Smoking Cues in Anti-Tobacco PSAs
Tobacco use is the greatest preventable cause of death worldwide, yet about 1 in 5 adults in the U.S. continue to smoke. Message campaigns, such as public service announcements (PSAs), designed to increase the awareness of the health harms of tobacco use, have shown initial promising findings. This has led to exploration of the features of anti-tobacco PSAs that are most and least persuasive, including types of appeals and PSA components or features. Anti-smoking mass media campaigns often present smoking-related cues (e.g., seeing someone smoke) to illustrate the negative consequences of smoking. However, these smoking-related "cues" can elicit strong smoking urges, providing a key motivation for continued smoking in the face of a desire to quit. Our preliminary research shows that when chronic smokers view anti-smoking PSAs that include smoking cues, their urges to smoke increase significantly – if the central argument of the PSA is weak. To extend this research, we propose to study the effects of smoking cues in anti-tobacco PSAs on smoking urges, message processing, persuasion, and smoking behavior in a sample of 300 chronic smoking adults. This laboratory-based study will use a 3 (smoking cue: no cue/peripheral cue/central cue) x 2 (low/high argument strength) factorial (between-subject) design. PSAs in different conditions will be balanced for multiple explicit and implicit ad features. The primary outcomes are: 1) smoking urges, 2) message processing (recall, perceived ad effectiveness), and 3) persuasion (attitudes, self-efficacy, intentions). In addition, participants will be monitored for physiological arousal (heart rate, skin conductance) elicited during cue presentation. Following the session, we will permit participants to smoke in the smoking research laboratory and assess latency and consumption, thus serving as a behavioral measure of PSA smoking cue effects. Results from this study may importantly provide empirical support for better development of anti-smoking PSAs, and to support restrictions on tobacco industry advertising.
Back to top
The following research was conducted under CECCR I and serves as the foundation upon which we have built the next phase of research for CECCR II, Smoking Cues in Anti-tobacco PSAs.
Anti-Tobacco PSA Coding. An ad archive created by the Centers for Disease Control and Prevention (CDC) media resource center was used as the total population of anti-smoking ads. It contained 589 antismoking PSAs broadcast between 1998 and 2002. Trained research technicians viewed and coded each of these PSAs on the following six criteria (Intercoder reliability (kappa =0.89, p<0.001): (a) 30 second, (b) English-speaking, (c) targeting adult smokers, (d) not targeting adolescents, (e) aiming at cessation and/or treatment seeking, and (f) not primarily about second-hand smoking. These ads formed the majority of ads that targeted adult smokers, and resulted in a set of 99 PSAs for further coding and evaluation.
Message Sensation Value. Trained raters viewed the 99 PSAs and rated each for MSV features using a scoring template (visual range 0-10, audio range 0-5, content range 0-5) based on work by Morgan and colleagues. Counts of cuts and edits were converted to low (0-6), moderate (7-14), and high (> 15) before MSV summary scores were created. Inter-rater reliability of MSV scoring between pairs of raters was high (Kendall’s Tau = 0.906, p < 0.001). MSV is an important control valuable across conditions in the proposed study to ensure effects on psychological, and cognitive processing are due to factors unrelated to the formatting features of the ads.
Argument Strength. First, two trained raters viewed the PSAs and transcribed the explicit and implicit messages of each PSA. Next, two different raters viewed the PSAs while reviewing the corresponding messages and edited these messages to generate a single statement reflecting the central arguments of each PSA (e.g.: ‘If the health harms of smoking are not enough to get you to quit, consider quitting for your children and those you love’; and, ‘Although you may think smoking helps you cope, if you don’t soon quit you will eventually die’) (Kappa = .90). To test the adequacy of this procedure, 20 of the 99 arguments were randomly selected and judged by untrained viewers (N=24). They watched the PSAs and reported on the arguments made within the ad compared to the ones extracted. On average, the participants identified the “correct” extracted argument as the best summary of the ad’s content 96.4% of the time. The “correct” arguments were also judged to be much more accurate representations of the ads (M = 91.9%) than the linguistically “close” (M = 20.1%) and “far” (M = 13.1%) options (p < .05 for all PSAs). These results indicated a high level of agreement between the researchers and untrained participants with respect to the central arguments of the PSAs. This procedure allows the separation of a PSA’s arguments against smoking from the way those arguments are presented.
Next, we conducted a shopping mall intercept survey of 300 current smokers to collect ratings of the transcribed central arguments. Participants were presented with 12 of the 99 central arguments in a balanced design such that each argument was rated 36 times. After being presented with each argument (in text-only form), participants were asked if the statement: a) is a reason for quitting smoking that is believable; b) is a reason for quitting smoking that is convincing; c) is a reason for quitting smoking that is new to me; d) is a reason for quitting smoking that applies to me; e) gives a reason for quitting smoking that is important to me; f) put thoughts in my mind about quitting smoking; g) put thoughts in my mind about wanting to continue smoking; h) helped me feel confident about quitting smoking; i) would help my friends quit smoking; j) made me want to quit smoking; and k) is a strong or weak reason. Participants responded to each item using a 5-point scale (1=strongly disagree to 5=strongly agree). Consistent with thought-listing scoring procedures, we subtracted the thoughts about continuing to smoke score (item g) from the thoughts about quitting smoking score (item f) to create a thoughts item. Argument strength scores were created by summing the responses to the above items. An overall argument strength score was then created for each PSA by taking the mean of the 36 individual scores for that PSA.
Death Appeals. By some estimates, nearly 50% of anti-smoking ads discuss long term consequences of smoking such as cancer, heart disease, and death. Priming thoughts of death and cancer may not have beneficial consequences in health behavioral choices. A subset of 32 of the 99 PSAs described above was selected for coding for death appeals and for subsequent testing. An ad was coded as using a “death appeal” if there was a visual and/or verbal mention of death or death-related illness (e.g. cancer, funeral imagery, “kills,” etc.) in connection to smoking. Non-life-threatening illnesses were ignored (e.g. asthma), as well as implicit threats, such as ads that focused on health or longevity benefits. Ads were reliably coded (Krippendorff’s a = 0.87) and 16 were found to have death appeals and 16 not. A sample of 383 adult smokers viewed 4 randomly selected ads from the set of 32 evaluating them for perceived effectiveness, transportation, vulnerability to health risks, and fear while viewing the ad. Ads were also separately rated for their argument strength and coded for message sensation value. Ads with death appeals produce significantly (a) greater perceived effectiveness, (b) more fear, (b) higher vulnerability to health risks, and (d) higher levels of transportation (all significant at p < .001, ?2 ranging from .10 to .17).
Ads high in MSV that contained death appeals also produced higher ratings of vulnerability, transportation, and fear. However, controls for argument strength (from a separate sample of raters) accounted for all the differences between death and non-death ads suggesting that death appeals are ones that participants see as strong arguments against smoking. While this finding does not suggest that all the strong arguments used in PSAs will be death appeals, it does suggest a caution in testing strong and weak arguments. Designs need to control or balance death appeals in PSAs used across experimental conditions.
Visual Verbal Redundancy. A subset of the 99 PSAs selected for intensive study was evaluated for the degree of redundancy between the visual channel and the verbal channel of information. Following McGuire’s elaboration of the persuasion process, we assumed that redundant visual and verbal streams would enhance encoding of information in the message leading to greater effectiveness. A rating procedure was developed, tested, and refined in three studies. An initial 12 –item scale was used by 24 undergraduate raters to successfully discriminate anti-smoking PSAs chosen a priori to be redundant and non-redundant by a panel of experts. Four of the initial 12 items were selected for the final rating scale (repeat, consistent, say the same thing, and in sync). These four distinguish between the redundant and non-redundant ads sharply (p < .001) and are highly reliable (alpha > .90). Study 2 used a sample of 315 adult smokers and 32 anti-smoking PSAs. Each smoker rated a set of 5 randomly selected PSAs on several measures including visual-verbal redundancy. Correlations among the ads (N=32) showed that visual-verbal redundancy (VVR) showed discriminate validity in that it was unrelated to message sensation value (r =0.08, ns) and to argument strength (r= .17, ns).
The ratings of VVR were correlated with a measure of transportation (r= .57, p < .001) indicating convergent validity. Study 3 used a different set of 32 anti-smoking PSAs. A sample of 427 adult smokers each saw and rated 4 ads. Smokers rated the effectiveness of the ads higher when the ads were higher in VVR (ß= .58, p< .001) and when the ads employed stronger arguments (ß= .54, p< .05).
The VVR rating scale will be adapted to compare the degree of redundancy between visual smoking cues in PSAs and the verbal information the ad presents. These ratings will allow a validation of the distinction between peripheral smoking cues (non-redundant with the central argument) and central smoking cues (redundant with the central argument).
Smoking Cues. Next the 99 PSAs were coded for the presence of different types of smoking cues (there was 100% correspondence for ratings of cue presence). Overall, 39.4% of these ads contained some type of smoking scenes (N = 39). Fifty nine percent of these smoking-cue ads presented the actual smoking behavior (N = 23), 28.2% showed pure presence of smoking-related material (N = 11), and 12.8% presented holding and handling of cigarettes (N = 5). To check whether the prevalence of smoking cues in this set of ads can be generalized to the other ads, another set of 99 PSAs was randomly selected from the whole set of ads (excluding the above 99 ads). Smoking cue was present in 46.5% of the ads (including presence, holding and smoking behavior, N = 46). 41.3% of these smoking-cue ads presented the actual smoking behavior (N = 19), 41.3% showed pure presence of smoking-related material (N = 19), and 17.4% presented holding and handling of cigarette (N = 8). Thus smoking scenes were relatively common in antismoking PSAs. A smoking cue is considered as a peripheral cue under the condition that it is not completely enmeshed with the central argument. Our previous coding indicates this non-redundant cue category often covers the majority of the ads, and may have importance related to the effect cues have on urges.
Effects of MSV and Argument Strength: Four groups of PSAs were then selected based on the coding above: 1) high MSV-high argument strength (AS), 2) high MSV-low AS, 3) low MSV-high AS, 4) low MSV-low AS (note, as this study was completed before the cue pilot, these ads were not matched for presence of smoking cues or death appeals). PSAs exceeding one standard deviation from the mean on each of the two dimensions were selected for use in the current study; 16 PSAs met this criterion (4 in each group). These PSAs were presented to 200 current smokers in a 2 (MSV) by 2 (argument strength) factorial design. Physiological responses, specifically heart rate, skin conductance, zygomaticus major and corrugator supercilli were assessed while viewing the PSAs and cognitive measures were administered following viewing.
Message Sensation Value (MSV). High MSV PSAs, compared to low MSV PSAs produced significantly greater increases in corrugator (frowning) activity (p = .012). In addition, high MSV PSAs were associated with less cognitive processing (based on a self-report measure asking how much the PSAs made them think about quitting) (p=.01), and were rated less positively (p=.006) and more negatively (p=.001) than low MSV PSAs.
Argument Strength (AS). High AS PSAs produced greater increases in heart rate (p = .053) and skin conductance (p=.02), compared to low AS PSAs. Sensory, but not cognitive processing, was significantly higher for participants viewing the high AS PSAs, compared to low AS (p=.04). In addition, an MSV by AS interaction (p=.001) indicated that recall was worse for PSAs low in AS when MSV was low; there was no AS effect when MSV was high. There were no effects of MSV and AS conditions on quitting intentions.
Effects of Smoking Cues: The study had a 2 (AS: high versus low, between-subject factor) by 2 (smoking cue: presence versus absence, within-subject factor) mixed design. Ninety-six participants were randomly assigned to 1 of the 2 argument strength (AS) conditions and each participant viewed 6 PSAs: 3 of which had smoking cues and 3 did not. No-cue ads were presented before smoking-cue ads so as to reduce the possibility of contamination. The smoking cues were all non-redundant with the central argument of the PSA (i.e., peripheral cues). Cue and no-cue ads were chosen to have similar levels of MSV. As in study 1 (above), physiological recording was performed during PSA viewing. Smoking urge was measured before and after each set of PSAs. The repeated measures analysis revealed a significant quadratic main effect of smoking cue (p < .05) and a marginally significant linear interaction between smoking cue and AS (p = .06) on smoking urge. As shown below, smoking urge declined or remained low until the final set of 3 PSAs which contained smoking cues. For participants in the strong argument condition, the cues had no effect on urge. However, for those in the weak argument condition, the cue PSAs led to a sharp increase in smoking urge.
In addition, there were significant cue by AS interactions for heart rate (p < .02) and corrugator response (index of negative affect) (p = .004). Most importantly, as also shown, we observed a significant interaction between smoking cues and AS on ad effectiveness (p=.04) (86). Effects of smoking cues on measures of message processing and persuasion could not be properly evaluated due to the within subject design (this will be addressed in the proposed study by the between-subject design). However, the AS manipulation produced several positive effects, including: increased recall (p=.05), positive attitudes toward quitting (p=.05), and increased self-efficacy (p=.005).
Summary of Study 1 and 2. We draw the following tentative conclusions from these data: (a) smoking cues within anti-tobacco PSAs produce significantly increased smoking urges and reductions in perceived ad effectiveness when the central arguments are weak; (b) argument strength is positively associated with attitudes toward quitting and quitting self-efficacy (when cues are balanced across conditions, as in Study 2); and (c) MSV is negatively associated with beliefs about the negative consequences of quitting and positively associated with a physiological measure of negative affect. It should be noted that we are not surprised that these effects did not translate into changes in quitting intentions following only a single exposure. However, our data do suggest that PSA features such as argument strength do influence attitude and self-efficacy measures that are known to predict quitting intentions and behavior. The within-subject design of the cue study (study 2) precludes proper evaluation of effects on these measures (because all participants viewed no cue and cue ads). Thus, the proposed study will manipulate both cue presentation and AS in a between-subject design, providing a stronger test of the effects of smoking cues cue type, in PSAs on smoking urges, message processing and persuasion. MSV and death appeals will be balanced across conditions.
Effect of Marijuana Scenes in Anti-Marijuana PSAs: Using coding procedures similar to those described above, we evaluated a series of anti-marijuana PSAs and compared those that included marijuana smoking scenes (n=15) to those that did not (n=36). Ads were balanced for MSV and AS. There was a significant main effect of the presence vs. absence of marijuana scenes on all 3 measures of ad effectiveness; adolescents perceived ads with cues as less effective (p=.02), liked these ads less (p=.005), and had fewer positive thoughts about the ads (p=.008); however, the simple presence of marijuana in the PSAs (as opposed to acts of smoking it) had no significant effects on these measures. These data provide further support for the hypothesized effects of smoking cues on PSA effectiveness.
Additional Data on Smoking-Related Ads: The study team has also conducted research on the effects of tobacco-related advertisements. An initial investigation of smokers' (n=200) health-related beliefs about a novel potential reduced exposure produce (PREP; Quest cigarettes) reported that smokers erroneously believed the produce to be lower in tar, healthier, and less likely to cause cancer, compared to their regular brand cigarette . In a subsequent study, 500 Quest-naive, current smokers were randomized to view one of three digitally-edited versions of the Quest advertisement to investigate if smokers' beliefs could be manipulated with how a product is marketed . Those viewing an advertisement with no text were more likely to be correct about the health risks, importantly demonstrating that cigarette and PREP advertisements can mislead smokers about health risks, as well as illustrating the need to better understand what makes an advertisement effective.
Back to top
Current Research Aims
Tobacco use is the greatest preventable cause of death worldwide. Despite widespread awareness of the health harms of tobacco use, about 1 in 5 adults in the U.S. continue to smoke. Nicotine, the addictive chemical in tobacco has direct rewarding effects. Moreover, through a behavioral conditioning process, stimuli associated with smoking become rewarding as well (e.g., the sight or smell of cigarettes). These smoking-related “cues” can elicit strong smoking urges, providing a key motivation for continued smoking in the face of a desire to quit.
Anti-smoking mass media campaigns, such as public service announcements (PSAs), often present smoking-related cues to illustrate the negative consequences of smoking (e.g., a person smoking or a pack of cigarettes). Research in our CECCR shows that when chronic smokers view anti-smoking PSAs that include smoking cues, their urges to smoke increase significantly – if the central argument of the PSA is weak. To extend this research, we propose to study the effects of smoking cues in anti-tobacco PSAs on smoking urges, message processing, persuasion, and smoking behavior in a sample of 300 chronic smoking adults. This larger study will use a 3 (smoking cue: no cue; peripheral cue; central cue) x 2 (low/high argument strength) factorial (between-subject) design. PSAs in different conditions will be balanced for multiple explicit and implicit ad features. The primary outcomes are smoking urges, message processing (recall, perceived ad effectiveness), and persuasion (attitudes, self-efficacy, intentions). In addition, participants will be monitored for physiological arousal (heart rate, skin conductance) during the PSA viewing. Following the session, we will permit participants to smoke in the smoking research laboratory and assess latency and consumption; this will serve as a behavioral measure of PSA smoking cue effects. We will offer them the opportunity to sign up for a smoking cessation program as an additional behavioral measure.
Our Primary aim in the study is to determine the effects of smoking cues presented in a series of anti-tobacco PSAs on smoking urges, message processing, and persuasion.
Our hypotheses are as follows:
H1a (cue main effect): Compared to smokers viewing PSAs which do not include smoking cues, those who view PSAs containing smoking cues will report greater increases in smoking urges.
H1b (mediation effects): Cue-induced smoking urges will, in turn, reduce message processing (recall, ad effectiveness), and have a negative influence on persuasion (i.e., less positive attitudes toward quitting and lower quitting self-efficacy). These effects, may contribute independently to lower intentions to quit.
H1c (cue by argument strength interaction): The negative effects of smoking cues in PSAs will be significantly more pronounced for PSAs with weaker arguments.
H1d (cue by argument strength interaction): The negative effects of smoking cues in PSAs will be significantly more pronounced for PSAs with peripheral cues and those with central cues and weak arguments, compared to PSAs with central cues and strong arguments.
H1e (cue by argument strength interaction): The control (no cue) PSAs with strong arguments will yield more favorable effects than the PSAs with strong arguments and central smoking cues.
Our secondary aim is to determine the effects of smoking cues in anti-tobacco PSAs on physiological measures of cue-reactivity and on smoking behavior.
Our hypotheses under the secondary aim are as follows:
H2a: Compared to smokers viewing PSAs which do not include smoking cues, those who view PSAs containing smoking cues will exhibit increases in heart rate and skin conductance (measures of general arousal). These effects may be more pronounced in ads with weaker arguments.
H2b: Following viewing of the PSAs, smokers exposed to ads containing smoking cues will have a shorter latency to smoking and smoke more intensely, than those who viewed PSAs that do not contain cues.
H2c: The effects of smoking cues on subsequent smoking latency and intensity will be stronger for those viewing ads with weaker arguments.
H2d: The effects of smoking cues on urge and physiological measures will be greater for those PSAs with peripheral smoking cues, compared to those with central cues. In addition to these aims, we will test a hypothesized conceptual model that integrates features of the Elaboration Likelihood Model and the Integrative Model of Behavior. Further, we will explore the effects of individual difference variables, such as nicotine dependence level and gender on hypothesized cue effects. Our paradigm will also permit us to link smoking cue exposure to physiological responding by examining responses that are "time-locked" to cue presentation within the PSAs. The finding that smoking cues in anti-tobacco PSAs increase smoking urges and reduce the persuasive effects of these interventions could have a significant impact on future PSA design as well as on policy.
Back to top
Overview of Study Design
We will use a 3 (smoking cue: no cue; peripheral cue; central cue) x 2 (low vs. high argument strength [AS]) between-subject factorial study design. 150 male and 150 female smokers will be randomly assigned to 1 of these 6 conditions, and will view a series of 4 PSAs. Because of potential effects of nicotine dependence on cue-reactivity, we will stratify the randomization by nicotine dependence level (<4 vs. 4 > on the Fagerstrom Test for Nicotine Dependence). PSA sets will be balanced for MSV and death appeals, and the order of ads within each condition will be randomized to control for the possible order effect. The primary outcomes include smoking urges, message processing and persuasion. Secondary outcomes include physiological reactivity and smoking behavior immediately following the session. Finally, participants will be recontacted one week after the session to reevaluate message effects for recall, intention to seek treatment, and smoking behavior.
Back to top