Abstract
In addition to the more obvious forms of sexism in advertising, media critics and scholars raise concerns about various forms of nonobvious, or latent, sexism (e.g., “dismembered” body parts; makeup possibly resembling a bruise; women in potentially dangerous locations; bodies decorated as products). There is, however, no evidence that the public considers these ads sexist or is affected by them. To determine whether ads promote sexism even if the content is not identified as sexist, participants were exposed to ads containing no sexism, overt sexism, or latent sexism (i.e., content considered sexist by media experts, but not identified as sexist by a lay sample) and then read two vignettes describing incidents of sexual assault and sexual coercion. Participants exposed to ads with latent sexism showed greater acceptance of the sexual assault than did those in the no sexism ad condition and in the overt sexism ad condition. Regarding the sexual coercion vignette, latent sexism did not have the same effects; instead, participants exposed to ads with overt sexism were less likely to minimize the seriousness of the incident than participants in the other ad conditions. Therefore, acceptance of sexual assault can be increased by sexist content in ads even if the content is not identified as sexist. In fact, the evidence suggests that the types of latent sexism in this study produce more deleterious effects than sexism that is easily recognized.
Media critics (Cambridge Documentary Films, 1979, 1987; Kilbourne, 1999; Media Education Foundation, 2000, 2010) and scholars alike (e.g., Benokraitis & Feagin, 1995; Conley & Ramsey, 2011; Cortese, 2008; Goffman, 1979; Kang, 1997; Lindner, 2004; Stankiewicz & Rosselli, 2008) believe sexism in advertisements is revealed not only through overt, easily recognizable content, such as scantily clad women or women in stereotypical roles, but also through less obvious, latent content that may only become apparent through deconstruction of the ads (e.g., a woman seated next to a standing man; eye makeup that may be seen as resembling a bruise; a woman’s legs shown without the rest of her body). Furthermore, these experts claim the less obvious forms of sexism perpetuate stereotypes of women, women’s inferior status, and even acceptance of abuse and sexual assault of women. Yet, despite the widespread acceptance by experts, there is no evidence that the public considers the latent content to be sexist, nor is there any evidence that the public is affected by these images. Unlike ads containing obvious sexism, whose effects are well established (e.g., American Psychological Association [APA], 2007; Ward & Harrison, 2005), there has been no research into the presumed effects of latent sexism in advertisements. Therefore, this research was designed to determine whether latent sexism in print ads (i.e., content considered sexist by experts, but not identified as sexist by laypersons) promotes acceptance of sexual assault.
The notion that ads convey meaning about gender without viewers’ awareness is not new. In his influential Gender Advertisements, Goffman (1979) argued that advertisements symbolically reflect social-cultural constructions of gender through displays of posture, positioning, facial expressions, and social roles: Sitting at a man’s feet (ritualized subordination), gazing off absently (licensed withdrawal), or gently caressing an object (feminine touch) all demonstrate women’s inferior status. According to Goffman, however, we fail to recognize the sexism in these images precisely because they reflect our unexamined assumptions about gender. Nonetheless, these gender displays allegedly perpetuate sexist stereotypes. Despite the complete lack of empirical evidence showing that these images promote sexist beliefs or attitudes, Goffman’s analysis is widely accepted by scholars, who regularly employ his taxonomy of gender displays to establish the prevalence of sexism in the media (e.g., Belknap & Leonard, 1991; Kang, 1997; Krassas, Blauwkamp, & Wesselink, 2003; Lindner, 2004; Millard & Grant, 2006; Plous & Neptune, 1997).
Building on Goffman’s (1979) work, influential media critic Jean Kilbourne (1999; Cambridge Documentary Films, 1979, 1987; Media Education Foundation, 2000, 2010) explicitly claims that sexism in advertising affects us unconsciously. In fact, the sexist content in many of her illustrative ads only becomes apparent through her deconstruction of them; as a result, the lay public may not even consider the ads sexist. For example, she argues that violence against women is normalized and trivialized by ads using only female body parts (the body “hacked apart” or “dismembered”), images of startled women in dark or potentially dangerous locations, ostensibly playful scenes of men chasing women, women’s bodies decorated as products (objectification), and dark eye makeup (possible bruises from assault). The message is that women are less than human, they are victims, and may even enjoy being assaulted. She states, “Images like this . . . don’t directly cause violence. They create a climate in which it is seen as more tolerable” (Cambridge Documentary Films, 1987, p. 27:24).
Kilbourne’s lectures, presented in the Killing Us Softly video series (Cambridge Documentary Films, 1979, 1987; Media Education Foundation, 2000, 2010), have been viewed by millions and she has been named by Time magazine as one of the three most popular speakers on college campuses (http://www.mediaed.org/cgi-bin/commerce.cgi?preadd=action&key=241). Her videos are used extensively by educators in psychology, communications, and gender studies (Conley & Ramsey, 2011). In short, she is widely viewed as an expert and her ideas enjoy widespread acceptance. Yet, little research has examined her specific claims. Although there is support for some of her claims about the prevalence of sexist imagery in ads (Conley & Ramsey, 2011), her claims of effects from ads with unrecognized (i.e., latent) sexism remain untested. The present research is designed to examine these claims by determining whether acceptance of sexual assault is increased by exposure to ad content with latent themes of abuse and victimization of women.
There is, of course, ample research demonstrating myriad, robust effects of overt, easily recognized sexism (APA, 2007; Ward & Harrison, 2005). Images of women as sex objects have been shown to increase men’s attributions of responsibility to female rape victims (Wyer, Bodenhausen, & Gorman, 1985) and produce more favorable attitudes toward interpersonal violence, rape myth beliefs, and gender stereotypes (Lanis & Covell, 1995; MacKay & Covell, 1997). In addition, print images of women in traditional homemaker roles caused women’s attitudes toward political participation to become less favorable (Schwarz, Wagner, Bannert, & Mathes, 1987). Exposure to sexist television commercials produced comparable effects among women, including decreased body satisfaction (Lavine, Sweeney, & Wagner, 1999), reduced achievement aspirations for the future (Geis, Brown, Jennings, & Porter, 1984), diminished leadership aspirations (Davies, Spencer, & Steele, 2005), and lower preference for quantitative careers (Davies, Spencer, Quinn, & Gerhardstein, 2002). After viewing sexist commercials, men assigned to the role of interviewer exhibited more sexist behavior toward a female confederate (L. A. Rudman & Borgida, 1995). Even sexism in music videos has been shown to produce more gender-stereotyped perceptions of cross-sex social interactions (Hansen & Hansen, 1988). Yet, the stimuli used in all these studies are widely recognized as sexist; in fact, they are selected deliberately for their obvious sexual content, or the sexism is validated through manipulation checks (i.e., images presented to participants are independently rated as highly sexist by another group of participants). In contrast, the present research investigates possible effects of latent sexism, operationalized as content considered sexist by media critics, such as Kilbourne (1999; Cambridge Documentary Films, 1979, 1987; Media Education Foundation, 2000, 2010), and scholars (e.g., Cortese, 2008; Goffman, 1979; Lindner, 2004; Stankiewicz & Rosselli, 2008), but not identified as sexist by laypersons.
There is some indirect support for the idea that sexist ad content may have effects even if it is not considered sexism. In its review of the literature, the APA (2007) identified implicit, automatic, unconscious processes as mechanisms by which sexist media may have its effects. Similarly, in several studies where stimuli were overtly sexist and demonstrably affected participants, researchers concluded that their participants nevertheless did not recognize the sexism (Hansen & Hansen, 1988; Jennings-Walstedt, Geis, & Brown, 1980; Lavine et al., 1999; L. A. Rudman & Borgida, 1995; Schwarz et al., 1987). Lavine et al. (1999) even postulated automatic activation of gender stereotypes to explain why feminists were just as vulnerable as nonfeminists to the effects of sexist TV ads.
Other research suggests the effects of sexism may be unrecognized, whether or not the sexism itself is. For example, research on face-ism shows that the bias toward facially oriented depictions of men and full-body portrayals of women (e.g., Dodd, Harcar, Foerch, & Anderson, 1989; Hall & Crum, 1994) promote the endorsement of gender stereotypes in both men and women (Levesque & Lowe, 1999; Schwarz & Kurz, 1989). Yet, it is not clear whether face-ism effects meet the criteria for latent sexism because there is no direct evidence that the public is unaware of this sexist bias. Similarly, the same pin-up style images of women legally designated as constituting sexual harassment if displayed in the workplace are accepted without question in publicly displayed advertisements (Rosewarne, 2007). Yet, although the effects of these images may be unrecognized, there is no evidence that the public does not recognize the images as sexist (the criteria for latent sexism).
Similar claims about effects of unrecognized sexism are made in domains other than advertising as well. For example, some forms of sexist behavior, such as the use of sexist language and sexist jokes, have been labeled as subtle sexism (Benokraitis & Feagin, 1995) because they are not always recognized as sexist (Swim, Mallett, & Stangor, 2004). These forms of subtle sexism can affect gender-related behavior and perception, though again there does appear to be some recognition by the public that these behaviors are sexist (Stout & Dasgupta, 2011). Likewise, it has been argued that the glamorization of prostitution and sex work through television shows, pimp/ho fashion, language, and video games, constitutes a form of symbolic violence against women that is unrecognized because of its widespread acceptance within popular culture (Coy, Wakeling, & Garner, 2011). Further evidence is necessary, however, to establish these forms of sexism as unrecognized by the public (i.e., latent), and as having the claimed effects.
Therefore, the purpose of this research is to determine whether latent sexism in print ads (i.e., ad content considered sexist by experts, but not identified as sexist by laypersons) can promote sexist attitudes and beliefs. To address this question, we focused specifically on the latent theme of abuse and victimization of women. If Kilbourne (1999; Cambridge Documentary Films, 1979, 1987; Media Education Foundation, 2000, 2010) is correct, latent abuse themes in ads should increase acceptance of sexual assault and sexual coercion (i.e., more blame attributed to the victim and less to the perpetrator, and minimization of the seriousness of the incidents). On the contrary, if the public is correct, and there is actually nothing in the ad content promoting sexism, then the ads should have no effect on beliefs about the victims, perpetrators, and seriousness of the incidents.
We examined Kilbourne’s (1999; Cambridge Documentary Films, 1979, 1987; Media Education Foundation, 2000, 2010) claim using a paradigm established by earlier studies of sexist advertising and perceptions of rape. Those studies demonstrate short-term perceptual and attitudinal effects following exposure to sexist print ads (Garst & Bodenhausen, 1997; Lanis & Covell, 1995; MacKay & Covell, 1997; Schwarz et al., 1987). Typically, participants view either sexist or nonsexist ads, after which their endorsement of victim blame, rape myth acceptance, or sexual aggression is compared. This design was expanded in the current research to include a latent sexism condition, with ads containing content considered sexist by Kilbourne and others (e.g., Cortese, 2008; Stankiewicz & Rosselli, 2008), but with the stipulation that laypersons do not identify the ads as sexist. As in studies of rape perceptions (e.g., Schneider, 1992; Simonson & Subich, 1999; Wyer et al., 1985; Xenos & Smith, 2001), participants then make judgments regarding victims, perpetrators, and the incidents.
Kilbourne’s (1999; Cambridge Documentary Films, 1979, 1987; Media Education Foundation, 2000, 2010) claims lead to the primary hypothesis that, compared with the no sexism ad condition, the latent sexism ad condition will express greater acceptance of sexual assault and sexual coercion (i.e., the victims are to blame, the perpetrators are not responsible, and the incidents are not serious). Based on previous research, overtly sexist ads are also predicted to produce greater acceptance of the incidents than the nonsexist ads. An additional research question is whether latent sexism and overt sexism differ in the strength of their effects. Greater effects from overt sexism may suggest that latent sexism is simply a weaker form of sexism, whereas greater effects from latent sexism may indicate that it is a distinct and possibly more deleterious form of sexism.
Method
Participants
The sample included 226 women (77%) and 66 men who participated in the actual experiment. An additional 33 women (70%) and 14 men participated in the stimulus ad validation process. All participants were undergraduates at a university college (a primarily 2-year institution, with some 4-year degrees). The percentages of students reporting they were in their first, second, third, and fourth year of study were 39, 46, 12, and 1, respectively. Participants were offered entry into a raffle as an incentive and a majority also received course credit. They ranged in age from 18 to 51, with a mean and median age of 22.5 and 20 years, respectively. The ethnic composition of the sample was Caucasian (44%), Asian (26%), East Indian (18.5%), along with members of nine additional ethnic groups (11.5%). Less than 1% did not identify their ethnicity. On average, participants had lived in Canada for 17.8 years (Mdn = 19.0). Data from two participants were excluded because it was discovered they were only 17 years old (ethics approval stipulated a minimum age of 18).
Procedure
Participants, who were told they would be taking part in two separate studies, were randomly assigned to one of three ad conditions: Overt sexism, no sexism, or latent sexism. To disguise the true purpose of the ads, the first task was introduced as a study of perceptions of advertisements. Each participant proceeded through a binder containing either 10 overt sexism ads, 10 no sexism ads, or 10 latent sexism ads, and rated each ad for interest, amount of product information, design, and persuasiveness. Ratings were made on 7-point Likert-type scales. In addition to preventing participants from recognizing the connection between the ads and their perceptions of the subsequent vignettes, these ratings provided data for addressing potential confounds in the ad condition effects. To the extent it was logistically feasible, an effort was made to keep the experimenters blind to participants’ condition through random assignment of binders, which were identified only by a code number.
When the participants finished the ad ratings, the experimenter introduced the next task as a study of perceptions of sexual encounters. To reinforce the cover story and convince participants that this task was unrelated to the advertisements, a second informed consent was obtained. Participants then received a booklet with two vignettes and questionnaires assessing beliefs about the victims, incidents, and perpetrators. The order of vignettes was counterbalanced. In the primary vignette, adapted from Wyer et al. (1985), a woman is sexually assaulted in her office after hours by a man who enters her office uninvited, closes the door, forcibly grabs her, and engages in intercourse without any indication of consent. To allow for some ambiguity, she is described as dressing to attract men, knowing her assailant, and not resisting (out of fear for her safety). The second vignette was a slight modification of an incident of verbal sexual coercion from Luddy and Thompson (1997) and was far more ambiguous. In this scenario the woman, who twice states she does not want to have sex, nevertheless agrees to engage in sexual intercourse to preserve her friendship with the man (her business colleague). Although the man is manipulative, no physical force or threat of force is used. This vignette, which is atypical in studies examining acceptance of sexual assault, was included to examine the generalizability of the results.
After reading each vignette, participants answered 10 questions relating to the victim, the perpetrator, and the incident. Questions were drawn from previous studies in which participants read sexual assault or sexual coercion vignettes (Schneider, 1992; Simonson & Subich, 1999; Wyer et al., 1985; Xenos & Smith, 2001) and generally reflected three commonly measured dimensions: Victim blame, seriousness of the incident, and responsibility of the perpetrator. Following the vignette task, participants completed a short demographic questionnaire asking for their age, gender, year of study, major, ethnic background, and years lived in Canada. Participants were then thoroughly debriefed and provided with information about counseling to assist with any possible distress created by the vignettes.
Stimulus Ads
A three-step process was used to select the print ads to be used as stimulus materials for the three experimental conditions (i.e., latent sexism, overt sexism, no sexism). In the first step, the lead author drew a large sample (N = 83) of full-page ads, believed to represent latent sexism, overt sexism, and no sexism, from popular women’s magazines (e.g., InStyle, Glamour, Cosmopolitan). To narrow the category of latent sexism, Kilbourne’s (1999; Cambridge Documentary Films, 1979, 1987; Media Education Foundation, 2000) analysis of ads depicting women’s victimization and abuse in nonobvious ways served as the basis for selecting ads with latent sexism. These ads included images of dismembered legs, makeup that could resemble a bruise, women in potentially dangerous locations, women’s bodies decorated with product logos, women presented alongside violent ad copy (e.g., “pow,” “bang”), and women held by men in vulnerable positions. Because the research purpose was to determine possible effects of latent sexism, and not its prevalence, no effort was made to select a representative sample of ads. Ads representing overt sexism were selected according to traditional criteria, such as women portrayed as sex objects, inferior to men, unintelligent, or trivialized. Nonsexist ads were defined by the absence of sexist content and included depictions of women as active and competent (e.g., as athletes or in business), or performing the same activity as a man.
In the second step of the ad selection, a lay sample of 40 untrained university students (27 women, 13 men) viewed the 83 ads and rated each on a scale of 1 (definitely not sexist) to 7 (definitely sexist). In the third step, a sample trained to recognize latent sexism examined all ads with mean sexism ratings less than 4.0 from the lay sample (n = 46 ads) and coded each dichotomously as either displaying no sexism or exhibiting some form of latent sexism. The trained sample of six female and one male fourth-year psychology students attended an information session on latent sexism and used a checklist identifying categories of latent sexism to code the ads. The average kappa for the 21 pairings of raters was .59, indicating moderate to substantial reliability according to benchmarks reviewed by Stemler (2001). All kappas were significant, 18 at the .001 level. The criteria for the final selection of ads for the three experimental conditions, along with mean sexism ratings, were as follows:
Overt sexism—The 10 ads with the highest mean sexism ratings from the lay validation sample (with the qualification that no brand name occurred more than once; M = 6.22, SD = 0.81)
No sexism—The 10 ads with the lowest mean sexism ratings from the lay validation sample and high agreement (six or seven out of seven) among trained coders that the ads contained no latent sexism (M = 2.12, SD = 0.85)
Latent sexism—The 10 ads with the lowest mean sexism ratings from the lay sample and high agreement among trained coders that the ads contained latent sexism (M = 2.36, SD = 0.81)
As final validation of the ads selected for the ad conditions, ANOVA confirmed significant differences in the mean sexism ratings from the lay sample, F(2, 76) = 368.91, p < .001, η2 = .91. The final set of no sexism ads and latent sexism ads were both rated by the lay sample as significantly less sexist (ps < .001) than the overt sexism ads. Moreover, confirming the criteria for latent sexism, the set of latent ads were rated as no more sexist than the set of nonsexist ads (p = .35). In fact, the latent mean of 2.36 on a 7-point scale was more than 1.5 standard deviations below the midpoint, indicating the lay sample clearly did not identify the latent ads as sexist. Women and men did not rate the ads differently, F(1, 38) = 2.44, p = .13, and gender did not interact with ad type, F(2, 76) = 2.09, p = .13. Cell sizes for the latent sexism, overt sexism, and no sexism conditions were 71, 71, and 84, respectively, for women, and 19, 26, and 21 for men.
Perceptions of Sexual Assault and Sexual Coercion
To assess perceptions of victim blame, seriousness of the incident, and perpetrator responsibility for the sexual assault and the sexual coercion, participants answered 10 questions about each incident. Questions were coded so that higher scores were consistent with greater acceptance of sexual assault and sexual coercion. The questions were as follows: “To what extent was Mary (Susan) responsible for what happened?” 1 (not at all responsible) to 7 (very responsible); “To what extent did Mary (Susan) desire sexual intercourse with Phil (Tom)?” 1 (not at all) to 7 (very much); “To what extent was it possible for Mary (Susan) to prevent or stop the act of sexual intercourse?” 1 (not at all possible) to 7 (very possible); “To what extent did Mary’s (Susan’s) behavior encourage Phil (Tom) to have intercourse with her?” 1 (not at all) to 7 (very much); “To what extent will this incident cause serious psychological problems for Mary (Susan)?” 1 (not at all likely) to 7 (very likely; reverse coded); “To what extent did Phil (Tom) violate Mary’s (Susan’s) rights?” 1 (not at all) to 7 (very much; reverse coded); “To what extent would you characterize this incident as rape?” 1 (definitely not rape) to 7 (definitely rape; reverse coded); “To what extent was Phil (Tom) responsible for what happened?” 1 (not at all responsible) to 7 (very responsible; reverse coded); “To what extent was Phil’s (Tom’s) behavior due to his misunderstanding of Mary’s (Susan’s) behavior or desires?” 1 (not at all) to 7 (very much); “To what extent was Phil’s (Tom’s) behavior due to psychological problems?” 1 (not at all) to 7 (very much). An additional question asking for recommended jail time for Phil (Tom) was excluded due to extreme, positive skew.
Exploratory factor analysis was carried out to determine the factors underlying the 20 items across both vignettes, and to create scales representing the factors for use as dependent variables. Analysis was conducted with SPSS Factor, utilizing principal axis factoring and direct oblimin rotation. The question, “To what extent was Phil’s (Tom’s) behavior due to psychological problems?” was eliminated from the analyses because it did not load on any factors (using 0.4 as the threshold), thus leaving 18 items. Following the procedures set out by Courtney (2013), several techniques were used to determine the proper number of factors; however, the results were not entirely conclusive. The parallel analysis, optimal coordinate, comparison data, and Kaiser’s rule techniques indicated four factors, whereas the minimum average partial technique determined three. In addition, parallel analysis conducted on a randomly generated dataset with 18 variables and 292 cases also identified the cutoff at three factors. Further examination of the fourth factor (eigenvalue = 1.40; 7.77% of variance accounted for) revealed a single item with no cross-loadings (a negative loading for the sexual coercion item “possible to stop or prevent the incident”), and one item with a slightly higher cross loading on Factor 3 (a negative loading for the sexual coercion item of victim responsibility). Therefore, we elected to exclude this factor from the analyses in favor of the clearer three-factor solution.
The first factor (eigenvalue = 5.04; 28% variance accounted for) consisted of all nine questions relating to the sexual assault, and none of the sexual coercion items. This factor includes the beliefs that the incident was not very serious (it was not rape, her rights were not violated, she is unlikely to suffer serious psychological consequences), that the victim was to blame (she is responsible, she could have stopped/prevented it, she may have desired it, she encouraged it), and that the perpetrator is not to blame (he is not responsible, he misunderstood her). This factor may be interpreted as a general acceptance of sexual assault, with a consistent set of supportive beliefs. The nine items loading on this factor were averaged to create an acceptance of sexual assault scale with very good reliability (α = .85).
The second factor (eigenvalue = 2.53; 14.05% variance accounted for) consisted of four items relating to the sexual coercion incident, including the belief that the victim’s rights were not violated, it was not a rape, she is unlikely to suffer any serious psychological consequences, and the perpetrator was not responsible. This factor may be interpreted as minimization of the seriousness of sexual coercion. The four items loading on this factor were averaged to create a scale measuring minimization of sexual coercion with acceptable reliability (α = .73).
The third factor (eigenvalue = 1.66; 9.23% variance accounted for) also consisted of four items relating to the sexual coercion. It included the belief that the victim of sexual coercion desired intercourse with the perpetrator, his behavior was due to misunderstanding of her behavior and desires, her behavior encouraged him, and she was responsible. This factor may be interpreted as blame for the victim of sexual coercion. The four items were averaged to create a victim blame for sexual coercion scale, with acceptable reliability (α = .70).
Results
To determine whether the different types of sexist ads affected perceptions of the sexual assault and sexual coercion, the scales measuring acceptance of sexual assault, minimization of sexual coercion, and victim blame for sexual coercion were each analyzed with a 3 (ad type [latent sexism, overt sexism, no sexism]) × 2 (participant gender) unweighted means ANOVA. Checks of assumptions indicated no issues with normality and no violations of homogeneity of variance. The least significant difference (LSD) method (i.e., protected t test) was used for post hoc comparisons following significant effects. Effect sizes are given as
Mean Ratings of Acceptance of Sexual Assault, Minimization of Sexual Coercion, and Victim Blame for Sexual Coercion by Ad Type and Gender.
Note. All means are on a 7-point scale. Standard deviations are in parentheses; 95% confidence interval in brackets. Means with different subscripts within the same row differ at p < .05.
Higher scores indicate less victim blame.
Sexual Assault
Analysis of the acceptance of sexual assault measure revealed a significant main effect for ad type, F(2, 286) = 4.23, p = .015, η2 = .03. Confirming the primary hypothesis, participants exposed to ads with latent sexism (M = 3.43, SD = 1.28, CI = [3.18, 3.77]) expressed greater acceptance of sexual assault than did participants exposed to nonsexist ads (M = 2.96, SD = 1.09, CI = [2.71, 3.26]), p = .02, d = .40. The latent sexism ad condition also expressed greater acceptance of sexual assault than did the overt sexism ad condition (M = 2.93, SD = 1.15, CI = [2.67, 3.19]), p = .007, d = .42. This provides additional evidence that latent sexism is distinct from overt sexism. The overt sexism and no sexism ad conditions did not differ. The main effect of gender was just above the .05 alpha level, F(1, 286) = 3.53, p = .06, η2 = .01, and the interaction of ad type and gender was nonsignificant, F(2, 286) = 1.67, p = .29, η2 = .01.
Sexual Coercion
The minimization of sexual coercion scale was analyzed with a 3 (ad type) × 2 (participant gender) ANOVA. The main effect of ad type was again significant, F(2, 286) = 6.28, p = .002, η2 = .04. Contrary to the primary hypothesis, and the results of the acceptance of sexual assault scale, the latent sexism ad condition (M = 3.29, SD = 1.36, CI = [2.99, 3.59]) did not show greater minimization of sexual coercion than the no sexism ad condition (M = 3.13, SD = 1.18, CI = [2.85, 3.41]), p = .46. The overt sexism ad condition (M = 2.61, SD = 1.05, CI = [2.35, 2.88]) did, however, show less minimization than both the latent sexism condition (p = .001; d = .56) and the no sexism condition (p = .009; d = .43). The main effect of gender was also significant, F(1, 286) = 10.59, p = .001, η2 = .036, indicating that men (M = 3.28) minimized the sexual coercion to a greater extent than did women (M = 2.74), d = .45. The interaction of ad type and gender was not significant, F(2, 286) = 1.46, p = .23, η2 = .01.
The 3 (ad type) × 2 (participant gender) ANOVA on the measure of victim blame for sexual coercion produced no significant main effect of ad type, F(2, 286) < 1.0; no main effect of gender, F(1, 286) = 3.05, p = .08; and no significant interaction, F(2, 286) < 1.0. Thus, there was no evidence that latent sexism increases victim blame for sexual coercion, and no indication of effects of latent sexism distinct from those of overt sexism on this measure.
Internal Validity
In sum, there is clear support for the prediction that ads with latent sexism produce greater acceptance of sexual assault compared with nonsexist ads. There is also evidence that the effects of latent sexism on acceptance of sexual assault and minimization of sexual coercion are distinct from the effects of overt sexism. Yet, because the ads in the latent, overt, and no sexism conditions differed in ways other than the type of sexist content, the internal validity of the ad effects remains a concern. To address this, participants’ ratings of the ads (see “Procedures”, above) were analyzed for differences that might account for the significant ad effects. Average ratings of interest, amount of product information, design, and persuasiveness were analyzed as DVs in separate 3 (ad type) × 2 (gender) unweighted means ANOVAs. Results indicated that men and women had different perceptions of the ads in the different conditions. Whereas women viewed the nonsexist ads more positively than the overtly sexist ads on all dimensions (i.e., more interesting, more product information, better design, and more persuasive), men rated the overtly sexist ads as the most interesting, but containing the least amount of product information. Notably, men did not differentiate between the latent and nonsexist ads, and the same was true for women, except on ratings of persuasiveness (latent ads were rated as less persuasive than nonsexist ads, but more persuasive than overtly sexist ads).
Despite the different ratings of the ads, however, none of the four dimensions was significant when included as a covariate in the 3 (ad type) × 2 (participant gender) ANOVAs of acceptance of sexual assault, trivialization of sexual coercion, or victim blame for sexual coercion (all ps > .15). Moreover, the significant ad effects reported above remained unchanged in all analyses with ratings of interest, amount of product information, design, and persuasiveness entered individually as covariates. Thus, these results provide support for the experiment’s internal validity by eliminating these dimensions as confounds from the interpretation of the ad condition effects on acceptance of sexual assault and minimization of sexual coercion.
Discussion
For years, media critic Jean Kilbourne (1999; Cambridge Documentary Films, 1979, 1987; Media Education Foundation, 2000, 2010) and other experts (Cortese, 2008; Goffman, 1979) have deconstructed ads to reveal symbolic, latent sexism, and have claimed that these ads contribute to sexist attitudes. Despite the widespread acceptance of these claims by scholars and educators alike (e.g., Belknap & Leonard, 1991; Conley & Ramsey, 2011; Kang, 1997; Krassas et al., 2003; Lindner, 2004; Millard & Grant, 2006; Plous & Neptune, 1997; Stankiewicz & Rosselli, 2008), the present research provides the first empirical support for their claims. More specifically, Kilbourne has argued that ad images of women’s “dismembered” legs, women in potentially dangerous locations, women’s bodies decorated as products, women presented alongside violent ad copy, and makeup possibly resembling a bruise contribute to greater acceptance of sexual assault. In our experiment, participants who saw ads with this content indicated greater acceptance of sexual assault than did participants who viewed nonsexist ads, even though the ads with latent sexism were not considered sexist within this population (as per the definition of latent sexism). These effects remained even after controlling for potential confounds stemming from the use of actual ads for the ad conditions. Therefore, even when the public does not recognize latent sexism in print ads, the content can lead to greater acceptance and tolerance for sexual assault.
Furthermore, the latent sexism ads not only increased acceptance of sexual assault relative to the nonsexist ads but also increased acceptance of sexual assault compared with the ads considered to be highly sexist. This was certainly unexpected given the extensive evidence of adverse effects resulting from exposure to overt sexism (APA, 2007). The lack of effects from overt sexism—the acceptance of sexual assault did not differ from the nonsexist ad condition—may reflect a broader shift in attitudes toward sexism in advertising. For example, Zimmerman and Dahlberg (2008) found that a contemporary sample of young, educated women perceived greater sexual objectification of women in advertising than did a comparable sample in 1991; however, the contemporary women also reported being less offended by the use of sexual objectification in advertising. The failure to replicate adverse effects of overtly sexist ads may also be a reflection of the atypical, diverse Canadian sample in this experiment. Canadians on average hold more liberal attitudes than citizens of the United States on issues of sexuality, including acceptance of premarital sex, same-sex marriage, legalized prostitution, and parents’ desire for sex education in schools (e.g., Bibby, 2006; McKay, Beyers, Voyer, Humphreys, & Markham, 2014), though there is insufficient evidence to indicate how these differences may impact responses to sexism in ads. Furthermore, our study did not include the large sample sizes from different ethnic categories that would be required to statistically examine their effects, so the role of our particular sample in producing these unique results remains speculative. Of course, the lack of effects from overt sexism may only appear exceptional because of the bias toward reporting significant results. In any case, the findings of this experiment clearly establish latent sexism as distinct from more obvious forms of sexism, and not simply a less extreme variation. In fact, these results suggest that ads with latent sexism have more deleterious effects than do ads with widely recognized, overt sexism.
These conclusions are of particular concern because of the apparent ubiquity of latent sexism in print ads. W. J. Rudman and Verdi (1993) found that half of the ads in their sample of fashion and fitness magazines showed women’s bodies dismembered in some way. Although it is impossible to determine the percentage of these ads meeting the criteria for latent sexism (i.e., not considered sexist by the lay public), even a small percentage would represent a substantial number of ads. In addition, Stankiewicz and Rosselli (2008) found a category of “distressed,” which included women appearing afraid or vulnerable, present in at least 2% of ads featuring women across all categories of magazines, and up to 3% of ads in women’s magazines. (These percentages are derived from their tabled values, using the “non-sexualized” category to more closely approximate the criteria for latent sexism.) These figures represent a considerable number of ads with latent themes of abuse, and that does not include other types of ads with the same message, such as images of women’s bodies as products, makeup resembling bruises, violent ad copy, and women held in vulnerable positions by men. Thus, there is clearly a need for researchers to expand their focus beyond overt sexism to the effects and prevalence of these different types of latent sexism.
In fact, the present study only examined the latent theme of women’s victimization, whereas experts have discussed other common themes unlikely to be recognized as sexism by the lay public. For example, Goffman’s (1979) categories of symbolic displays of gender stereotyping, such as feminine touch and licensed withdrawal, remain highly prevalent in print ads (Kang, 1997; Lindner, 2004). Similarly, ads portraying women with their mouths covered, which, according to Kilbourne (1999; Media Education Foundation, 2000, 2010), convey the underlying message that women should be silent, account for nearly 10% of ads found in popular women’s fashion/beauty magazines (Millard & Grant, 2006), and 17% of popular men’s magazines (Conley & Ramsey, 2011). Furthermore, in 1.5% of ads containing children, girls are presented as adult, sexual women (O’Donohue, Gold, & McKay, 1997), which Kilbourne and others (APA, 2007) have argued contributes to greater tolerance of sexual abuse of children (as do portrayals of grown women as little girls). Beyond the realm of advertising, others have argued that the glamorization of prostitution prevalent in pimp/ho fashion, television (e.g., Pimp My Ride), video games, and contemporary language are unrecognized forms of sexism (Coy et al., 2011). Given the results of the present study, there is clearly a need for research to determine whether other types of ads and cultural representations of women constitute other categories of latent sexism and, if so, whether they produce the effects claimed by experts.
Additional research is also necessary to determine why latent sexism increases acceptance of sexual assault, especially because overt sexism had no effect in our experiment. In particular, do the effects of latent sexism occur unconsciously, as Kilbourne (Media Education Foundation, 2000, 2010) claims? Although the present study was not designed to examine the mechanism by which latent effects occur, the findings are consistent with automatic activation of implicit stereotypes. Stereotypes may be primed automatically, producing unconscious effects, whether or not the individual endorses the stereotype (APA, 2007; Devine, 1989). Using a standard lexical decision task, researchers have shown that sexist commercials prime stereotypes of women (Davies et al., 2005; L. A. Rudman & Borgida, 1995) and many scholars have used priming as the theoretical framework to explain the effects of overt sexism in ads (for reviews, see APA, 2007; Ward, 2003).
There is no doubt the stereotype of woman-as-victim, even willing victim, is pervasive (Cortese, 2008; Kilbourne, 1999; Stankiewicz & Rosselli, 2008). Representations of sexualized violence against women permeate mainstream culture, from fashion, television, and video games to popular music (Coy et al., 2011; Horeck, 2014). Violence against women is, therefore, normalized and acceptance of rape myths is prevalent (Hayes, Abbott, & Cook, 2016; Suarez & Gadalla, 2010). Thus, images containing latent sexism, including women possibly injured, dismembered, “packaged” like a product, or in potentially dangerous situations, may promote acceptance of sexual assault by priming associations with the ubiquitous ideas about women as victims. When the sexism is unrecognized, it may be even more difficult to override or reject these automatic thoughts than if the sexism is obvious.
Of course, the methodology of the present study leaves open the possibility that participants in the validation pilot who rated the sexism in the stimulus ads recognized the sexist content in the latent ads, but for whatever reason did not rate the ads as sexist. Although this seems unlikely, given that even obvious sexism is not always recognized (Hansen & Hansen, 1988; Jennings-Walstedt et al., 1980; L. A. Rudman & Borgida, 1995; Schwarz et al., 1987), research employing different priming techniques and outcome measures is necessary to establish what is being activated by the latent content and whether it is indeed occurring unconsciously.
Unlike the acceptance of sexual assault, the latent sexism ads did not produce more sexist perceptions of the verbal sexual coercion incident. Compared with participants who saw nonsexist ads, those who viewed ads with latent sexism did not minimize the seriousness of the incident, nor did they blame the victim more. The different effects of ads for the two vignettes may simply be a reflection of the different perceptions of sexual assault and verbal sexual coercion more generally. In response to vignettes, participants are less likely to label verbal sexual coercion as rape (Luddy & Thompson, 1997; Russell, Oswald, & Kraus, 2011) and more likely to perceive it as controllable (Katz, Moore, & Tkachuk, 2007). Moreover, actual victims of verbal sexual coercion perceive their experiences as less serious than do victims of physical sexual assault (Abbey, BeShears, Clinton-Sherrod, & McAuslan, 2004). Therefore, it might be expected that sexist ads would affect perceptions of physical sexual assault and verbal sexual coercion differently.
In fact, our data suggest that perceptions of sexual assault and verbal sexual coercion are organized somewhat differently. When factor analyzed, questions about the seriousness of the assault and coercion incidents, together with questions about victim blame and perpetrator responsibility, revealed a single factor for the sexual assault (i.e., acceptance of sexual assault, comprised of seriousness, victim blame, and perpetrator responsibility items). The same questions about the sexual coercion, however, yielded two separate factors for seriousness of the incident and victim blame. Apparently, perceptions of verbal sexual coercion and sexual assault differ in important ways not yet examined in the literature.
Perhaps most unexpected was the finding that exposure to ads with overt, widely recognized sexism actually moderated the tendency to minimize the seriousness of the sexual coercion incident. Possibly, the exposure to blatant sexism in the ads activated thoughts about the damaging effects of sexism more generally, and this influenced perceptions of the severity and harm of the sexual coercion. It should also be noted that the apparent sensitizing effect of overt sexism may be unexpected, but it is not unprecedented, at least among women. In two studies, women expressed significantly less rape myth acceptance after viewing overtly sexist images (Lanis & Covell, 1995; Wyer et al., 1985). Notably, however, the overtly sexist ads in our study did not similarly moderate the amount of blame participants assigned to the victim of sexual coercion.
Of course, the lack of adverse effects from overt sexism may raise some questions about our manipulations and measures. With the use of actual ads as stimuli, it is impossible to rule out confounds; however, all effects remained after controlling for differences in ratings of interest, design, persuasiveness, and information in the different ad conditions. Furthermore, the use of images selected from actual media is standard practice (Davies et al., 2002; Davies et al., 2005; Lanis & Covell, 1995; Lavine et al., 1999; MacKay & Covell, 1997; Schwarz et al., 1987; Wyer et al., 1985). We also included a rigorous validation procedure to select ads with overt sexism, latent sexism, and no sexism. With an average sexism rating greater than six on a 7-point scale, there was widespread agreement that ads in the overt sexism condition were highly sexist. In addition, to prevent participants from recognizing the true purpose of the research, the experiment also included a deception similar to that used by others (Lanis & Covell, 1995; MacKay & Covell, 1997; Schwarz et al., 1987). The questions about the sexual assault and verbal coercion incidents were identical to, or similar to, items from previous studies in which participants were also exposed to scenarios of sexual assault (e.g., Schneider, 1992; Simonson & Subich, 1999; Wyer et al., 1985). We took the additional step of factor analyzing the questions and creating reliable measures based on the results. Therefore, we believe the manipulations and measures are methodologically sound, and the results are valid.
One noteworthy limitation, however, involves the diverse set of ads used in the latent sexism ad condition. The set of 10 latent sexism ads included images of women in potentially dangerous locations, the female body decorated as a product, a woman held in a vulnerable position by a man, makeup possibly resembling a bruise, violent ad copy, and dismembered legs. Although all of these images are alleged to contribute to the normalization and acceptance of abuse and assault against women (e.g., Kilbourne, 1999; Media Education Foundation, 2010), it is impossible to determine from our results whether all, or only some of these images contribute to increased acceptance of sexual assault. Vance, Sutter, Perrin, and Heesacker (2015) examined the effects of ads depicting women’s bodies morphed into products—similar to one of our latent sexism ads—and found no effect on acceptance of rape myths or likelihood of committing rape. Yet, it is unclear whether their ads depicting the body as product would meet the criteria for latent sexism. Therefore, further research is required to elucidate the effects of the different types of latent content included in our study.
Another limitation of our research is the reliance on a primarily first-year university sample. Obviously, there are questions about the degree to which our results would generalize to other age groups and demographics. Older individuals may be more likely to detect the sexism we identified as latent, and they may be less affected by it. Yet, many media literacy programs are delivered in high schools or at university, and their effectiveness is not dependent on age (Jeong, Cho, & Hwang, 2012). Young university students may, therefore, be better equipped to critically evaluate sexism in advertising than older, nonuniversity members of the public. In any case, there is a need for additional research to establish the generalizability of our findings. In the meantime, the findings of this study may provide somewhat greater assurance to educators who teach about Kilbourne’s (Media Education Foundation, 2000, 2010) ideas, to the media critics who bring attention to sexist content not recognized by the public, and to scholars who advance theory and research on various forms of latent sexism.
Footnotes
Acknowledgements
Thanks to the following for assistance in conducting the study: Oscar Astete, Aman Bassi, Gena Davies, Marissa de Waal-Malefyt, Joshua Guyer, Adam Paton, Jamie Rich, Shayna Rusticus, Niloufar Safari, Sara Salter, Maggie Wang, Nicole Weiss, and Mira Williams. Thanks also to anonymous reviewers for their valuable suggestions.
Authors’ Note
Sexist content in print ads that is not recognized as sexist by the public (e.g., images of women in potentially dangerous locations, women’s bodies decorated as products, makeup possibly resembling a bruise, incomplete bodies) can still lead to increased acceptance of sexual assault. In fact, this type of latent sexism had more adverse effects on perceptions of sexual assault than did sexist ads that were widely recognized as sexist by the public.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Thanks to the Minor Research Grant Committee and Office of Research and Scholarship at Kwantlen Polytechnic University for financial support.