|Home | About | Journals | Submit | Contact Us | Français|
Exposure to smoking in movies has been linked to adolescent smoking uptake. However, beyond linking amount of exposure to smoking in movies with adolescent smoking, whether the way that smoking is portrayed in movies matters for influencing adolescent smoking has not been investigated. This study experimentally examined how motivation for smoking depicted in movies affects self-reported future smoking risk (a composite measure with items that assess smoking refusal self-efficacy and smoking intentions) among early adolescents.
A randomized laboratory experiment was used. Adolescents were exposed to movie scenes depicting one of three movie smoking motives: social smoking motive (characters smoked to facilitate social interaction); relaxation smoking motive (characters smoked to relax); or no smoking motive (characters smoked with no apparent motive, i.e., in neutral contexts and/or with neutral affect). Responses to these movie scenes were contrasted (within subjects) to participants’ responses to control movie scenes in which no smoking was present; these control scenes matched to the smoking scenes with the same characters in similar situations but where no smoking was present. A total of 358 adolescents, aged 11–14 years, participated.
Compared with participants exposed to movie scenes depicting characters smoking with no clear motive, adolescents exposed to movie scenes depicting characters smoking for social motives and adolescents exposed to movie scenes depicting characters smoking for relaxation motives had significantly greater chances of having increases in their future smoking risk.
Exposure to movies that portray smoking motives places adolescents at particular risk for future smoking.
Adolescents’ exposure to smoking in movies is a significant global health concern (World Health Organization, 2009) Adolescents are exposed to hundreds of smoking impressions in movies annually (Anderson et al., 2009; Sargent et al., 2007) and exposure to smoking in movies (assessed using survey methods) has been linked with adolescent smoking in both cross sectional and prospective studies (Dalton et al., 2003, 2008; Hanewinkel and Sargent, 2007; Jackson et al., 2007; Sargent et al., 2005; Thrasher et al., 2009; Titus-Ernstoff et al., 2008).
Beyond linking amount of exposure to smoking in movies with adolescent smoking, it is important to investigate whether the way that smoking is portrayed in movies influences adolescent smoking. Movie characters who smoke convey the message that smoking serves particular motives (Hazan et al., 1994; Stockwell and Glantz, 1997; Worth et al., 2007). Smoking to manage negative affect and smoking to facilitate social interaction are the most common motives portrayed in movies (Worth et al., 2007) and managing negative affect and facilitating social interaction are motives that underlie adolescent smoking more generally (Guo et al., 2010; Johnson et al., 2003; Piko et al., 2007; Wills et al., 1999). Cognitive social learning theory (Bandura, 2006) suggests that adolescents learn about these smoking motives from observing these movie portrayals. However, almost no research has investigated smoking motives in movies influence adolescent smoking: a small correlational study (Shadel et al., 2010) found that movies that portrayed relaxation smoking motives were positively associated with increased desire to smoke among young adolescents, whereas movies that portrayed social or rebellious smoking motives had no association with desire to smoke.
The current experiment investigated whether exposure to movie smoking that is portrayed as helping characters to facilitate social interaction or to facilitate relaxation affects future smoking risk differently than movie smoking that is portrayed as having no motive. Our measure of future smoking risk, composed of items tapping smoking intentions and smoking resistance self-efficacy, has been found to be a potent predictor adolescents’ future smoking (Choi et al., 2001; Gilpin et al., 2005; Jackson, 1998; Pierce et al., 1996). Smoking intentions and smoking refusal self-efficacy are independent powerful predictors of adolescent smoking onset (e.g., Hiemstra et al., 2010; Wakefield et al., 2004). As such, future smoking risk assessed with this composite measure is a logical choice of an outcome in an experimental laboratory study with adolescents. We expected that adolescents exposed to movies that emphasized social smoking motives and relaxation smoking motives would experience higher future smoking risk compared to when no smoking motive was present. Finding support for this hypothesis in an experimental context would provide an important first step in understanding the role of movie smoking motives in adolescent smoking.
The study utilized a three-group, between-subjects design and took place across two sessions, separated by about one week (see Figure 1 for study schematic). Adolescents were randomly assigned to one of three movie smoking motive experimental conditions (between subjects): social smoking motive (characters smoked to facilitate social interaction); relaxation smoking motive (characters smoked to relax); or no smoking motive (characters smoked with no clear motive). Participants viewed the experimental movie smoking scenes during session 2. During session 1, participants viewed matched-control scenes (within-subjects control). Control scenes were matched to experimental scenes in that they portrayed the same actors under similar circumstances but with no smoking. The main dependent variable was the difference between participants’ future smoking risk score following exposure to the experimental scenes (viewed during session 2) versus control scenes (viewed during session 1).
This study was approved and monitored by the Human Subjects Protection Committee at RAND. The study used an authorized deception and a two-stage informed consent procedure to balance the internal validity and ethical integrity of the research. During the stage one consent process that occurred at session 1, participants and their parents were told about the general parameters of the study (e.g., that it involved assessing adolescents’ reactions to movies; the study involved minimal risk; payment would be $15 for session 1 and $25 for session 2), and were told that there were aspects of the study that they could not be told about because that knowledge could affect the study results. They were told that they would be provided with all information about the study at the second session. Their stage one consent/assent indicated agreement to participate in the study without full knowledge of the study details. At the beginning of session 2, only the children’s parents participated in the stage two consent process whereby they (parents only) were fully informed about the study (i.e., that it was a study of movie smoking and that their children would be briefly exposed to scenes of movie smoking during Session 2). Parents had to provide written informed consent for their children to complete the study procedures during session 2. No parents refused to allow their child to complete the procedures at the second session. At the end of session 2, participants were fully debriefed.
To be included in the study, participants had to be between the ages of 11 and 14, have no physical or psychiatric problem that would interfere with completing the study (based on parent report), have written parental consent (both consent stages), and assent to their own participation (stage 1 consent). Adolescents were enrolled irrespective of their experiences with smoking. A total of 570 adolescents were screened, of whom 547 (96%) were eligible and chose to participate. The vast majority of individuals who were excluded were either too young (<11) or too old (>14). A total of 419 adolescents attended Session 1, and 358 of those who attended Session 1 also attended session 2. This study reports on the 358 participants who had complete data for both sessions. Descriptive information on the sample (by experimental condition) is provided in Table 1.
We conducted a pilot study to develop the movie scenes that were used as the control and experimental stimuli in this study (see Shadel et al., 2010). Smoking and nonsmoking scenes were selected from 28 wide release movies (rated PG to R). A total of 32 smoking scenes (8 per motive category) were selected and initially sorted by study team consensus into four smoking motives categories (see Worth et al., 2007; categorizations used for the experiment were determined by examining participant data; see below): a) characters smoking to relax; b) characters smoking to facilitate social interaction; c) characters smoking to appear rebellious; or d) characters smoking where no motive was apparent. Next, 32 non-smoking scenes were selected from the same movies where the same characters that appeared in the smoking scenes were present and where the scene setting was similar to that of the identified smoking scene. The smoking and nonsmoking scenes were trimmed into two minute segments that provided context and character development. None of the scenes contained sexual, profane, or violent content.
Participants (n = 77; ages 11–14) in the pilot study viewed scenes and rated them on variables that determine individuals’ responsiveness to advertising (e.g., how the scene makes them feel; how interesting was the scene; how much the scene makes them think; how much they liked the scene; how realistic the scene was; and how much they wanted to see the movie from which the scene was drawn; see Moore and Lutz, 2000). They then sorted each smoking movie scene into one of the four smoking motive categories; they were provided with definitions of each motive to facilitate this process.
Three scenes that were most consistently identified as social smoking scenes (67% agreement) and three that were most consistently identified as relaxation smoking scenes (74% agreement) were used as experimental stimuli for the social smoking motive condition and relaxation smoking motive condition, respectively. Three scenes that were most highly identified as having no clear smoking motive were placed into the no smoking motive category. These scenes did not differ with respect to the advertising response variables that were measured (Moore and Lutz, 2000). (For this experiment, we omitted the scenes in which a desire to appear rebellious was seen as the main motive for the characters’ smoking because these scenes portrayed only Caucasian males smoking cigarettes and because smoking to appear rebellious is much less common compared to the other smoking motives (Worth et al., 2007)).
Each three-scene dose within a smoking motive condition was approximately three minutes and 30 seconds long, and smoking appeared in each scene dose condition approximately 54% of the time. Total time for the three scenes within the non-smoking control condition dose was approximately three minutes and 30 seconds. The smoking and non-smoking scenes did not differ on the advertising variables described above and a diversity of actors (i.e., gender and race) appeared in all scenes. Thus, the smoking motive conditions differed only by the type of smoking motive presented, and the smoking and nonsmoking scenes differed only by the presence of smoking. Scenes were presented in random orders within condition during the experiment.
All of the measures described below were administered at the start of Session 1 and evaluated as potential covariates in the analyses (descriptive information for each measure is provided in Table 1). Because the distributions of these baseline covariates were skewed, we dichotomized them for purposes of analysis in the regression models.
The future smoking risk measure was administered at baseline (as a potential covariate) and after exposure to the dose of non-smoking (session 1) and smoking scenes (session 2).
Three items, drawn from a scale used in studies predicting adolescent smoking from smoking attitudes (Stacy et al., 1994), were administered. Subjects responded to the stem “Smoking is…” using scales anchored by the following bipolar adjectives: very good(1)–very bad(5), very clean(1)–very dirty(5), and very nice(1)–very awful(5). Responses to these items were summed such that a higher score indicated a more negative attitude toward smoking (α = 0.74).
The following item, drawn from the Legacy Media Tracking Survey (http://www.americanlegacy.org/167.htm), was used: “Out of every 10 people your age, how many do you think smoke?” Higher responses to this item are associated with adolescent smoking status (Sussman et al., 1988).
The following two items were used (Romer and Jamieson, 2001): “If you were a smoker, would your smoking be…” and “If you were a smoker, would smoking every day be…”. Responses to both items are made on a 4-point scale (1 = not at all risky for my health; 4 = very risky for my health) and summed to produce an overall perceived smoking risk score (α = .70). Higher scores indicate a belief that smoking is risky to one’s health. These two items have significantly predicted smoking in adolescents (Romer and Jamieson, 2001).
Participants used a 4-point scale (1 = I would definitely smoke to 4 = I would definitely not smoke) to rate their confidence to resist smoking in the following situations (a) your best friend is smoking; (b) your date is smoking; (c) you are bored at a party; and (d) all your friends at a party are smoking. Responses to these items were summed to form a measure of resistance self-efficacy on which higher scores reflect greater confidence to resist smoking (α = .93). This measure has been shown to predict smoking in adolescents (Tucker et al., 2002).
Smoking outcome expectancies were measured using five items from an established measure (Copeland et al., 2007). Two items were about positive outcomes (“Smoking at parties is fun”; “Smokers are more fun to be around than nonsmokers”) and three items were about negative outcomes (“Smoking looks dumb”; Smokers have health problems”; Once people start smoking, it is hard for them to stop”). Responses were made on a five point scale (1 = strongly agree to 5 = strongly disagree), the negative outcomes items were reverse scored, and responses were summed such that higher scores reflect more positive smoking outcome expectancies (α = .60).
Future smoking risk was assessed using a 3-item scale adapted from a scale developed by Choi and colleagues (2001) and shown to be predictive of future adolescent smoking: “Do you think you will try a cigarette anytime soon?”, “Do you think you will smoke a cigarette anytime in the next year?”; and “If one of your best friends offered you a cigarette, would you smoke it?”. Responses were made on a 1 (Definitely Not) to 10 (Definitely Yes) scale and summed to produce a measure of future smoking risk on which higher scores indicate greater risk of future smoking (α = .95).
Previous exposure to movie smoking was assessed with an item adapted from the Center for Disease Control’s on-line database (see http://apps.nccd.cdc.gov/QIT/QuickSearch.aspx): “During the last 30 days, about how often have you seen someone smoking in movies?” To respond to this item, participants chose from the following options: never, hardly ever, some of the time, and most of the time.
Smoking status was ascertained by asking participants whether they “…had ever smoked, even a puff in your [their] life?”. Participants who responded “no” were classified as never smokers (n = 329); participants who responded “yes” were classified as ever smokers (n = 29) (see Table 1).
Adolescents were recruited using print advertising that contained no information about smoking, cigarette advertising, or smoking in movies to guard against recruitment biases. Parents of interested participants telephoned the study center to complete a brief telephone screening. Individuals who met the inclusion/exclusion criteria were scheduled for session 1.
Participants completed the study in a small group setting. Group sessions were held in conference rooms that were arranged with participants facing a projection screen on which the movie scenes were played. During session 1, participants first completed the stage 1 informed consent/assent procedures with their parents present (see Figure 1). After parents provided written informed consent and the child provided written informed assent, parents were excused to a waiting area. Next, participants completed a baseline questionnaire packet. This packet consisted both of target measures related to smoking and a number of filler measures that were similar to the smoking measures but assessed other behaviors (e.g., consumption of alcohol and fast food). Filler items were included to disguise the true focus of the study from participants. Participants were then randomized to one of the experimental conditions and exposed to their assigned non-smoking control movie scenes. After exposure to the non-smoking control scenes, they completed the dependent measure (future smoking risk) along with filler items. They were provided with a $15 gift card for completing session 1. Participants returned for session 2 approximately one week later. During session 2, participants’ parents completed the stage 2 consent process, after which their children were exposed to their assigned smoking motive movie scenes. After exposure to the smoking scenes, adolescents completed the dependent measure and filler items. Following completion of these procedures, participants were fully debriefed and provided with a 45-minute interactive media literacy intervention focusing on cigarette advertising and movie smoking (Primack et al., 2009). They were provided with a $25 gift card for completing session 2. Transportation costs for both sessions were reimbursed to parents.
Descriptive information by condition information appears in Table 1 (there were no significant differences between participants who completed the study and those who dropped-out on any of these variables). Randomization was generally successful in ensuring parity across experimental conditions; however, there was a greater proportion of ever smokers (ever smoked, even a puff) vs. never smokers (never smoked, even a puff) in the no smoking motive condition compared with the social and relaxation smoking motive conditions. Thus, we included smoking status as a covariate in all multivariable analyses (there was no interaction between smoking status and the experimental conditions).
We subtracted participants’ self-reported future smoking risk score following exposure to the matched-control non-smoking scenes from their future smoking risk score following exposure to the experimental smoking scenes to use as the dependent measure. However, responses to the dependent measure, future smoking risk, were restricted in all conditions. Preliminary analyses suggested that a zero-inflated Poisson regression model would provide the most appropriate fit to the data, given the relatively low frequency of future smoking risk increases (i.e., 87% of participants had a decrease or no change). Using the dependent variable future smoking risk change between the experimental and control conditions, then, we estimated a multivariable zero-inflated Poisson regression model to test the effect of exposure to smoking motives in movies on adolescents’ future smoking risk. This two part model first tests for the effect of movie smoking motive exposure on whether there is an increase in smoking risk (i.e., a logistic regression of the likelihood of increase), and then conditional on having an increase in risk, it tests for the effect of movie smoking motive exposure on the size of the risk (i.e., Poisson regression). Preliminary bivariable analyses with the variables in Table 1 revealed that baseline smoking attitudes, smoking resistance self-efficacy, smoking outcome expectancies, future smoking risk, and race were all associated with the dependent measure. When these variables were entered into a multivariable zero-inflated Poisson regression model as covariates, only smoking resistance self-efficacy, smoking outcome expectancies, and baseline future smoking risk were associated with the dependent measure. Thus, we retained this reduced set of covariates in the final model. The final model results are presented in Table 2. There were significant effects of movie smoking motives on adolescents’ future smoking risk. Compared with participants exposed to movie scenes with no smoking motive depicted, participants exposed to movie scenes that depicted social smoking motives were significantly more likely to experience an increase in their future smoking risk (log-odds = 1.28 [approximate odds ratio =3.6], p=0.016) but participants exposed to movie scenes that depicted relaxation smoking motives did not show a significant likelihood of an increase in future smoking risk (p = .281). However, among participants with an increase in future smoking risk, being exposed to movie scenes that depicted relaxation smoking motives resulted in a surge in their future smoking risk levels (log-odds = 0.75 [approximate odds ratio = 2.1], p = .011); participants exposed to movie scenes that depicted social smoking motives did not show this surge in their future smoking risk level (p = .101).
This study provides the first experimental evidence that motivated smoking in movies causally affects future smoking risk in young adolescents. As such, it represents an important first step in understanding the effect that smoking motives in movies have on adolescent smoking. Middle school students who were exposed to movie scenes that portrayed smoking as facilitating social interaction or as facilitating relaxation experienced increases in their future smoking risk compared with adolescents who were exposed to movie scenes that portrayed smoking as having no motive. However, there was a difference in how these two different smoking motives affected future smoking risk. Movie scenes depicting social smoking motives were most effective at moving adolescents with no risk of future smoking toward having some level of future smoking risk. In contrast, movie scenes depicting relaxation smoking motives seemed not to affect adolescents who had no future smoking risk; rather, movie scenes with relaxation smoking motives were most effective at moving adolescents with some level of smoking risk toward higher levels of risk. The observed differences in how social and relaxation smoking motives affected smoking risk is consistent with the results of a prior study that found that youth smoking is more likely to be driven by a motivation to regulate affect than to facilitate social interaction (Piko et al., 2007). Additional research that specifies both smoking motives and uses those motives to predict different levels of smoking risk (as in the current study) and actual smoking behavior (e.g., with a prospective study) would further advance this literature.
Our findings suggest that even brief exposures to specific kinds of motivated smoking in movies can affect movement in a key predictor of smoking uptake (Choi et al., 2001; Gilpin et al., 2005; Jackson, 1998; Pierce et al., 1996). Adolescents are exposed to hundreds of smoking impressions in movies annually (Sargent et al., 2007), a nontrivial percentage of which portray socially-motivated smoking and smoking to manage negative affect (Worth et al., 2007). To the extent that cognitive variables like intentions and refusal self-efficacy mediate the movie exposure-smoking behavior relationship (cf., Wills et al., 2008), our results suggest that multiple such exposures over time could incrementally change adolescents’ cognitions about smoking in a way that eventually moves them to try smoking. In this context, it is important to note that non-cognitive factors (i.e., behavioral, contextual) also likely mediate the effects of movie smoking on adolescent smoking (see Vakratis and Ambler, 1999). As such, future research needs to investigate the complex array of cognitive, behavioral, and contextual factors that mediate the effect of movie smoking exposure on adolescent smoking.
Our findings support the notion that adolescents learn about motives for smoking from exposure to motivated smoking in movies. In theory (Bandura, 2006), such learned motives come to regulate smoking behavior (Guo et al., 2010; Johnson et al., 2003; Piko et al., 2007; Wills et al., 1999). This interpretation is consistent with findings that greater exposure to cigarette smoking in movies increases positive expectations about smoking, which then predicts smoking onset (Wills et al., 2008). From a policy or intervention standpoint, then, it may be important to focus on portrayals of motivated smoking in movies as they may be most likely to influence adolescents’ future smoking behavior (see Chapman, 2008).
The findings of the current study extend the findings of our previous correlational study (Shadel et al., 2010). That study found that only relaxation motives were significantly associated with adolescents’ “desire to smoke”. The difference in results between the previous study (Shadel et al., 2010) and the current study may be due to differences in study design (correlational vs. experimental) and dependent variable (desire to smoke vs. future smoking risk). However, the effect size for social smoking motives in the previous study was, in fact, larger than the effect size for relaxation smoking motives, but failed to reach significance due to its larger variance. Taken together, then, the important point from this set of studies seems to be that motivated smoking in movies has a more potent effect on adolescent smoking risk compared to movie smoking with no attendant motive.
The majority of the nearly 100 studies conducted to date have used survey methods to study the relationship between movie smoking and adolescent smoking (e.g., Dalton et al., 2003, Sargent et al., 2005). Experimental studies are extremely rare in this domain (only two studies have been executed with adolescents; see Hanewinkel, 2009; Pechmann and Shih, 1999) but have substantial value because they help to strengthen the causal inferences made about the relationship between movie smoking and adolescent smoking, an urgent need in this domain of inquiry (Nelson, 2010). Regardless of study design, however, none of these studies has examined how smoking motives in movies differentially affect adolescent smoking. Thus, the current study makes a unique contribution by utilizing a strong design (an experiment) to examine an understudied question (the role of smoking motives in movies on adolescents) in an understudied population (middle adolescents).
Despite these strengths, there are limitations to this study. First, the sample of movie scenes was selective. Therefore, these results may not generalize to other instances of movie smoking. Second, the study employed a reactively recruited sample of early adolescents with almost no smoking experience; our findings may not generalize to other populations of adolescents. Finally, although the experimental design brings strength in terms of the causal inferences that can be drawn about the relationship between how smoking is portrayed in movies and adolescent smoking risk, such a design naturally suffers from lack of ecological validity. Future research using representative samples of adolescents and actual smoking initiation and escalation as dependent measures in prospective designs (rather than a cognitively-defined variable such as future smoking risk in an experiment) would further advance knowledge in this important domain of inquiry.
Role of Funding Source. This research was supported by R01DA022496. The NIDA had no further role in study design; in the collection, analysis and interpretation of data; in the writing of the report; or in the decision to submit the paper for publication.
Special thanks are due to Jill Schaefer, Justin Greenfield, Rachel Burns, and Michelle Horner for their invaluable assistance in executing the procedures of this research.
Conflict of Interest. There are no conflicts of interest.
Contributors. William G. Shadel, Steven Martino, Amelia Haviland, and Brian Primack designed the original study. Claude Setodji executed the analyses reported in this paper. All authors contributed to the conceptualization of the analysis and the writing of this paper. All authors contributed to and approve of the final version of the manuscript.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
William G. Shadel, RAND Corporation, 4570 Fifth Avenue, Suite 600, Pittsburgh, PA.
Steven Martino, RAND Corporation, 4570 Fifth Avenue, Suite 600, Pittsburgh, PA.
Claude Setodji, RAND Corporation, 4570 Fifth Avenue, Suite 600, Pittsburgh, PA.
Amelia Haviland, RAND Corporation, 4570 Fifth Avenue, Suite 600, Pittsburgh, PA.
Brian Primack, University of Pittsburgh, 230 McKee Place, suite 600, Pittsburgh, PA 15213.
Deborah Scharf, RAND Corporation, 4570 Fifth Avenue, Suite 600, Pittsburgh, PA.