Long-standing concerns over the negative effects of televised sex and violence have resulted in theorizing about how content might be constructed to minimize negative outcomes. Perhaps the most often cited strategies stem from social cognitive theory (SCT; Bandura, 1986, 2002b), which suggests that depictions of liked characters experiencing either rewards or punishments for their actions will influence viewers to be either more or less likely to perform those behaviors themselves. However, despite this assertion, surprisingly few studies manipulate televised message content to test those predictions. This is particularly troubling as there is an interesting dilemma associated with employing SCT to discourage negative behaviors via media depictions. Specifically, viewers are expected to model the behaviors of liked characters, often a protagonist in a TV series. Yet, in the interest of maintaining entertainment value, protagonists are perhaps the least likely to experience enduring negative consequences for their risky behavior. Thus, even if such characters appear in a particular scene or episode to experience negative consequences, these outcomes are likely to be minimized over the course of a series. If viewers grow to expect protagonists to “bounce back” despite their negative experiences, the persuasive impact of negative behavioral reinforcement may be undermined, potentially resulting in behavior modeling, regardless of the valence of the consequences displayed. The purpose of this research is to empirically test these ideas. Specifically, we draw from SCT and schema theory to assess whether or not televised depictions of risky sexual behaviors that are met with either rewards or punishments in fact alter viewers’ expectations of their own future sexual behaviors.
Despite extensive discussion of the potential for viewers’ social learning of risky health behaviors from TV programming, there has been relatively little effort to test behavioral modeling predictions. That is, it is not clear whether televised depictions of negatively reinforced undesirable behaviors (e.g., unsafe sex), in fact, influence the value viewers attach to those behaviors and their future likelihood of performing them, as social cognitive theory (SCT) would predict. Indeed, we argue that social learning in likely minimal in such contexts due to the programming schemas audiences bring to the viewing experience. In Study 1, the contents of serial TV programming schemas are examined. Results suggest that viewers expect main characters to ultimately survive and thrive, despite the adversity they face. In Study 2, competing predictions from SCT and schema theory are tested. College women (N= 400) were exposed to various portrayals of promiscuous sexual behavior (1-night stands) that were edited to display more or less positive or negative outcomes. Results suggested that, consistent with SCT, those with direct experience with the behavior were not influenced by the portrayals. However, inconsistent with SCT, but consistent with the schema argument, those without direct experience were more likely to expect future participation in the unsafe behavior, regardless of the valence of the reinforcement depicted. Implications for future research on media, SCT, and risky behaviors are discussed.
Sexual portrayals on TV
Content analyses demonstrate that TV programming is highly saturated with sexual content. Kunkel, Eyal, Finnerty, Biely, and Donnerstein (2005) found that in 2004–2005, 70% of network and cable TV programs featured some form of sexual content, up from 56% in 1998 and 64% in 2002. Such sexual depictions generally involve flirtation and talk about sex (Cope-Farrar & Kunkel, 2002; Cope-Farrar et al., 2003; Kunkel, Cope, & Biely, 1999), though up to 45% depict sexual behavior, including sexual intercourse (Kunkel et al., 2005). However, only a small proportion of programs with sexual content—14% in 2005—reference sexual risks, such as unwanted pregnancies or sexually transmitted diseases, or sexual responsibility, such as condom use (Kunkel et al., 2005). Further, Cope-Farrar and Kunkel found that 75% of characters engaging in sexual behaviors were not shown to experience any clear consequences for their actions, and of those who were, 75% were shown to experience positive outcomes.
The result of these arguably unrealistic portrayals of the consequences of risky sexual behavior is that young viewers may begin to see such behaviors as relatively benign. Indeed, cultivation-based research on exposure to sexually oriented TV programming suggests that repeated viewing of such portrayals is associated with greater acceptance of risky attitudes toward sex, overestimation of adolescent peers’ sexual activity, and females’ earlier expectations for their own sexual experience timing (Aubrey, Harrison, Kramer, & Yellin, 2003; Ward, 2002; Ward & Friedman, 2006; see Ward, 2003, for a review). Further, Collins et al. (2004) found that adolescents who viewed more sexual content were more likely to initiate intercourse and report greater sexual experience over a 6-month period.
The concern over the effects of unrealistic portrayals of the consequences of sexual behavior has led to discussion of how such depictions might be altered to promote healthier sexual attitudes and behavior in young viewers. Given the limited discussion of sexual risks and responsibilities in sexual portrayals on TV (e.g., Kunkel et al., 2005), a leading recommendation focuses on increasing depictions of the negative consequences associated with the risky behaviors. From a theoretical standpoint, SCT helps to explain why such depictions might have positive influence on audience behavior.
SCT and sexual initiation
SCT revolves primarily around the functions and processes of vicarious learning (Bandura, 1986, 2002b). That is, by observing others’ behaviors, including those of media figures, one may develop rules to guide one’s own subsequent actions. According to Bandura, observational learning is guided by four processes, which are moderated by observers’ cognitive development and skills. First, attention to certain models and their behavior is affected by source and contextual features, such as attractiveness, relevance, functional need, and affective valence. Second, retention processes focus on the ability to symbolically represent the behavior observed and its consequences, along with any rehearsal of that sequence. Third, production focuses on translating the symbolic representations into action, reproducing the behavior in seemingly appropriate contexts and correcting for any errors based on feedback received. Finally, motivational processes influence whether symbolically represented behaviors are enacted based on the nature or valence (positive or negative) of the reinforcement. Such reinforcement may come from the feedback generated by one’s own behavior, the observed feedback given to others, or internal incentives (e.g., self-standards). As observational learning occurs via symbolic representations, the effects are believed to be potentially long lasting.
SCT, as applied to media contexts (Bandura, 2002b; Stiff, 1986) and TV depictions of sexual behavior specifically (Bandura, 2002a), suggests that to positively affect viewers’ behaviors, the audience must pay attention to attractive models who are displaying relevant behaviors. To the extent positive behaviors are portrayed (e.g., practicing safe sex), the model’s behavior should be positively reinforced (e.g., through the display of positive outcomes, such as greater interest by the sexual partner or the model’s enhanced self-respect). To the extent negative behaviors are portrayed (e.g., practicing unsafe sex), the model’s behavior should be negatively reinforced (e.g., through the display of negative outcomes or experiencing negative affect, such as guilt or regret). Although both portrayals are theoretically possible, the latter may be more likely to appear in TV depictions, given norms of dramatic programming.
Although SCT is often used as a theoretical explanation for the potential effects of TV programming on adolescents’ sexual attitudes and behaviors, most empirical work tests cultivation-based predictions in which quantity of media exposure (both overall and certain genres) serves as the explanatory mechanism for such effects (Ward, 2003). Indeed, some studies that appear to rely on SCT to derive predictions offer tests better suited to cultivation or priming frameworks. For example, in their 6-month longitudinal study of the effects of TV viewing on initiation of sexual intercourse among adolescents, Martino, Collins, Kanouse, Elliott, and Berry (2005) found that those who watched more televised sexual content had fewer negative expectations of the consequences of engaging in sexual intercourse and higher safe sex self-efficacy, both of which were associated with higher likelihood of initiating intercourse. However, although SCT suggests that specific message constructions provide guidelines for behavior, Martino et al. only considered general exposure to sexually oriented TV programs. Further, sexual initiation was measured by whether or not the adolescents had sex but not whether they had practiced safe or unsafe sex. Thus, though the authors concluded that, seemingly consistent with SCT, watching sexual TV content accelerates sexual initiation, they were unable to demonstrate that different depictions differentially affect this outcome, as SCT suggests, nor were they able to demonstrate that the nature of the behavior depicted (which content analyses suggest is unsafe sex) is similar to the nature of the behavior the adolescents’ performed. Ultimately, to test whether SCT is a viable explanation for the effects of mediated depictions of sexual behavior, these issues must be addressed.
This research marks an initial effort at testing the predictions of SCT in that we manipulate the depictions of early sexual timing to reflect positive versus negative reinforcement to determine their relative influence on viewer’s perceptions of the appropriateness of engaging in promiscuous sexual behaviors. We must disclose, however, that we are not fully convinced that SCT-based predictions will be borne out. Indeed, we argue that there is a good chance that they will not be, not because SCT is not a powerful predictor of modeled behavior but rather because its transference to the media context—serial programming in particular—may be problematic. Specifically, we assert that the schemas that viewers have of serial programming may color their perceptions of particular scenes such that negative and positive outcomes will not necessarily have differential impact on viewer attitudes or behaviors.
Schema and serial programming
Schemas are hypothetical cognitive structures that contain and organize information related to a concept or object (Fiske & Taylor, 1991). They vary in their degree of organization and development across topics, both within and between individuals, and are subject to continual change in response to interactions in the social and mediated worlds. In turn, they influence people’s perceptions and responses to their environments. As such, schema approaches to the study of media might reasonably take one of two forms. Media messages might shape schema development on any number of issues, which is the foundation for any social influence or persuasive outcome resulting from media exposure. Conversely, the schemas about various topics that viewers bring to media messages might influence message interpretation (e.g., see Potter, Pashupati, Pekurny, Hoffman, & Davis, 2002). Classic studies of how different audiences decode the same programming to arrive at different meanings support this view (e.g., Liebes & Katz, 1990; Morley, 1980).
However, a third and related way in which schema might influence message interpretation is via the schema of serial TV programming itself. Viewers of such programming, based on their viewing experience, develop ideas of the likely course of events in such programs. These schemas, then, are likely used in the assessment of any one program in ways that impact perceived content and entertainment value. Very scant attention has been paid to this approach, though evidence does support this perspective. Most notably, Liebes and Katz (1990) found that Japanese viewers’ expectations of the televised “home drama” (p. 137) reduced their enjoyment of the U.S. evening drama Dallas as they found the stories to be too complex, violent, and unresolved, and they failed to end on a happy note.
Building on this perspective, we argue that program schemas do not simply influence program enjoyment but also, and perhaps more importantly, what viewers expect might happen to characters in future episodes. These expectations, in turn, might alter their perception of and reactions to program events as they are viewed. For example, let’s assume that TV program schemas contain the notion that main characters are rarely killed. Thus, no matter how dire the circumstances a main character finds him or herself in (e.g., Jack Bauer, the main character in the counterterrorism-focused U.S. series 24, who faces life-threatening situations multiple times per episode), we expect he or she will find some way to escape. This explains why we are so surprised when a main character is killed (e.g., President Palmer, a main character in earlier seasons of 24). Similarly, when a character faces a health threat (e.g., the breast cancer diagnosis of Samantha, one of four female lead characters in the popular HBO series Sex and the City) or an unexpected event (e.g., the unplanned pregnancy of Rachel, one of six single friends in the popular New York City-based sitcom Friends), we do not expect him or her to succumb or for the event to meaningfully alter the course of the character’s life (i.e., to become a dominant theme in later programs) as we recognize these obstacles as dramatic devices to enhance the entertainment value of that particular episode. Although this argument may make intuitive sense, the literature does not document the contents of serial programming schemas or their effects on plot expectations. Thus, we begin by investigating the contents of these schema.
Study 1: Serial drama schema content
We begin with the hypothesis (H1) that TV viewers believe main characters in fictional TV series will experience positive outcomes, despite the adversity they face. Assuming we find support for H1, we then ask RQ1: What explanations do TV viewers offer for expecting positive outcomes?
Procedures and measures
Sixty undergraduates at University of California, Santa Barbara completed a survey in June 2007, after the season finales of the most recent TV season had aired, but before the new season’s premieres. They were first asked to think about the main characters in fictional TV series generally and to indicate on 1 (not at all likely) to 7 (extremely likely) scales how likely the main character is to suffer long-term consequences for their actions, bounce back quickly from adversity, survive seemingly impossible situations, die if stricken with an illness, be killed in a threatening situation, and experience a happy ending. We then presented very brief, one-sentence descriptions of the cliff-hangers of several popular American TV evening programs (Desperate Housewives, Grey’s Anatomy, Lost, Heroes, and CSI). We asked how likely (on 1–7 scales) the featured character was to experience a particular outcome that was either positive (e.g., survive a suicide attempt) or negative (e.g., die from cancer). We then asked in an open-ended form, why they thought that. Each open-ended response was coded for whether or not it included the notion that bad things do not happen to main characters or main characters are necessary to keep the show interesting. An undergraduate coder blind to the purposes of the study coded the open-ended responses. A reliability check with 10% of the surveys coded by a second coder indicated sufficient reliability (Cohen’s κ= .84). The survey concluded with measures of viewing frequency of the featured programs, age (M= 21.77, SD= 1.75), gender (91% female), year in school (69% seniors), and race (67% White).
Analyses and results
One-sample t tests comparing each item’s mean to its scale midpoint suggested support for H1. Respondents were unlikely to believe main characters would suffer long-term negative consequences, would die if stricken with an illness, or would be killed in threatening situations (p < .001). Conversely, respondents believed main characters were likely to bounce back from adversity quickly, to survive seemingly impossible situations, and to experience happy endings (p < .001). The specific story line expectations were also predominantly consistent with H1 (see Table 1 for complete results).
|Item||M (SD)||n||t Value|
|Will suffer long-term negative consequences||2.43*** (1.13)||59||−10.69|
|Will bounce back from adversity quickly||5.71*** (0.95)||59||13.87|
|Will survive seemingly impossible situations||5.91*** (1.04)||59||14.16|
|Will dies if stricken with an illness||2.07*** (1.01)||59||−14.63|
|Will be killed in a threatening situation||1.93*** (0.94)||59||−16.82|
|Will experience a happy ending||6.14*** (1.07)||59||15.27|
|How likely do you think it is that:|
|Lynette (Desperate Housewives) will die from cancer?||1.95*** (1.01)||61||−15.89|
|Edie (Desperate Housewives) will be saved from her suicide attempt?||5.67*** (1.45)||61||9.03|
|Cristina and Burke’s relationship (Grey’s Anatomy) is truly over?||3.92 (1.55)||61||−0.41|
|George (Grey’s Anatomy) will return?||5.07*** (1.54)||60||5.37|
|Charlie (Lost) will survive?||4.53* (1.78)||60||2.32|
|Syler (Heroes) is truly dead?||3.15*** (1.63)||60||−4.03|
|Sara (CSI) will survive her ordeal?||5.84*** (1.00)||61||14.30|
Given viewers tend to expect positive outcomes for main characters, we next examined why this might be. Across the seven story lines, the notion that the main character was necessary to keep the show interesting constituted on average 41% of the responses. In addition, an average of 20% of responses noted that bad things do not happen to main characters.
The results of Study 1 lend support to the idea that TV viewers have schemas about serial TV programming that contain the beliefs that even when faced with great adversity, main characters bounce back and experience happy endings as they are important to the plot and maintaining interest in the program. If we accept that most people’s serial programming schemas contain these beliefs (indeed 100% of the sample said this at least once across their responses), it is likely this colors the way events in particular episodes are perceived. Specifically, behavioral consequences, though perceived as negative, might be minimized as temporary setbacks easily overcome or resolved. Thus, when main characters lose their jobs, we do not expect such characters will, for example, lose their homes. When main characters are shot, we fully expect them not only to survive but also to have no lasting repercussions as a result of the trauma. Although we might see the injury as life threatening, we do not accord it its full negative impact as we have learned that though the immediate consequences seem bad, the longer term consequences are not especially bothersome.
We should be clear that the concept of schema is not the only one in the literature that allows us to assert that currently held beliefs influence perception of stimuli. Indeed, prominent lines of social influence research, such as cognitive consistency (Festinger, 1957), information integration (e.g., Anderson, 1968), expectancy value (e.g., Fishbein & Ajzen, 1975), and perhaps most especially, social judgment (e.g., Sherif & Hovland, 1961), might allow us to make similar arguments. However, as our goal is not to examine how beliefs about TV are formed and combined with other related beliefs but rather to consider how such beliefs, once they exist, might influence judgments of what is viewed, the notion of schema offers both a parsimonious and appropriate framework for this task. Indeed, it is wholly consistent with the full range of theories that suggest that current beliefs shape perceptions of external and internal stimuli.
With this in mind, we can now revisit how viewers might interpret a scene designed from a SCT perspective to promote safer sexual behavior. Imagine a situation in which a character has unprotected sex and has put herself at risk for AIDS or an unwanted pregnancy. According to SCT, for those without direct experience in these matters, the negative value associated with the behavior would result in viewers being less likely to perform the behavior (unprotected sex) in the future. Indeed, prosocial programming recommendations would argue that more of these sorts of scenes should be included in TV programming. However, with serial TV programming schema in mind, we might arrive at a very different conclusion. That is, these sorts of scenes would have little impact on attitudes or behaviors because the viewers, though they may recognize the dangers associated with unsafe sex, will also presume that ultimately the negative outcomes will have little if any persisting negative implications.
These two different views allow for a test of competing hypotheses. If Bandura’s SCT is appropriately applied to these media contexts, we should expect negative consequences of undesirable behaviors to reduce viewers’ intended action relative to those same behaviors associated with more positive consequences, assuming viewers have not directly experienced the target behavior. If, however, the serial programming schema explanation is more appropriate, we would expect the valence of the reinforcement for the risky behavior (positive or negative) to be inconsequential, and, given the minimization of the negative consequences expected to occur, both depictions should increase intentions for risky behavior for those without direct experience.
We recognize that the expectation of “no difference” between the positive and the negative reinforcement conditions poses its own unique set of concerns. However, if the manipulation clearly depicts, and if the audience perceives, differences in reinforcement value; if there are differential effects for those with and without direct experience; and if there are still no differences in the effects of the different conditions on those without direct experience, we can be reasonably assured that it is the predictions of SCT, rather than operationalization of its components, that are problematic.
Study 2: Testing SCT versus schema theory
In essence, we expect that, consistent with SCT, direct experience will trump vicarious experience. Therefore, depictions of certain experiences should influence attitudes and behaviors only for those who have not had those experiences. For this research, we focus on promiscuous sexual behaviors, specifically, initiating sex on the first date (i.e., having a one-night stand [ONS]). We chose this behavior given its frequent depiction in TV programming and the likely range of experience in this behavior among young women. Thus, we expect the following:
H2: After exposure to fictional portrayals of one-night stands, women who have previously had one-night stands will exhibit no change in their expected future likelihood of engaging in such behavior, whereas women who have not previously had a one-night stand will exhibit an increase in their expected future likelihood of engaging in such behavior.
This, of course, assumes that such behaviors are generally depicted with positive, or certainly with no negative, consequences (Cope-Farrar & Kunkel, 2002). However, according to SCT, the valence of the depiction should influence the likelihood of behavioral modeling for those without past direct experience. Specifically, SCT predicts that those who have not engaged in a depicted behavior who see that behavior positively reinforced for a character with whom they identify will be more likely to engage in that behavior in the future compared to those who see that behavior negatively reinforced. Thus, we predict:
H3: After exposure to a positive (vs. negative) outcome portrayal of a one-night stand, women who have not previously had a one-night stand will report a greater likelihood of engaging in a one-night stand in the future.
However, the serial programming schema interpretation of this dynamic contradicts the SCT interpretation. Specifically, the schema perspective suggests that the valence of the specific outcomes associated with the behavior will not differentially influence viewers’ behavioral expectations. Thus, we offer a competing hypothesis to that generated from SCT.
H3alt: After exposure to a portrayal of a one-night stand, women who have not previously had a one-night stand will report a greater likelihood of engaging in a one-night stand in the future, regardless of the valence of the outcome portrayed.
Clearly, the data can support only H3 or H3alt. If an interaction is found between past behavior and valence of the behavioral outcomes such that positive outcome depictions have greater influence than negative outcome depictions for those without, but not those with, past experience, SCT predictions will be supported. If there is a main effect for past experience such that program exposure evinces differential effects on those with, versus those without, direct experience, but no main effect for the outcome valence of the depiction and no interaction emerge, the results will be more consistent with schema-based predictions.
Participants and procedures
Four hundred undergraduate women at the University of California, Santa Barbara were recruited to participate in exchange for communication course credit. The majority of respondents were Caucasian (67%) and ranged in age from 18 to 30 years (M= 19.75, SD= 1.51; Mdn= 19.50). Twenty-four percent were freshmen, 34% were sophomores, 29% were juniors, and 12% were seniors. Ninety-four percent of participants reported experience with dating, and 70% reported having been sexually active in the past. Particularly relevant to this research, 30% (n= 121) reported having had at least one ONS in the past, whereas 70% (n= 277) did not. Two respondents did not respond to this item. Unfortunately, due to a formatting issue, many women who had not previously had an ONS skipped a key pretest measure regarding future likelihood of engaging in an ONS (n= 159). Thus, for some of the analyses, the sample size for those who had not had an ONS previously is 120. However, it is reassuring to note that the women who did and did not answer this pretest item did not differ in the posttest measure of this same variable (p= .45).
The study was introduced as research about attitudes concerning norms of dating relationships. In groups of 2–12, participants completed a survey that assessed, among other measures, demographics, relationship status, past sexual behaviors including ONS, and TV viewing habits. Participants then watched one of the six edited episodes of a popular fictional comedic drama in which the protagonist has sex on a first date after which they were asked to indicate their emotional reactions to the segment, their attitudes toward the protagonist and her behavior, and their personal likelihood of engaging in an ONS in the future.
This between-subjects study contained six conditions based on three manipulations of intrinsic or extrinsic punishments and rewards: high versus low level of expressed regret, strong versus weak indication of lesson learned, and negative versus positive social outcomes. The initial two sets of manipulations both reflected intrinsic punishment versus reward and were contained within one story line of an episode of the TV series (see below), whereas the third manipulation reflected extrinsic punishment and reward and was contained within a separate story line in a second episode of the same program.
Two episodes of the HBO series Sex and the City were edited to reflect three different manipulations of intrinsic and extrinsic punishments and rewards (described above). Episode 1 was edited down to 7 minutes and captured the first two manipulations, resulting in four stimulus tapes. The beginning of the episode was nearly identical in each condition and features Carrie (the main character who is a New York City newspaper sex columnist and the narrator of the series) joining a female friend for dinner where she hits it off with an out-of-town visitor. They go on a date that ends with sex, even though Carrie makes it clear that she does not normally sleep with men on a first date. The next morning she finds her date has left her a note along with $1,000 in cash. At this point, the conditions diverge, and the first manipulation appears. In the “high-regret” condition, Carrie bemoans her decision, calling herself a whore, and her friends seem to agree with her. In the “low-regret” condition, she merely tells her friends what happened, and they respond as if it is not a big deal. The second manipulation took place in the last minute of the segment. In the “strong lesson learned” condition, Carrie again runs into her friend and meets another man who offers to fly her to Venice. Carrie expresses discomfort and walks away from the situation indicating (to herself) that she is disgusted with her past behavior and will not compromise her morals again (intrinsic negative reinforcement). In the “weak lesson learned” condition, Carrie states (to herself) with equanimity that there was nothing to be done about her previous behavior and jokes that she established her rate for an ONS (intrinsic positive reinforcement). The two manipulations were crossed creating four conditions.
In Episode 2, which was edited down to a 5-minute story line containing the extrinsic reward/punishment manipulation, Carrie is shown before and after having sex on the first date with the man who ultimately becomes the program’s leading man, Mr. Big. In both conditions, Carrie prepares for her date. Shortly, after getting into Mr. Big’s car, they begin to kiss and end up at his apartment where they have sex. In the “negative-outcome” condition, Carrie bemoans the fact that Mr. Big has not called. Later, as she and her friends are waiting on the sidewalk to see her new bus ad, her friend Charlotte comments that Mr. Big did not show up because Carrie slept with him too soon. When the bus arrives, there is graphic graffiti drawn on the side of Carrie’s face in the ad, and the segment ends with Carrie looking dejected and humiliated. In the “positive-outcome” condition, Carrie is thrilled when later Mr. Big calls and asks her to go on a “real first date,” and she happily proclaims that she has avoided “the sex on the first-date curse.”
As this research was part of a larger study on sexual depictions on TV, only those measures relevant to the analyses presented here are described below. All reliability coefficients reported are based on the data from this study.
Respondents were first asked about their current dating relationship status and for basic demographic information. They were then asked whether they have been sexually active, the age of first intercourse, and how many sexual partners they have had in the past 12 months and ever. Respondents were then asked whether they have ever had an ONS. If yes, they were asked how long ago the last one took place. Respondents were then asked how many ONSs they have had, and how likely they would be, compared to the past, to have an ONS in the future on a 1 (much less likely) to 7 (much more likely) scale.
After indicating their feelings or anticipated feelings about personal ONS and completing a trait regret scale, respondents were asked how frequently they watch/have watched each of 40 programs during the peak of their viewing, including Sex and the City, on a 4-point scale (0 =never watch; 1 =watch rarely; 2 =watch sometimes; 3 =watch frequently). Daily hours of TV viewing was assessed by asking respondents how many hours of TV they typically watch during the average weekday and weekend day during four time periods (6 a.m. to 12 p.m., 12 p.m. to 6 p.m., 6 p.m. to 12 a.m., 12 a.m. to 6 a.m.). These data were combined by weighting the “average week day” questions by a factor of five and the “average weekend” question by a factor of two. To measure how real the respondents perceived the media depictions to be, we included Rubin, Perse, and Taylor’s (1988) perceived realism scale (five items; α= .82).
After viewing one of the six stimulus tapes and providing their emotional reactions, respondents indicated how much they enjoyed the segment by completing four items (enjoyable, entertaining, pleasurable, and captivating) on 1 = (not at all) to 7 (very much) scales (α= .93).
The likeability of the main character Carrie was assessed with three 7-point semantic differentials (friendly/unfriendly, likable/unlikable, and pleasant/unpleasant; α= .84).
As a manipulation check for the extrinsic reward/punishment manipulation, participants were then asked to rate the acceptability of Carrie’s sexual behavior with four 7-point semantic differentials (right/wrong, appropriate/inappropriate, fair/unfair, and safe/risky; α= .83). As a manipulation check for the intrinsic reward/punishment manipulations, respondents were asked how regretful Carrie seemed about her behavior on a 1 (not at all) to 7 (very much) scale.
The key dependent measure repeated the item from the pretest that asked how likely, compared to the past, respondents would be to have an ONS in the future. We then asked respondents to imagine how they would react were a close friend to disclose that she had had an ONS using three 7-point items (accepting/not accepting, judgmental/not judgmental, and supportive/not supportive). After reverse coding the judgmental item, the three items formed a reliable index (α= .72) representing an indirect indicator of how acceptable respondents believed ONSs to be, with numbers closer to 1 indicating greater acceptance.
Participants then reported how similar they felt to the protagonist using McCroskey, Richmond, and Daly’s (1975) perceived homophily scale (eight items; α= .80). Finally, respondents were asked how many times per month they watched Sex and the City and whether they had previously seen the episode they viewed during the study.
The hypotheses were tested with paired t tests as well as with analyses of covariance (ANCOVAs) in which we consider each manipulation separately. However, as a result of our partially crossed design, we also examine, when relevant, the high/low regret and strong/weak lesson learned manipulations together to determine if the two interact. Only the pretest measure of likelihood of having an ONS in the future is included as a significant covariate when relevant. Statistical assumptions of these tests were examined and met, with the exception of some minor deviations from normal distribution in the likelihood measure of having a future ONS. However, the tests performed are robust in the face of all but severe deviations. Indeed, parallel nonparametric tests fully reinforce the results reported below.
In preliminary analyses, we found participants did not differ across conditions in whether or not they were sexually active (p= .64), had previously had an ONS (p= .86), pretest likelihood of having a future ONS (p= .46), frequency of watching Sex and the City (p= .27), or whether they had seen the particular stimulus episode before (p= .45).
To determine whether the intrinsic and extrinsic punishments were perceived as intended, we first considered whether or not respondents perceived Carrie as experiencing more negative affect in the high (vs. low) regret and the strong (vs. weak) lesson learned conditions. Indeed, t tests indicated that respondents who viewed the high regret version reported that Carrie displayed more regret (M= 4.30, SD= 1.48) than those who viewed the low regret version (M= 3.87, SD= 1.63), t(271) = 5.19, p= .02, η2= .02. Similarly, respondents in the strong lesson learned condition reported that Carrie showed more regret (M= 4.41, SD= 1.47) than those in the weak lesson learned condition (M= 3.79, SD= 1.61), t(271) =−3.03, p= .001, η2= .04. Of note, analysis of variance (ANOVA) indicated no significant interaction between the two manipulations on perception of Carrie’s regret level, p= .25. Given the manipulations occurred in different parts of the episode and given the lack of interaction between the two, the evidence indicates that these are, in fact, separate manipulations.
We then examined whether respondents who viewed the extrinsic punishment/reward manipulation in the second episode of Sex and the City saw Carrie’s behavior as more or less undesirable. Indeed, those who viewed the negative outcome condition reported her behavior to be less desirable (M= 3.60, SD= 1.29) than those who viewed the positive outcome version (M= 3.11, SD= 1.28), t(125) =−2.13, p= .035, η2= .04. Of note, they also viewed her as more regretful about her behavior (M= 4.15, SD= 1.58 vs. M= 3.21, SD= 1.42), t(125) = 3.53, p= .001, η2= .09. Thus, all manipulations were successful.
Despite these intended variations, ANOVAs indicated no significant differences in enjoyment of the segments among the six conditions (p= .32) nor were there differences in how likable participants found Carrie to be (p= .12) or how similar they felt they were to her (p= .50). Of note, though participants saw themselves as only moderately similar to Carrie (M= 3.61, SD= 1.14), they found her to be very likable (M= 6.06, SD= 1.21), suggesting her character would serve as a viable behavioral model.
Episode viewing and expected behavioral modeling
First, we used paired t tests to consider whether viewing a depiction of promiscuous sexual behavior, regardless of its outcome, influenced likelihood of future promiscuity among viewers with versus without direct experience with ONSs (H2). Across the full sample, we found a significant increase in the likelihood of having an ONS from pre- to postviewing, M= 2.24, SD= 1.58 versus M= 2.55, SD= 1.51, t(238) =−3.74, p < .001, Cohen’s d= .24. Looking only at those who had previously had an ONS, no increase in future intentions of having an ONS emerged, M= 2.43, SD= 1.67 versus M= 2.43, SD= 1.52, t(118) = 0.00, p= 1.00. However, for those who had not previously had an ONS, a paired t test indicated an increased likelihood of having an ONS pre- and postviewing, M= 2.05, SD= 1.48 versus M= 2.66, SD= 1.51, t(119) =−4.67, p < .001, Cohen’s d= .43.
We also performed three ANCOVAs (one for each manipulation) that included two independent variables—past ONS experience and viewing condition—and initial likelihood of having a future ONS as the covariate. A main effect of direct experience on likelihood of a future ONS would support H2. For each ANCOVA, a main effect for previous ONS experience was evidenced at or near significance, indicating that those who had not previously had an ONS were more likely to report future intentions to have one (M= 2.55, SD= 1.51) compared to those who had past ONS experience (M= 2.43, SD= 1.46): degree of regret, F(1, 156) = 7.58, p= .007, η2= .05; extent of lesson learned, F(1, 156) = 6.84, p= .01, η2= .04; and social outcome valence, F(1, 73) = 3.47, p= .066, η2= .05. These two sets of results support H2 and are consistent with the SCT-derived notion that media depictions may sway those who have no direct experience but are unlikely to influence those with direct experience.
Positive and negative reinforcement and expected behavior modeling
Next, we examined whether different depictions of the outcomes associated with the promiscuous sexual behavior influenced expressed future likelihood of having an ONS for those without past direct experience. H3, derived from SCT, suggested that viewing a positive outcome portrayal of an ONS will increase the likelihood of having a future ONS relative to viewing a negative outcome portrayal, whereas H3alt, derived from notions of serial programming schema, predicted increased likelihood of engaging in the behavior, regardless of the depiction’s valence.
These contrasting hypotheses were tested with the three ANCOVAs described above. Support for H3 would be indicated by a significant main effect for previous ONS experience and a significant interaction between past experience and the nature of depiction, whereas support for H3alt would be indicated by the presence of a main effect for experience but no main effect for the nature of the depiction and no significant interaction. As noted above, for each ANCOVA, a main effect for previous ONS experience was evidenced at or near significance. However, no main effect for the nature of the depictions was found in these analyses (ps = .22–.72). Further, no significant interaction emerged for the lesson learned (p= .27) or social outcome valence (p= .23) manipulations. Although a significant interaction did emerge for the expressed regret manipulation, F(1, 156) = 4.66, p= .03, η2= .03, an examination of the interaction itself indicated that, counter to SCT predictions, those with previous experience with an ONS were less likely to expect a future ONS if they were exposed to the high-regret (M= 2.13, SD= 1.09) versus the low-regret condition (M= 2.70, SD= 1.68), whereas no differences appeared for those without previous experience (M= 2.50, SD= 1.45 and M= 2.61, SD= 1.58, respectively). These results are consistent with H3alt, but not H3.
Paired t tests within each condition supported the above findings (see Table 2 for Ms, SDs, and ns). For those who reported having had a past ONS, there were no significant changes in likelihood of having one in the future: high regret (p= .18), low regret (p= .37), strong lesson learned (p= .90), weak lesson learned (p= .50), negative social outcome (p= .29), and positive social outcome (p= .43). However, for those without past ONS experience, most conditions evidenced an increase in future intentions at or near statistical significance: high regret (p < .001; Cohen’s d= .67), low regret (p= .07; Cohen’s d= .28), strong lesson learned (p= .002; Cohen’s d= .53), weak lesson learned (p= .03; Cohen’s d= .33), negative social outcome (p= .08; Cohen’s d= .43), and positive social outcome (p= .15; Cohen’s d= .33).
|Outcome Depiction||Had ONS||Not Had ONS|
|High regret||38||2.37 (1.57)||2.13 (1.09)||38||1.74*** (1.25)||2.61 (1.59)|
|Low regret||43||2.56 (1.69)||2.70 (1.68)||42||2.07† (1.39)||2.50 (1.45)|
|Strong lesson learned||43||2.30 (1.49)||2.33 (1.44)||40||1.88** (1.44)||2.73 (1.48)|
|Weak lesson learned||38||2.66 (1.77)||2.55 (1.48)||40||1.95* (1.34)||2.38 (1.53)|
|Negative social outcome||20||2.45 (1.82)||2.35 (1.73)||19||2.21† (1.65)||2.95 (1.51)|
|Positive social outcome||18||2.28 (1.74)||2.56 (1.62)||21||2.43 (1.66)||2.81 (1.54)|
Attitude toward one-night stands
The above evidence suggests that exposure to any depiction of promiscuous behavior might boost expected likelihood of future promiscuous behavior among those without past experience. To further explore any additional constructs that might illuminate this issue, we examined whether or not attitudes toward promiscuous behavior were influenced by the different outcome portrayals. We focused on two dependent measures: respondents’ attitudes toward Carrie’s sexual behavior and how supportive respondents would be of a friend who disclosed having had an ONS. Three ANCOVAs with the same architecture as those reported above, but with attitudes toward Carrie’s behavior as the dependent measure, indicated significant main effects for having previously had an ONS for all three analyses. In each case, those who had previously had an ONS indicated greater acceptance of Carrie’s behavior (M= 4.20, SD= 1.06) than those who had not (M= 3.39, SD= 1.19), Fs(1, 156) = 10.31–11.61, ps = .001–.002, η2= .06–.14. No significant main effects for the nature of the portrayals (expressed regret p= .34; lesson learned p= .35; social outcome valence p= .23), and no interactions between past experience and nature of portrayal (expressed regret p= .31; lesson learned p= .25; social outcome valence p= .36), emerged.
We then performed three additional ANCOVAs with architecture similar to those above but with judgment of a close friend who had had an ONS as the dependent measure. In each case, those who had previously had an ONS indicated less negative assessment of a friend’s ONS (M= 2.29, SD= 1.16) than those who had not (M= 3.30, SD= 1.28), Fs(1, 156) = 14.44 – 25.65, p < .001, η2= .13 – .26. However, no significant main effects for the nature of the portrayal (expressed regret p= .27; lesson learned p= .37; social outcome valence p= .39) and no significant interactions between past experience and nature of portrayal emerged (expressed regret p= .51; lesson learned p= .83; social outcome valence p= .69), suggesting the nature of the depiction did not influence attitudes toward ONS.
In sum, the lack of main effects for the different positive and negative outcome portrayals, as well as the lack of significant interactions, offers further evidence consistent with the schema perspective relative to SCT.
Two concerns might be raised about the behavioral modeling analyses. First, many of those who had not previously had an ONS did not complete the pretest measure of future ONS likelihood (56%) and thus could not be included in the above analyses. We assumed this omission was due to a formatting issue in the survey (in which an indention seems to have implied that only those who had previously had an ONS should answer the set of questions) rather than a relevant systematic difference between these respondents. To be sure, we compared those who completed the item and those who did not along a range of variables. The two groups did not differ in age (p= .79), race (p= .89), or current dating situation (p= .24). Although those who skipped the pretest item were less likely to be sexually active compared to those who did not (51 vs. 70%, p= .001), this variable was not a significant covariate in any of the analyses. Most important, the two groups did not differ on the posttest measure of future ONS likelihood (skipped: M= 2.52, SD= 1.50; answered: M= 2.66, SD= 1.51, p= .45). Thus, we conclude that this methodological glitch does not compromise the validity of the reported findings.
A second concern involves the potential influence of having seen the stimulus episodes before. Nearly, half (48%) of the respondents had previously seen their episodes and thus might have known that Carrie’s character rebounds well from both of her sexual escapades. Previous viewing was not a significant covariate in any of the analyses and including it did not meaningfully alter the results. Still, we reran the ANCOVAs with only those respondents who had not previously seen the episodes. Despite the diminished power associated with reducing the sample size by half, results were identical, if not stronger in magnitude, to those based on the complete sample. All three analyses revealed a main effect for previous ONS experience, suggesting those without past experience would be more likely to have an ONS in the future compared to those with experience (expressed regret p= .08, η2= .04; lesson learned p= .05, η2= .05; social outcome valence p= .03, η2= .12). In addition, the SCT-inconsistent interaction for the expressed regret condition noted earlier again emerged (p= .02, η2= .07), suggesting that those who had previously had an ONS were particularly influenced by the regret stimulus compared to the other depictions. However, no main effects for the portrayal outcomes (expressed regret p= .58; lesson learned p= .43; social outcome valence p= .54), and no other significant interactions (lesson learned p= .68; social outcome valence p= .90), emerged. Thus, previous exposure to the stimulus does not explain the pattern of effects found in these data.
This research was designed to offer an experimental test of the long-standing assumption, derived from SCT, that negative portrayals of the outcomes associated with risky behaviors will discourage inexperienced viewers from performing them. Consistent with SCT, the results suggested that the portrayals of sexual behavior were likely to affect only those without direct experience with the target behavior. However, inconsistent with SCT, the valence of the portrayal outcomes did not differentially affect attitudes or intentions regarding that behavior. Specifically, for those who had not previously had an ONS, viewing fictional depictions of this behavior significantly increased expectations of the likelihood of having an ONS in the future, regardless of the positive and negative outcomes associated with it. This finding, evidenced across three different operationalizations of consequences, is inconsistent with SCT, though consistent with what might be expected from the application of serial programming schema.
Before concluding that SCT is inappropriately applied in this media context, we are bound to explore other possible explanations for these findings. For example, it might be argued that the reason for the lack of a significant interaction between past experience and nature of reinforcement may stem from the popularity of the program, which may have undermined the reliability of the manipulations as women who regularly watch Sex and the City know that the protagonist does not, in fact, suffer any long-term serious consequences for her actions. However, recall first that the manipulation checks indicated that the conditions were seen differently by the viewers. Second, our data suggested that regular program viewing was not a significant covariate in the analyses. Further, additional analyses indicated that the results were as strong, if not stronger, for those who had not previously seen the episodes. Thus, our failure to find a significant interaction between past experience and outcome valence of the portrayal cannot be explained away by weak manipulations or past viewing experience.
Related to these concerns, we might first also consider whether promiscuous sexual behavior was not a relevant enough issue to the audience for modeling to be a realistic possibility or, second, whether the consequences portrayed, though perceived as different, were not negative enough to dissuade the audience from being tempted by the behavior. The fact that nearly one third of the sample had already engaged in the target behavior, 70% had been sexually active, and college students are generally living unsupervised as they explore their social relationships and sexuality, it is unlikely that sexual behavior was not relevant to the majority of the participants.
As to the consequences, we did choose to operationalize negative consequences in a somewhat mild manner, that is, through internal punishment of regret and shame and through external social sanctions. Had we chosen more explicit consequences, such as pregnancy or sexually transmitted disease, it is possible our results may have been different. However, it is not clear that such manipulations would have external validity as they do not appear to be consistent with the type of messages that are disseminated in TV programming (see Cope-Farrar & Kunkel, 2002; Kunkel et al., 2005). That is, rarely are liked main characters subjected to truly negative, long-term consequences. Although they may have pregnancy scares or brief brushes with STDs (both of which have been depicted on Sex and the City), these potential threats are generally resolved positively by the end of an episode or the effects in the long term are portrayed as more benign than they really are. Indeed, it is unlikely that producers would be open to displaying long-term negative consequences given they are antagonistic to the norms of entertainment. We further point out that the negative consequence conditions (high regret, strong lesson learned, negative social outcome) seemed to generate slightly stronger interest in the risky behavior than the positive consequence conditions (see Table 2, column 7). Thus, it is possible that portrayals with more severe consequences might not only fail to evidence the desired outcome but might also be prone to backfiring.
As we consider these outcomes, it is possible that the messages encourage the risky behaviors, regardless of the valence of the consequences associated with them, because of the very fact that liked characters perform them, which may infuse that behavior with positive value, despite of the consequences depicted. Of course, this sets up another troubling dilemma for the application of SCT. That is, if liked characters are those most likely to be modeled, yet their behaviors, despite the actual consequences depicted, are ascribed a positive value, it may not be reasonable to expect to use such characters as models to encourage audiences to avoid negative behaviors. Of course, the converse of this logic is that liked TV characters may be particularly well suited for modeling positive behaviors, such as quitting smoking, eating healthy, or practicing safe sex. However, given these behaviors are not usually as dramatic, it is unlikely that they will naturally occur in TV programming as often as the more exciting, and often more risky, behaviors.
Given that the manipulations were successful, yet audience members, even those who had not seen these episodes before, did not evidence the SCT-based expected difference in reaction to viewing positive versus negative behavioral reinforcement, we must consider alternative theoretical perspectives that would not simply explain but predict these results. The notion of serial programming schema does so. If people bring to the viewing environment the story schema that liked characters will survive and thrive despite the adversity they face, then it is likely that, despite seeing an outcome as negative, they will not see its long-term consequences as such, which will likely undermines the impact of that negative consequence in the short run.
Future research would be well served if the lines of reasoning outlined above are explored by further investigating the contents of serial programming schema as well as viewers’ perceived short-term and long-term consequences of a liked protagonist’s actions. Indeed, research into whether schema content or simple priming serves as the operative mechanism explaining these sorts of outcomes would be important. The fact that differential effects were revealed for inexperienced versus experienced women suggests a more complicated cognitive process than mere exposure, or priming, would suggest, but this is still an explanation that warrants attention, especially for those with direct experience, given their response to the regret depiction. This area of inquiry would also be well served by considering other forms of manipulation of punishment or risk, keeping in mind what is likely to occur in mainstream entertainment programming.
There are certainly limitations to this study. First, we did not measure actual behavior. Thus, we cannot know if the increased projected likelihood of having an ONS translated into action. Second, the lesson learned and expressed regret manipulations were embedded in the same episode, and though the manipulation checks were successful, perhaps we should be cautious in assuming we had three distinct manipulations of behavioral reinforcement. Third, as noted earlier, the measurement of pretest ONS likelihood was problematic, though we believe our evidence suggests this was not a serious threat to the findings presented. Finally, we considered only one behavior, and though conceptually the arguments should apply more broadly, it is unclear the extent to which these results generalize to other behaviors.
Despite these concerns, we feel confident concluding that, contrary to SCT, even when behaviors are negatively portrayed, audiences may be motivated to model them anyway. We in no way mean this as a condemnation of SCT. Rather, we hope this research stimulates greater care in the application and testing of psychological theories to the study of media contexts and effects. If it is the case that the positive value ascribed to risky behaviors as a result of liked celebrities performing them outweighs the negative outcomes displayed for those behaviors or that the negative outcomes are ultimately minimized in the minds of audiences, then the oft-proffered suggestion of showing the negative consequences of risky behaviors is not sufficient to minimize their negative impact on susceptible viewers. It is critical that the scholarly community test these notions before recommendations are made and followed not only to avoid compounding the negative effects of the media that currently exist but also to begin to work toward creating portrayals that both enhance entertainment value and minimize risk.
The authors thank Michael Pfau and two anonymous reviewers for their insightful comments, which have undoubtedly improved the clarity of this article.