalized from tobacco advertising orfrom peer-group smoking can be counteracted far more effectivelywith antismoking messages before those pro-smoking cognitionsbecome anchored in a behavioral commitment to cigarette smok-ing. However, this focus on reducing smoking initiation among theadolescent population ignores a large and apparently growingpopulation of college-age smokers for whom cigarette smokingmay already be a fairly well-established and highly personallyrelevant activity. In addition to targeting would-be adolescentsmokers, future research therefore needs to focus more intensivelyon this important group, tailoring persuasive strategies to targetpro-smoking attitudes that have already found root in the habitualbehaviors, current self-concepts, and future possible selves ofcollege-age smokers.

References

Aloise-Young, P. A., Hennigan, K. M. & Graham, I. W. (1996). Role ofthe self-image and smoker stereotype in smoking onset during earlyadolescence: A longitudinal study. Health Psychology, 15, 494-497.

Amos, A., Gray, D., Currie, C., & Elton, R. (1997). Healthy or druggy?Self-image, ideal image, and smoking behaviour among young people.Social Science and Medicine, 45, 847-858.

Barton, J., Chassin, L., Presson, C. C., & Sherman, S. J. (1982). Socialimage factors as motivators of smoking initiation in early and middleadolescence. Child Development, 53, 1499-1511.

Bumkrant, R. E., & Unnava, H. R. (1995). Effects of self-referencing onpersuasion. Journal of Consumer Research, 22, 17-26.

Burt, R. D., & Peterson, A. V. (1998). Smoking cessation among highschool seniors. Preventive Medicine, 27, 319-327.

Chaiken, S. (1987). The heuristic model of persuasion. In M. P. Zanna,J. M. Olson, & C. P. Herman (Eds.), Social influence: The OntarioSymposium (Vol. 5, pp. 3-39). Hillsdale, NJ: Elibaum.

Chaiken, S., Liberman, A., & Eagly, A. H. (1989). Heuristic and systematicprocessing within and beyond the persuasion context. In J. S. Uleman &J. A. Bargh (Eds.), Unintended thought (pp. 212-252). New Yoik:Guilford Press.

Cohen, J. (1977). Statistical power analysis for the behavioral sciences.New York: Academic Press.

Contrada, R. J., & Ashmore, R. D. (Eds.). (1999). Self, social identity andhealth: Interdisciplinary explorations. New York: Oxford UniversityPress.

Cooper, J., & Fazio, R. H. (1984). A new look at dissonance theory. In L.Berkowitz (Ed.), Advances in experimental social psychology (Vol. 17,pp. 168-189). Reading, MA: Addison-Wesley.

Cross, S., & Markus, H. (1991). Possible selves across the life span.Human Development, 34, 230-255.

Davis, A., & Fowler, B. (1999, March 4). Collegians lighting up: Smoke-free zones spreading across nation's campuses. USA Today, p. 1.

Falomir, J. M., & Invemizzi, F. (1999). The role of social influence andsmoker identity in resistance to smoking cessation. Swiss Journal ofPsychology, 58, 73-84.

Festinger, L. (1957). A theory of cognitive dissonance. Evanston, IL: Row,Peterson.

Frey, D., & Stahlberg, D. (1986). Selection of information after receivingmore or less reliable self-threatening information. Personality and SocialPsychology Bulletin, 12, 434-441.

Gibbons, F. X., & Gerrard, M. (1991). Downward social comparison andcoping with threat. In J. M. Suls & T. A. Wills (Eds.), Social compar-ison: Contemporary theory and research (pp. 317-345). Hillsdale, NJ:Erlbaum.

Gibbons, F. X., & Gerrard, M. (1995). Predicting young adults' health riskbehavior. Journal of Personality and Social Psychology, 69, 505-517.

Gibbons, F. X., Gerrard, M., Blanton, H., & Russell, D. W. (1998).Reasoned action and social reaction: Willingness and intention as inde-pendent predictors of health risk. Journal of Personality and SocialPsychology, 74, 1164-1180.

Goldman, L. K., & Glantz, S. A. (1998). Evaluation of antismokingadvertising campaigns. Journal of the American Medical Association,279, 772-777.

Halpern, M. (1994). Effect of smoking characteristics on cognitive disso-nance in current and former smokers. Addictive Behaviors, 19, 209-217.

Hansen, W. B., & Malotte, C. K. (1986). Perceived personal immunity:The development of beliefs about susceptibility to the consequences ofsmoking. Preventive Medicine, 15, 363-372.

Hooker, K. (1992). Possible selves and perceived health in older adults andcollege students. Journal of Gerontology, 47, P85-P95.

Hooker, K., & Kaus, C. R. (1994). Health-related possible selves in youngand middle adulthood. Psychology and Aging, 9, 126-133.

Howard-Pitney, B., Borgida, E., & Omoto, A. M. (1986). Personal involve-ment: An examination of processing differences. Social Cognition, 4,39-57.

Johnson, B. T., & Eagly, A. H. (1989). Effects of involvement on persua-sion: A meta-analysis. Psychological Bulletin, 106, 290-314.

Page 10: Defensive evaluation of antismoking messages among college-age smokers: The role of possible selves

SMOKING AND POSSIBLE SELVES 433

Johnston, L. D., O'Malley, P. M, & Bachman, J. G. (1999). Nationalsurvey results on drug use from the Monitoring the Future study,1975-1998. Volume 11: College students and young adults (NIH Publi-cation No. 99-4661). Rockville, MD: National Institute on Drug Abuse.

Johnston, L. D., O'Malley, P. M., & Bachman, J. G. (2000). The Moni-toring the Future national survey results on adolescent drug use: Over-view of key findings, 1999 (NIH Publication No. 00-4690). Rockville,MD: National Institute on Drug Abuse.

Kassarjian, H. H., & Cohen, J. B. (1965). Cognitive dissonance andconsumer behavior. California Management Review, 8, 55-64.

Kunda, Z. (1987). Motivation and inference: Self-serving generation andevaluation of evidence. Journal of Personality and Social Psychol-ogy. 53, 636-647.

Kunda, Z. (1990). The case for motivated reasoning. Psychological Bul-letin, 108, 480-498.

Lee, C. (1989). Perceptions of immunity to disease in adult smokers.Journal of Behavioral Medicine, 12, 267—277.

Levin, K. D., Nichols, D. R., & Johnson, B. T. (2000). Involvement andpersuasion: Attitude functions for the motivated processor. In G. R.Maio & J. M. Olson (Eds.), Why we evaluate: Functions of attitudes.Mahwah, NJ: Erlbaum.

Liberman, A., & Chaiken, S. (1992). Defensive processing of personallyrelevant health messages. Personality and Social Psychology Bulle-tin, 18, 669-679.

Litz, B. T., Payne, T. J., & Colletti, G. (1987). Schematic processing ofsmoking information by smokers and never-smokers. Cognitive Therapyand Research, 11, 301-313.

Lloyd, B., & Lucas, K. (1998). Smoking in adolescence: Images andidentities. London: Routledge.

Lloyd, B., Lucas, K., & Fembach, M. (1997). Adolescent girls* construc-tions of smoking identities. Implications for health promotion. Journalof Adolescence, 20, 43-56.

Lundgren, S. R., & Prislin, R. (1998). Motivated cognitive processing andattitude change. Personality and Social Psychology Bulletin, 24, 715-726.

Maikus, H., & Nurius, P. (1986). Possible selves. American Psycholo-gist, 41, 954-969.

McCoy, S. B., Gibbons, F. X., Reis, T. J., Gerrard, M., Luus, C. A. E., &Sufka, A. V. W. (1992). Perceptions of smoking risk as a function ofsmoking status. Journal of Behavioral Medicine, 15, 469-488.

McKenna, J., & Williams, K. N. (1993). Crafting effective tobacco coun-teradvertisements: Lessons from a failed campaign among teenagers.Public Health Reports, 708(Suppl. 1), 85-89.

McMaster, C., & Lee, C. (1991). Cognitive dissonance in tobacco smokers.Addictive Behaviors, 16, 349-353.

Oysennan, D., Cant, L., & Ager, J. (1995). A socially contextualizedmodel of African-American identity: Possible selves and school persis-tence. Journal of Personality and Social Psychology, 69, 1216-1232.

Oysennan, D., & Markus, H. R. (1990). Possible selves and delinquency.Journal of Personality and Social Psychology, 59, 112-125.

Pechmann, C. (1997). Does antismoking advertising combat underagesmoking? A review of past practices and research. In M. E. Goldberg, M.Fishbein, & S. E. Middlestadt (Eds.), Social marketing: Theoretical andpractical perspectives. Mahwah, NJ: Erlbaum.

Pervin, L., & Yatko, R. (1965). Cigarette smoking and alternative means ofreducing dissonance. Journal of Personality and Social Psychology, 2,30-36.

Petty, R. E., & Cacioppo, J. T. (1979). Issue involvement can increase ordecrease persuasion by enhancing message-relevant cognitive responses.Journal of Personality and Social Psychology, 37, 1915-1926.

Ruvolo, A., & Markus, H. (1992). Possible selves and performance: Thepower of self-relevant imagery. Social Cognition, 10, 95-125.

Shadel, W. G., & Mermelstein, R. (1996). Individual differences in self-concept among smokers attempting to quit: Validation and predictiveutility of t measures of the smoker self-concept and the abstainerself-concept. Annals of Behavioral Medicine, 18, 151-156.

Sherif, C. W., Sherif, M., & Nebergall, R. E. (1965). Attitude and attitudechange: The social judgment-involvement approach. Philadelphia:Saunders.

Sherif, M., & Hovland, C. I. (1961). Social judgment: Assimilation andcontrast effects in communication and attitude change. New Haven, CT:Yale University Press.

Stanton, W. R., Lowe, J. B., & Gillespie, A. M. (1996). Adolescents'experiences of smoking cessation. Drug and Alcohol Dependence, 43,63-70.

Stein, K. F., Roeser, R., & Markus, H. R. (1998). Self-schemas andpossible selves as predictors and outcomes of risky behaviors in adoles-cents. Nursing Research, 47, 96-106.

Strecher, V. J., Kreuter, M. W., & Kobrin, S.--C. (1995). Do cigarettesmokers have unrealistic perceptions of their heart attack, cancer, andstroke risks? Journal of Behavioral Medicine, 18, 45-54.

Stryker, S. (1987). Identity theory: Developments and extensions. In K.Yardley & T. Honess (Eds.), Self and identity: Psychosocial perspectives(pp. 89-103). New York: Wiley.

Tagliacozzo, R. (1979). Smokers' self-categorization and reduction ofcognitive dissonance. Addictive Behaviors, 4, 393-399.

U.S. Department of Health and Human Services. (1994). Preventing To-bacco Use Among Young People: A report of the Surgeon General.Atlanta, GA: U.S. Department of Health and Human Services, PublicHealth Service, Centers for Disease Control and Prevention, NationalCenter for Chronic Disease Prevention and Health Promotion, Office onSmoking and Health.

U.S. Department of Health and Human Services, Centers for DiseaseControl and Prevention. (1998). Media campaign resource center fortobacco control: 1998 video catalog. Atlanta, GA: Author.

U.S. Department of Health and Human Services, Centers for DiseaseControl and Prevention. (1999). Media campaign resource center fortobacco control: Wave III video catalog. Atlanta, GA: Author.

Wechsler, H., Rigotti, N. A., GledhiU-Hoyt, J., & Lee, H. (1998). Increasedlevels of cigarette use among college students: A cause for nationalconcern. Journal of the American Medical Association, 280, 1673-1678.

Wicklund, R. A., & Brehm, J. W. (1976). Perspectives on cognitivedissonance. Hillsdale, NJ: Erlbaum.

Worden, J. K., Flynn, B. S., Solomon, L. J., Seeker-Walker, R. H., Badger,G. J., & Carpenter, J. H. (19%). Using mass media to prevent cigarettesmoking among adolescent girls. Health Education Quarterly, 23, 453-468.