Tuning out tenderness: the influence of gender and friends on U.S. adolescents’ emotional self-socialization via film selection and avoidance

Abstract

Teen boys often face peer pressures to avoid “feminine” emotions, such as tenderness. Media selections may reflect such pressures and constitute emotional self-socialization into traditional gender roles. An online experiment with 402 13- and 14-year-olds, based on Knobloch-Westerwick’s SESAM model, tested hypotheses about gendered selections/avoidance of hostile and tender content. Randomized to imagine watching a film alone or with friends, teens rated their interest in different film emotions and their likelihood of viewing eight films (pre-tested hostile or tender), then selected and viewed four trailers. Boys (vs. girls and nonbinary youths) gave higher ratings to hostile films and lower ratings to tender films. Baseline tender affect (lower in boys vs. girls and nonbinary youths) negatively predicted number of hostile trailers viewed which in turn negatively predicted post-test tenderness, consistent with emotional self-socialization. Imagined viewing condition did not moderate gender differences except in post hoc interactions with gender mix of friends.

Bem’s (1974) Sex Role Inventory (BSRI), still widely used, characterizes aggression, assertiveness, and dominance as masculine emotional attributes, and affection, tenderness, and warmth as feminine emotional attributes. Although readers may well find such categorizations to be problematic, it is notable that these are emotions that may be elicited and reinforced by media use. Within the framework of Knobloch-Westerwick’s (2015) Selective Exposure Self- and Affect-Management (SESAM) model, the central question of this study concerned young teens’ film choices as a form of emotional self-socialization into (or out of) traditional gender roles.

Eisenberg and colleagues define emotional socialization as the process by which children learn to experience, express, and regulate their emotions “in a manner consistent with the socializers’ beliefs, values, and goals about emotion and its relation to individual functioning and adaptation in society” (Eisenberg et al., 1998b, p. 317). Eisenberg et al.'s (1998a) model focused mostly on parents as the primary agents of socialization for young children, but other work, reviewed below, has pointed to early adolescence as a moment of important developmental changes in peer influence and self-socialization via media use. We conducted an online experiment to examine young teens’ choices of films featuring “masculine” hostile emotions and “feminine” tender emotions when imagining watching either alone or with friends and if these selections reinforced gender differences in related emotional attributes.

Media as a resource for self-socialization

Larson (1995) argued that teens rely on private media use to navigate negative emotions that tend to intensify during adolescence, but also noted a shift from co-viewing with family to peer-oriented media use in early adolescence (Larson et al., 1989). Arnett (1995) proposed that media use (more than family and school) offered teens autonomy to undergo intense, vicarious experiences; select role models; and develop a distinctive, shared youth culture. Thirty years later, approximately half of teens report “almost constant” media use afforded by smartphones (Pew Research Center, 2023). Despite considerable research on the implications for teen mental health, little work has examined teen media use as a form of gendered emotional self-socialization and the ways it might be influenced by friend groups.

Like Arnett (1995), Knobloch-Westerwick (2015) considered media selections in terms of self-socialization. Whereas Arnett discussed identity socialization primarily in terms of the media exemplars teens select for themselves as behavioral models, the SESAM model proposes that individuals select media based on affective considerations and to activate or maintain salient self-concepts. It posits an iterative cycle whereby individuals’ working self-concepts and affect shape media selections and interpretations, which in turn influence self-concepts and affect, potentially compounding over time. Individuals are generally expected to prefer content that reinforces their pre-existing dispositions, self-concepts, and beliefs, though situational needs for self-enhancement or self-improvement might lead them to seek out models that afford flattering downward comparisons or inspiring upward comparisons (respectively). To illustrate, Knobloch-Westerwick (2015) offered the example of reading fitness magazines to activate one’s self-concept as an athlete, maybe increasing motivation to exercise (i.e., self-improvement).

Knobloch-Westerwick and colleagues have studied media self-socialization processes for various domains, including gender. In one experiment, Knobloch-Westerwick and Hoplamazian (2012) examined the impact of self-selected magazine reading on undergraduates’ pre-post changes in femininity and masculinity on the BSRI. Women spent more time reading female-typed (vs. male-typed or neutral) magazines (e.g., Glamour), which increased femininity scores, and men spent more time with male-typed magazines (e.g., Men’s Health), which predicted decreased gender differentiation, operationalized as femininity minus masculinity and thus, according to the authors, “indicating a shift toward masculinity” (p. 379). Knobloch-Westerwick et al. (2020) measured college women’s salience of their future selves as a parent, working professional, and romantic partner and found that salience predicted proportional time spent (over four daily sessions) reading parenting, business, and beauty magazines, respectively. Time spent reading beauty magazines predicted increased salience of their future self as a romantic partner four days later. These experiments offer some evidence of young adults’ self-socialization into gendered self-concepts through gender-typed magazines.

The current study seeks to extend the explanatory range of SESAM to include the internalization of affective or emotional attributes linked to salient self-concepts. To build on Knobloch-Westerwick’s example, perhaps athletes might also seek out media that reinforce emotional attributes (e.g., aggression, competitiveness) that they consider relevant to their athletic self-concept and abilities, and avoid media with emotional qualities (e.g., gentleness, warmth) that undercut their athletic-related goals. In our case, we focused on affective media selections as potentially resulting from and reinforcing gendered self-concepts. To replace the athlete example, we can imagine boys who prize a traditional notion of masculinity seeking out hostile content and avoiding tender content to conform with their views of what it is to be a man.

Thus far, research on the “affect” component of SESAM has primarily treated it as a state outcome or moderator distinct from self-concept. For example, Luong et al. (2021) measured positive and negative affect with items (e.g., scared, hopeful) unrelated to the self-concepts activated by the stimuli (e.g., career, friendship), and thus the findings that identity-related upward social comparisons increased positive affect offer evidence only of mood-type outcomes. Our study, conversely, seeks to probe gender as a motivation for affective media selections, thus further internalizing the explored emotions within one’s gendered self-concept. Further, we explore this process of emotional self-socialization during a developmental period marked by elevated felt pressure (from the self and from peers) to conform to gender norms, including affective experiences and displays, just as teens are transitioning away from parental mediation toward more private and socially oriented media use. Below, we briefly review the broader literature on gender, emotion, and peer groups before turning to media research in this area.

Gender identity and emotion

A considerable body of work finds gender differences in men and women’s reports of their emotional attributes. A recent meta-analysis of 937 samples comprising 254,465 participants (Hsu et al., 2021) found that women (vs. men) rated themselves higher on “feminine” communal qualities (e.g., affectionate, tender, warm; Hedge’s g = −.56), and men (vs. women) rated themselves higher on “masculine” agentic qualities (e.g., aggressive, assertive, dominant; Hedge’s g = .40) based on the BSRI (Bem, 1974) and Personal Attributes Questionnaire (Spence & Helmreich, 1978). Similar patterns have emerged in research on beliefs about emotion-related gender differences. Plant et al. (2000) found U.S. undergraduates rated women (vs. men) as more often experiencing and expressing 12 of 19 emotions (e.g., love, sadness, sympathy); men were rated as more frequently experiencing and expressing anger and pride. Durik et al. (2006) asked adults to rate norms for their own race/ethnicity and found that, across groups, women (vs. men) were expected to display a wider array of emotions, including love, sympathy, sadness, and fear.

Male avoidance of “feminine” emotions, such as sadness and tenderness, has long been a focus of research on traditional masculinity ideology (reviewed in Berke et al., 2018). Vandello et al. (2008) framed this avoidance in terms of the “precarity” of masculinity, having found that participants were more likely to agree that manhood (vs. womanhood) “is hard won and easily lost” (p. 1328). Relatedly, O’Neil (1981) argued that men experience considerable gender role conflict (ie, stress) often manifested as a fear of feminity. In a review of 232 studies, he noted that one key domain of such gender role conflict was restrictive emotionality, whereby men reduced their emotional expression because of perceptions of emotionality as “feminine, weak, and not part of being human” (O’Neil, 2008, p. 419). Restrictive emotionality has been linked to negative mental health outcomes (reviewed in Exner-Cortens et al., 2021), including decreased resilience (Galligan et al., 2010) and increased suicidality (Jacobson et al., 2011).

Research has found adolescence to be a formative period for boys’ socialization into restrictive emotionality (Rogers et al., 2021). In a qualitative longitudinal study, Way et al. (2014) found that most boys entered adolescence resistant to masculine norms, but that resistance often declined from pre- to late-adolescence, accompanied by a turn toward toughness and stoicism. Reigeluth et al. (2016) found mixed results observing 10th–12th grade boys in a dyadic disclosure task: about a third were emotionally expressive in their disclosures and met with affirming responses, but the remainder were either defensive, guarded, or disengaged when disclosing, and their partners typically responded in ways that were unsupportive or mirrored their emotional tone, including dismissing difficult situations with joking and laughter. Relatedly, nearly all of the boys interviewed by Oransky and Marecek (2009), most of them 15 years old and in 10th grade, reported restricting their own emotional displays and encouraging the same of their friends, with violations teased as “gay” or “girly.” A majority saw this enforcement of emotional restrictivity as helping each other get through things by avoiding difficult emotions.

Work not focused specifically on emotions has found pre-adolescent and adolescent boys and girls both report strong gender-related pressures, and while boys have generally reported experiencing greater gender pressures than have girls (e.g., Cook et al., 2019 ; Nielson et al., 2020), these pressures may vary by source and type. For example, studies from Jackson and Bussey (2020) and Jackson et al., 2021) found that boys (vs. girls) felt greater pressure to conform to own-gender behaviors and avoid other-gender behaviors—girls, conversely, did not feel pressured to avoid other-gender (i.e., masculine) behaviors—and felt more pressure around feminine behaviors from peers than from parents. Kornienko et al. (2016) additionally found felt gender pressures to spread within peer networks, such that having friends who reported elevated pressure for gender conformity predicted participants’ increases in felt pressure over time.

Much of this research on felt gender pressures was conducted with early adolescents, and it is unclear if and how these felt pressures change with age (e.g., Hoffman et al., 2019; Tanti et al., 2011). However, the shift from tween (8- to 12-years-old) focus on television toward teen (13- to 18-years-old) focus on social media and online videos (Rideout et al., 2022) supports Larson et al.’s (1989) theorizing of early adolescence as a period of transition from family- to peer-oriented and private media use. As such, young teens report feeling strong pressures from friends and from themselves (i.e., internalized pressures) to conform to certain gender norms just as parents are becoming decreasingly involved in their media selections. It remains unclear whether young teens’ media selections tend to reinforce or, alternatively, offer relief from the gender pressures that characterize this developmental stage, and whether their choices suggest felt pressures from friends to conform to traditionally gendered affect.

Media selections as emotional self-socialization into gender roles

Two meta-analyses on attractions to violent, aggressive, and hostile media have found gender differences, though in both instances, effect sizes were heterogeneous, suggesting the need to identify moderators. Hoffner and Levine (2005) found that male (vs. female) viewers and those who were more aggressive and lower in empathic concern tended to report more enjoyment of horror films. Weaver (2011) found similar results with violent content more generally: male and aggressive individuals were moderately more likely to view and enjoy violent media than were female and non-aggressive individuals. For example, four-year-olds expressed gendered preferences for violent books (Collins-Standley & Gan, 1996) and videotapes (Knobloch et al., 2005), and adolescent boys (vs. girls) reported greater interest in physically aggressive content (van der Wal et al., 2020) and greater preference for slapstick, disparaging, and aggressive humor, with gender differences increasing with age (from 10 to 17) (van der Wal et al., 2022).

There is also some evidence of media selections reinforcing gender-normative tendencies toward hostility and aggression. In a longitudinal study with adolescents, Slater et al. (2003) found that trait aggression predicted concurrent violent media use, which predicted aggression at a subsequent wave, and that boys showed slightly steeper increases in aggression compared to girls. In an experimental study, Knobloch-Westerwick and Alter (2006) demonstrated gender differences in U.S. undergraduates’ media selections following an anger induction. After “failing” a bogus test, half of participants were told they’d have a chance to confront the test’s designer after a period of reading news stories. Within this condition, men spent more time reading negative news leading up to the anticipated confrontation, whereas women spent more time on positive news; no such gender differences emerged among participants not anticipating a confrontation. The authors interpreted these selections as reflecting gender norms for managing interpersonal conflict, with men seeking to reinforce anger and women seeking to assuage it.

Evidence is slightly more mixed with regard to gender differences in response to content eliciting “feminine” emotions of tenderness or sadness. Oliver (2008) defined tenderness as “feelings associated with human connectedness and…vulnerabilities, [including] empathy, warmth, kindness, and connection” (p. 48). In surveys, women (relative to men) reported higher interest in watching “warm-hearted” TV programs (Mares et al., 2016), gave higher ratings of love and kindness in a film they had found meaningful (Janicke & Oliver, 2017), and were more likely to report having experienced eudaimonic affect (lump in throat, tears, feeling “light and bouncy”) in response to a meaningful film (Oliver et al., 2012). However, Oliver and Raney (2011) found no gender differences in eudaimonic versus hedonic media selection motivations.

Relatedly, Oliver (1993) found that women were more likely than men to seek out and enjoy sad films, but in a subsequent study (Oliver et al., 2000) found that these gender differences varied by plotlines, such that women showed greater interest than men for a movie about a friend diagnosed with leukemia but not for a movie about a star basketball player paralyzed in an accident. Oliver (2008) found that, regardless of gender, tender affective states (whether induced, observed, or hypothetical) increased anticipated enjoyment of sad films. However, Greenwood (2010) found that, when induced to feel sadness, men were more likely to select dark comedies, whereas women were more likely to select romantic dramas. That is, in line with social gender norms, men chose content that buffered sadness through comedic relief, whereas women chose content that, based on genre conventions, risked exacerbating sadness. Relatedly, in a two-month experience-sampling study, 7- to 17-year-old boys, but not girls, were more likely to watch “fun” content when upset, angry, sad, or nervous (Carpentier et al., 2008).

Most of the above studies involve undergraduates and adults, the exception being the work done on adolescents’ self-socialization into aggression. Largely missing from this research, however, is consideration of the role of peers during this developmental period, including their impact on selections in the context of co-viewing. Research on co-viewing, as synthesized by Tal-Or (2021), typically focuses on physical co-presence moderating media effects, for example, decreasing narrative transportation by dividing viewers’ attention, or amplifying affective responses through spontaneous behavioral mimicry. Co-viewer characteristics may also moderate effects, as in Zillmann et al.’s (1986) experiment showing that undergraduates reported heightened enjoyment of a horror film when co-viewing with an other-gendered confederate who responded to the film in gender-stereotypical ways (e.g., distress for the female confederate). Other types of gendered co-viewing experiences, such as boys’ performances of emotional stoicism in the presence of their male friends, might similarly amplify emotional socialization into toughness.

But even before these potential effects, anticipated co-viewing might influence which media teens select. Arnett (1995) suggested that adolescents use media to connect with each other and establish shared values and interests. If peers share certain values and assumptions about the gendering of affective interests, co-viewing would likely increase conformity to these assumptions. To rephrase this within a SESAM framework: if gender predicts affective media selections (van der Wal et al., 2020) and is also a salient part of social identity for young adolescents (Tanti et al., 2011) that can be influenced by and spread within peer networks (Kornienko et al., 2016), then co-viewing with friends might increase teens’ motivation to select gender normative affective experiences while avoiding affective experiences that violate gender norms and therefor risk social alienation. Weber (2013) found evidence of the latter interviewing German teens about their co-viewing behaviors: boys reported avoiding soap operas, which they characterized as being only for girls, or “pussy TV,” as one interviewee notably called it, mirroring the language boys use to police each others’ masculinity and emotional expression (Reigeluth & Addis, 2016). This line of theorizing exceeds the current predictive scope of the SESAM model, which to date has considered only private media selections. Manipulation of social viewing contexts, even in the hypothetical, could help illuminate how exogenous factors activate the self in ways that may contribute to long-term emotional self-socialization.

Summary, research gap, and the current study

To summarize, research suggests gender differences in the emotional attributes that adults ascribe to themselves and that are seen as normative. By early adolescence, these differences are evident not only in peer group interactions but also in internalized values, including, for boys, felt pressure to avoid “feminine” emotional displays. Although media socialization also occurs during childhood, the transition toward more private and peer-oriented media usage during early adolescence, combined with these heightened internalized and externalized pressures about gendered emotions, makes it important to understand how young teens’ media selections may reinforce and further internalize the gendering of these emotions.

Although there is a sizable body of work on gender differences in affective selections of media content, much of that work has focused on undergraduates and adults, especially when examining sad, tender, or mixed-affect content. This article aims to advance the explanatory and predictive power of the SESAM model by probing the link between the self and affective traits within emotional self-socialization while also considering how imagined co-viewing with friends might increase salience of the gendered self in ways that moderate emotional self-socialization behaviors. We conducted an online experiment with 13- and 14-year-olds, randomizing them to evaluate and select content to hypothetically view either alone or with friends. We chose this narrow age range to limit other age-related factors as an additional source of variance while zooming in on a key developmental moment where heightened gender pressures intersect with increased media selection autonomy. Given findings about boys’ felt pressure to avoid “feminine” emotions, we compared boys to girls and nonbinary youths.

Following baseline measures, teens evaluated what emotions they would like to experience while viewing a film. We anticipated significant gender differences, such that:

H1: Boys (vs. girls and nonbinary youths) will report (a) higher interest in hostile emotions and (b) lower interest in tender emotions.

We expected that these effects would be moderated by imagined viewing condition, given the significance of peer influence on gender norms, and additionally hypothesized that:

H2: Gender differences will be stronger when imagining viewing with friends (vs. alone) for interest in (a) hostile emotions and (b) tender emotions.

Teens were then presented with four pairs of films (one “hostile,” one “tender”) and were asked to rate how likely they would be to watch each one. Whereas the previous measure reported interest in discrete emotions not linked to any specific film, this measure asked about likelihood of viewing actual films, pretested to reflect these emotions. Parallel to H1 and H2:

H3: Boys (vs. girls and nonbinary youths) will be (a) more likely to watch hostile films and (b) less likely to watch tender films.

H4: Gender differences will be stronger when imagining viewing with friends (vs. alone) for reported likelihood of watching (a) hostile films and (b) tender films.

For each pair, participants selected the film they would most want to watch (alone or with friends) then viewed the trailer of the selected film. Unlike the above dependent variables, this selection forced a dichotomous choice between hostile and tender content. We predicted the same main effect and interaction as above:

H5: Boys (vs. girls and nonbinary youths) will view a higher number of hostile film trailers.

H6: Gender differences will be stronger when imagining viewing with friends (vs. alone) for number of hostile trailers viewed.

Finally, the SESAM model suggests media selections might reinforce emotional attributes; it is this cycle which constitutes media-based emotional self-socialization. Based on this, we predicted hostility and tenderness would change from pre- to post-test in the direction of the content selected, such that:

H7: Number of hostile trailers viewed will be associated with (a) increases in hostility and (b) decreases in tenderness.

Hypotheses, questionnaire, and planned analyses are preregistered at https://osf.io/7aynj/. Deviations from preregistered hypotheses and planned analyses are explained with additional analyses in Supplementary Appendix D. Data and syntax are also available on OSF.

Methods

Pretest: selecting the stimuli

Twelve film trailers were pretested with U.S. university undergraduates (N = 120), who watched two randomly assigned trailers (one “hostile” and one “tender”) and rated how much (1 not at all, 5 very much) the film would express different emotions. Anger and aggression were averaged to form a hostile score; love, sadness, and tenderness were averaged to form a tender score. Four tender films were chosen based on high tenderness (M ≥ 4) and low hostility (M ≤ 2.5): CODA (2021), Five Feet Apart (2019), Me and Earl and the Dying Girl (2015), and Rocks (2019). Four hostile films were chosen based on high ratings of hostility (M ≥ 4) and low ratings of tenderness (M ≤ 2.5): The Devil All the Time (2020), The Protégé (2021), Runt (2020), and Spree (2020). Films from each category did not significantly differ in ratings of excitement. Posters, keywords, and brief descriptions for these films (see Supplementary Appendix Figures G1 and G2) were also pretested for these same characteristics with an additional sample (N = 56). Just before the main study, CODA won the Academy Award for Best Picture; however, our results indicated that teens were no more likely to have previously seen this film than the others. See Supplementary Appendix B for additional information and pretest results.

Main study

Power analysis

Power analysis (using G*Power) for a repeated measures ANOVA with four groups (boys vs. girls and nonbinary; alone vs. with friends) and two uncorrelated repeated measures (hostile vs. tender) with α = .05, power = .95, and a small (f = .15) expected effect size suggested a minimum sample size of 388.

Participants

A sample of 431 13- and 14-year-olds from the U.S. was recruited via Qualtrics (an online polling company). Of these, 27 were dropped for failing Qualtrics’ quality checks (for missing or straight-line responses and/or low response times) and/or either of two attention check items. Two participants did not report their friends’ genders, which was used for post hoc analyses, and thus were removed from the study, resulting in N = 402. As approved by the authors’ Institutional Review Board, parents granted consent via the online consent page, then teens gave assent and completed the study. Teens were compensated in accordance with Qualtrics’ policy. Sample quotas were used to get roughly even distributions for age (49.5% 13-year-olds vs. 50.5% 14-year-olds), race/ethnicity (54.2% youths of color vs. 45.8% white), and gender (52.2% male, 46.3% female, 1.5% nonbinary). For details, see Supplementary Appendix A.

Procedure

Teens first reported demographics then their baseline emotional attributes. Teens were randomized within gender into one of two conditions (alone vs. with friends): “Imagine that you are [hanging out with your friends/home alone] and can watch a movie [together/by yourself].” They rated how interested they would be in that movie containing various emotions, then evaluated four randomized pairs of films, with one pre-tested as “hostile” and one pre-tested as “tender.” For each pairing, teens first rated their likelihood of watching each film, then selected the one they’d most likely watch and watched the trailer for that film. Prompts reminded participants of their imagined viewing condition (alone vs. with friends) at each stage of stimuli evaluation. Following evaluation and exposure, participants re-rated their emotional attributes.

All teens also answered questions about their friend group and recent media use. To strengthen the experimental manipulation, those assigned to imagine viewing with friends answered the questions about their friends immediately after the viewing instructions (i.e., after baseline measures and before any of the film-related ratings) and completed the media use questions at the end of the survey. Those in the solo-viewing condition answered the media use questions after the viewing instructions and answered questions about their friend group at the end of the survey. All teens completed manipulation check items for viewing condition and emotional tone of films after completing post-exposure measures.

Measures

The full questionnaire can be viewed on OSF. Additional measures, including background variables that did not differ by condition and were therefore not included, such as baseline mood and energy, are described with relevant statistics in Supplementary Appendix B.

Pre- and post-test emotional attributes

At both pre-test and post-test, six items assessed hostility: aggressive, angry, assertive, dominant, mean, and tough (α pre-test = .72, post-test = .80). Five items assessed tenderness: affectionate, compassionate, sympathetic, tender, warm (α pretest = .84, post-test = .86). Items were selected based on the BSRI (Bem, 1974) and Plant et al.’s (2000) gender-stereotyped emotions with structure confirmed using principal components analysis with varimax rotation. “Sadness” loaded on both measures and reduced reliability for tenderness and thus was excluded.

At pre-test, the 12 focal items were interspersed among 28 other items, including the rest of the short-form BSRI (Bem, 1981) and 15 filler items (e.g., happy, analytical, depressed). Teens were instructed to “rate how well these items describe you” (1 not at all, 7 very much). At post-test, after viewing four trailers, teens read, “we'd like you to rate yourself once more based on how you're feeling right now,” and re-rated themselves on 26 items, including the focal 12.

Gender mix of friends

We asked participants, if watching a movie with friends, how many friends there would be, from zero (0) to six or more (⁠$M=2.77, SD=1.42$⁠), and the number who were boys, girls, nonbinary, about the same age, older, and younger. Further questions asked from where they knew them (e.g., school, neighborhood), words to describe them, and typical shared activities.

The pre-registered plan was to use the percent of same-gender friends as a covariate, but this variable interacted with gender and condition and thus was instead explored as an additional moderator post hoc, as described below. Given a bi-modal distribution, (see Supplementary Appendix B Figure B1), a dichotomous variable was created to distinguish between teens who had all same-gender friends (n = 164) and those whose friends were mixed- or other-gender (n = 238).

Interest in film emotions

Having been randomized to imagine watching alone or with friends, teens were asked, “how interested would you be in watching a movie that is…” and rated 11 synonym pairs (e.g., gentle/tender) (1 not at all interested, 5 extremely interested). Items were chosen based on gender stereotypes of emotions. Principal components analysis with varimax rotation indicated a “hostile” component (angry/enraging, aggressive/hostile, brutal/violent, α = .85) and a “tender” component (gentle/tender, loving/caring, moving/touching, α = .85). Items were averaged for each component. “Sad/heartbreaking” cross-loaded on these two components and was omitted.

Rated likelihood of viewing hostile and tender films

Teens saw four randomized pairs of tender and hostile films. For each film, we presented the film poster, keywords based on Netflix search terms (e.g., heartfelt, violent, ominous), and a brief description, modified from its IMDb page (Supplementary Appendix Figures G1 and G2). Teens rated their likelihood of watching each film (alone or with friends) on a 3-point scale: No (1), Maybe (2), or Yes (3). Ratings were averaged across the four hostile films (α = .72) and four tender films (α = .80).

Number of hostile trailers viewed

For each film pairing, teens selected which they would most likely watch and viewed the trailer for that film. The number of hostile trailers viewed was summed, ranging from 0 to 4.

Results

Variable construction, principle components analyses, and reliability testing were performed in SPSS (version 29.0.1.0). Analyses of variance (ANOVAs) were performed for hypotheses 1 through 6 using the R package rstatix (0.7.2) (Kassambara, 2023). Simple effects comparisons were adjusted familywise following Holm’s (1979) Sequential Bonferroni Procedure. Hypothesis 7 was tested as a path analysis using lavaan (0.6–16) (Rosseel, 2012). Because we use a mixed design with observed and manipulated variables, we report effect size for ANOVAs as generalized eta squared (Olejnik & Algina, 2003).

Manipulation and randomization checks

Following the post-test measures, participants completed two manipulation checks.

Film tone check

Teens were shown four randomly selected film posters from the study (two hostile, two tender). For each, teens answered whether they had watched the trailer or just read the description and rated their perception of how much (1 not at all, 5 very much) the film would convey: anger, aggression, excitement, love, sadness, tenderness. Hostile (vs. tender) films were rated significantly higher on aggression and anger and lower in tenderness and love, both by about 1.5 points, with smaller but significant differences in sadness (higher for tender films) and excitement (higher for hostile films). Details are in Supplementary Appendix B.

Viewing condition check

Participants indicated if they had imagined selecting content to watch alone or with friends. Of those assigned to the alone condition, 168 (84.8%) chose “alone.” In the with friends condition, 187 (91.7%) chose “with friends.” Thus, 47 teens (11.7%) did not respond based on their assigned condition. Given that dropping subjects who fail a manipulation check can bias results (Aronow et al., 2019), we ran analyses twice: first based on assigned condition, then based on reported condition from the manipulation check. There were some minor differences in simple effects of gender within post hoc interactions between condition and gender mix of friends, but omnibus results did not differ when using reported rather than assigned condition. We report results for assigned condition here and results for reported condition in Supplementary Appendix F.

Randomization checks

There were no significant differences between conditions in distributions of background variables. For gender, only race/ethnicity varied significantly: more boys (53%, n = 113) were White than girls and nonbinary youths (38% n = 72), χ² = 10.13, p = .001. Including race as a covariate increased heteroskedasticity and reduced power without impacting significant results (see Supplementary Appendix G); thus, race was excluded from analyses.

Between-film differences

To check for possible confounds related to the gender of characters featured in each film, we conducted additional analyses comparing participants’ responses to each film. There were significant between-film differences, but none that supported character gender as a confound. A detailed reporting can be found in Supplementary Appendix G.

Descriptive statistics

As shown in Table 1, the overall means for rated interest in tender film emotions, reported likelihood of viewing hostile and tender films, and number of hostile trailers viewed hovered around the midpoint; interest in hostile film emotions was slightly lower. State tenderness was significantly lower for boys (vs. girls and nonbinary youths) at baseline, t(400) = 2.20, p = .028, d = .22, and posttest, t(400) = 3.51, p < .001, d = .35; gender differences for state hostility were not significant at baseline, t(400) = –1.62, p = .107, d = .16 or posttest, t(400) = –1.86, p = .064, d = .19. All measures of hostility positively correlated with each other (r > .32), as did measures of tenderness (r > .30). See Supplementary Appendix C for correlations with gender differences.

Pre-registered tests of H1–H4: emotional preferences in films

To test the hypotheses about gender effects and interaction with imagined viewing condition on interest in film emotions (H1–2) and interest in specific films (H3–4), we ran two Mixed Model ANOVAs with gender (boys vs. girls and nonbinary youths) and assigned viewing condition (alone vs. with friends) as between-subject factors and film emotion/tone (hostile vs. tender) as the within-subject factor. Table 2 summarizes these findings, with results in the top half of column one for H1–H2 and column two for H3–4. Results did not differ when excluding gender expansive (e.g., nonbinary, trans) teens from the sample (see Supplementary Appendix G).

Interest in hostile and tender emotions in films

Teens reported their interest in tender and hostile emotions if they were to view a film (alone or with friends). There was a main effect of emotion: rated interest in tender emotions (M = 3.12, SD = 1.18) was higher than interest in hostile emotions (M = 2.29, SD = 1.11), F(1, 398) = 118.93, p_adj < .001, $ηG2$ = .120.

H1 gender differences

H1 predicted that boys (vs. girls and nonbinary youths) would report (a) higher interest in hostile emotions and (b) lower interest in tender emotions in films. Consistent with H1, and as shown in Figure 1A, gender and emotion interacted: boys showed higher interest in hostile emotions compared to girls and nonbinary youths, F(1, 400) = 8.85, p_adj = .003, $ηG2$ = .022; girls and nonbinary youths showed higher interest in tender emotions compared to boys, F(1, 400) = 37.58, p_adj < .001, $ηG2$ = .086. However, within gender, boys reported significantly more interest in tender (vs. hostile) emotions, F(1, 209) = 10.47, p_adj = .001, $ηG2$ = .021. There was also a main effect of gender, F(1, 398) = 5.11, p_adj = .024, $ηG2$ = .006, with girls and nonbinary youths reporting higher interest in both film emotions (M = 2.80, SD = 1.26) compared to boys (M = 2.62, SD = 1.17).

H2 condition effects

H2 predicted stronger gender differences when imagining viewing with friends (vs. alone). In fact, there were no significant effects of condition, including the predicted interactions with gender and repeated measure of emotion. H2 was not supported.

Likelihood of viewing hostile and tender films

Teens rated their likelihood of viewing films pre-tested as hostile or tender. There was a main effect of film tone, with teens reporting a higher likelihood of viewing tender (M = 2.15, SD = .63) versus hostile films (M = 2.04, SD = .59), F(1, 398) = 10.42, p_adj = .001, $ηG2$ = .010.

H3 gender differences

H3 predicted that boys (vs. girls and nonbinary youths) would report being (a) more likely to watch the hostile films and (b) less likely to watch the tender films. Consistent with H3 and as shown in Figure 1B, gender and film tone interacted: boys gave higher ratings for the hostile films compared to girls and nonbinary youths, F(1, 400) = 6.23, p_adj = .013, $ηG2$ = .015; girls and nonbinary youths gave higher ratings for the tender films compared to boys, F(1, 400) = 37.25, p_adj < .001, $ηG2$ = .086.

Unlike interest in film emotions, boys reported being more likely to view the hostile (vs. tender) films, F(1, 209) = 7.22, p_adj = .008, $ηG2$ = .012. There was again a main effect of gender, F(1, 398) = 5.84, p_adj = .016, $ηG2$ = .008, with girls and nonbinary youths (M = 2.16, SD = .61) reporting higher likelihood of watching all films compared to boys (M = 2.04, SD = .61).

H4 condition effects

H4 predicted that those imagining viewing with friends would show greater gender differences in reported likelihood of viewing hostile and tender films. Again, there were no main or interaction effects of assigned viewing condition. H4 was not supported.

H5–6: number of hostile trailers viewed

For each pair of films (one hostile, one tender), teens made a binary selection of which of they would most want to watch (alone or with friends) and watched the trailer for that film. Number of hostile trailers viewed was the dependent variable. Test statistics for the two-way ANOVA (gender × condition) for preregistered H5–6 are in the top half of the final column of Table 2.

H5 gender differences

Supporting H5, and as shown in Figure 1C, boys viewed a higher number of hostile trailers compared to girls and nonbinary youths.

H6 condition effects

H6 predicted that gender differences would be stronger when imagining viewing with friends (vs. alone). As with other dependent variables, the predicted interaction was not significant, nor was there a significant main effect of condition.

H7: changes in state hostility and tenderness

After watching all four trailers they had selected, participants again rated their state hostility and tenderness to assess changes from baseline. H7 predicted that the number of hostile trailers viewed would predict (a) increases in state hostility and (b) decreases in state tenderness. To reflect the SESAM model, we ran a path analysis with the number of hostile trailers viewed mediating changes in hostility and tenderness from baseline to post-exposure, with gender predicting baseline measures. The model showed good fit, χ²(5) = 4.89, p = .429, CFI = 1.00, RMSE = .000 (90% CI: .000–.067), SRMR = .015.

As shown in Figure 2, male gender and baseline hostility and tenderness all predicted number of hostile trailers viewed. Male gender also predicted baseline tenderness but not hostility. Contrary to H7a, the number of hostile trailers viewed did not predict post-test hostility (controlling for baseline hostility). However, consistent with H7b, the number of hostile trailers viewed negatively predicted post-test tenderness, (controlling for baseline tenderness). This resulted in a significant indirect effect, β = .03, p = .019, such that higher baseline tenderness predicted fewer hostile trailers viewed which predicted higher post-exposure tenderness. There was also an indirect effect of gender on post-exposure tenderness, β = −.17, p < .001, such that boys watched more hostile trailers, resulting in lower tenderness. However, the complete indirect pathway from gender → baseline tenderness → number of hostile trailers viewed → post-exposure tenderness was not significant, β = −.01, p = .114. In short, number of hostile trailers viewed mediated affective changes for tenderness but not hostility.

Post-hoc analyses with gender-mix of friends

As noted earlier, we had asked all teens about the types of friends with whom they might watch a film (number, ages, genders). We anticipated treating gender mix of friends as a covariate but found that it interacted with gender and condition. In retrospect, it made sense that gender-mix of friends might moderate the effect of condition. In an experiment with young teens, Leszczynski and Strough (2008) found increases in boys’ BSRI feminity scores when they performed a task with a female (vs. male) confederate. As post-hoc exploration, we added gender-mix of friends (same vs. mixed/other) as a third between-subjects factor (in addition to gender and condition) for tests of H1–6, acknowledging that the study was not powered for these analyses. Results are described briefly below and summarized in the bottom half of Table 2. Supplementary Appendix E includes additional analyses, tables, and figures.

Interest in hostile and tender emotions in films

There was a significant interaction between gender, condition, gender-mix of friends, and film emotion type. For interest in hostile emotions, the gender difference (i.e., boys’ greater interest) was significant only for those imagining watching with all same-gender friends, F(1, 72) = 7.88, p_adj = .036, $ηG2$ = .099, but not for those with same-gendered friends watching alone or for those with mixed- or other-gendered friends in either condition (F > 2.52, $ηG2$ < .028). For interest in tender emotions, gender differences were significant for teens with same-gender friends (vs. mixed-/other-gender) but were stronger for those imagining watching with those friends, F(1, 72) = 41.98, p_adj < .001, $ηG2$ = .368, versus those imagining watching alone, F(1, 88) = 9.00, p_adj = .028, $ηG2$ = .093.

Likelihood of viewing specific hostile and tender films

There was a significant interaction between gender, condition, gender-mix of friends, and film tone to predict rated likelihood of watching hostile or tender films. For hostile films, gender differences were not significant regardless of condition and gender-mix of friends. For tender films, there were significant gender differences for teens with same-gender but not mixed- or other-gender friends, and again these differences were stronger for those imagining watching with those friends, F(1, 72) = 45.76, p_adj < .001, $ηG2$ = .389, compared to those imagining watching alone, F(1, 72) = 9.21, p_adj = .021, $ηG2$ = .095.

Put another way, participant gender explained about four times as much variance in ratings of tender emotions and tender films for teens imagining viewing with same-gender friends than for teens imagining watching with mixed- or other-gender friends.

Number of hostile trailers viewed

The number of hostile trailers viewed did not differ significantly by participant gender among teens imagining viewing with other- or mixed-gendered friends, F(1, 128) = 3.58, p_adj = .061, $ηG2$ = .027. For teens imagining viewing with same-gendered friends, gender differences in number of hostile trailers viewed were somewhat stronger for those imagining viewing with same-gender friends F(1, 72) = 23.82, p_adj < .001, $ηG2$ = .249, compared to teens imagining viewing alone (including those with same-gender friends, F(1, 88) = 11.52, p_adj = .002, $ηG2$ = .116, and those with mixed- or other-gender friends, F(1, 106) = 19.52, p_adj < .001, $ηG2$ = .156). In sum, we found a similar pattern to our mixed ANOVAs, but with less variation of gender effects at each level of interaction.

Discussion

This project examined young teens’ selection and avoidance of hostile and tender emotions in films. Within the framework of Knobloch-Westerwick’s (2015) SESAM, we conceptualized such selections as emotional self-socialization, predicting that media selections would reflect and reinforce gender pressures related to emotional experience and expression, such as male avoidance of tenderness (e.g., Reigeluth et al., 2016). To our knowledge, this is the first study to examine teen boys’ self-socialization away from tenderness and to consider how selections might vary by the imagined presence of friends as co-viewing partners.

We operationalized selective exposure in three ways: participants rated their interest in experiencing specific emotions while watching a hypothetical film, rated their likelihood of viewing actual films pretested to mirror these emotions, then made a binary selection to watch the trailer for either the tender or hostile film. Other than the forced-choice selection of trailers, teens could express equal interest in hostile and tender stimuli. Our findings consistently supported the hypothesized effect of gender (H1, H3, H5): boys, compared to girls and nonbinary youths, reported more interest in hostile emotions and less interest in tender emotions, rated themselves more likely to watch the hostile pre-tested films and less likely to watch the tender pre-tested films, and ultimately chose to watch more hostile trailers.

We also predicted that these gender differences would be larger for teens randomized to imagine selecting content to watch with their friends versus alone. In fact, we did not find viewing condition to moderate gender effects, failing to support our preregistered hypotheses (H2, H4, H6). However, condition effects emerged in post-hoc analyses when gender-mix of friends was added as a moderator. Across all three analyses, gender-mix of friends significantly interacted with gender, condition, and, for our repeated measures designs, film emotion/tone. Gender differences were consistently largest when teens evaluated tender content in the imagined presence of same-gender friends and consistently nonsignificant when teens evaluated hostile content in the imagined presence of mixed- or other-gender friends. These results hint that imagining or anticipating co-viewing with friends may moderate gender differences in affective preferences, but mostly in terms of boys avoiding tender content when with all male friends. Of course, we were not powered for such an interaction, and that a notable number of participants failed the manipulation check raises added concern about the strength of our manipulation. While we are encouraged that effects were consistent across stimuli and did not change when using reported rather than assigned condition, these results require replication.

After viewing all four selected trailers, participants completed post-exposure measures of state hostility and tenderness. Based on the SESAM model, we predicted pre-post changes in these emotions based on the number of hostile trailers viewed. Our path analysis did not reveal post-exposure changes in state hostility, failing to support H7a, but did reveal post-exposure changes in state tenderness, supporting H7b. Given the nature of the forced choice, the number of hostile trailers viewed was the inverse of the number of tender trailers viewed (from 0 to 4), and so it is not clear if tenderness increased after watching tender trailers or decreased after watching hostile trailers, or both, though mean scores overall decreased from pre to post.

Nonetheless, our model found two significant pathways of emotional self-socialization into and/or away from tenderness. Supporting the SESAM proposition that viewers are often motivated to maintain self-consistency, one indirect path found that lower baseline tenderness predicted viewing more hostile (and fewer tender) trailers, which predicted lower levels of post-exposure tenderness. Although gender predicted baseline tenderness, the complete indirect path from gender → pre-tenderness → selections → post-tenderness was not significant. However, gender was the strongest predictor of viewing hostile trailers in our model, resulting in a second, stronger indirect effect on post-exposure tenderness: controlling for baseline emotional attributes, teens tended to make gender-normative media selections (i.e., more hostile trailers for boys, more tender trailers for girls and nonbinary youths), leading to decreased tenderness for boys and increased tenderness for girls and nonbinary youths. These results suggest that, although teens do seek out disposition-congruent content, gender-normative selection or avoidance of tender content may be the more powerful source of teens’ emotional self-socialization.

This supports and extends the scope of SESAM by illustrating how self-concepts (in this case gender) motivate emotional self-socialization via selective exposure and avoidance. Past media research has tended to focus on teen boys’ self-socialization into hostility, aggression, and violence. In our study, tenderness, not hostility, varied most by gender: gender effect sizes for tender content were double those for hostile content in our repeated measures, as shown in Table 2. This could be a measurement issue–hostility had lower reliability than did tenderness and a lower mean at pre- and post-test–but these findings align with research on fear of femininity as driving restrictive emotionality within masculine socialization (e.g., O’Neil, 2008). Our study extends such research on restrictive emotionality into media uses and effects, suggesting that avoidance of tender content may be more central to adolescent boys’ emotional self-socialization than their preference for aggressive, violent, or hostile content. This is an area deserving of more scholarly attention, and future work might also consider additional downstream effects of such avoidance. For example, past studies linking tender emotions to prosociality (de Leeuw et al., 2023) or restrictive emotionality to maladaptive mental health outcomes (Exner-Cortens et al., 2021) further point to the potential harms of boys’ selective avoidance of certain emotional media experiences.

That said, such avoidance was more evident when teens were required to make a forced choice. As reflected in Table 2, when teens were allowed to express equal interest in both hostile and tender content, gender differences explained less variance in our omnibus tests (i.e., $ηG2$⁠) and also appeared somewhat less susceptible to moderation effects in our post hoc analyses (e.g., gender mix of friends moderated gender differences only in mixed ANOVA designs). In fact, when reporting on what type of emotions they would want to experience when watching a movie, boys expressed higher interest in tender over hostile emotions, although the opposite was true when evaluating pre-tested films. In line with this, it is possible that boys are less motivated by avoidance of tender emotions per se than by other characteristics of media tending to contain these emotions. Our exploratory probing found preferences significantly differed between films, albeit more so for hostile than for tender films. Future research may consider the types of tender content that are “permissible” and appealing to boys, such as by comparing plotlines (cf Oliver et al., 2000) and by examining different combinations of emotions.

Relatedly, we were surprised to find sadness cross-load onto pre- and post-test hostility and tenderness and interest in these emotions in films. Greenwood (2010) found sad men preferred dark comedies, which she noted may contain a hostile tone. Meier and Sharp (2024) noted the popularity of the hostile, sad, but decidedly non-tender film, The Joker (2019), among members of a highly misogynistic online community. Examining different combinations of sad, tender, happy, hostile, and other emotions, and measuring affective changes, might offer a clearer picture of the factors teens attend to in making selections, especially when navigating difficult emotions. Additionally, experiments involving full length features, more so than trailers, would be better able to inspect the effects of media exposure involving a multitude of emotions within a single viewing session, while multiple viewing sessions over a longer period would likely be necessary to parse short- from long-term outcomes.

There are other important limitations to the current study that suggest avenues for further research. One is the fact that this was an online experiment, and it was clear to participants that these were hypothetical, rather than actual, choices for solo or co-viewing. Further studies might examine the dynamic process by which friends actually negotiate viewing choices, or how gendered performances during co-viewing (cf Zillmann et al., 1986) might moderate emotional socialization processes. That said, our study was primarily interested in emotional self-socialization as rooted in one’s gendered self-concept. Unlike co-viewing research focused on the influence of a physically co-present other, our primary interest was in the extent to which teens select (or avoid) emotional media content that reinforces their gendered self, and if asking about friends, a known source of gender pressure, increases the salience of the gendered self within media selections. Our path analysis supported gender, above baseline emotional attributes, driving self-socialization, with post hoc findings giving some indication of the moderating influence of friends. While there’s much to learn about embodied co-viewing experiences, further studies focused on hypothetical or anticipated co-viewing, or even using a more traditional priming manipulation, might be equally helpful in probing how friends and peers might activate self-concepts in ways predictive of self-socialization.

We conceive of the current results as a first step in understanding emotional self-socialization as an identity-driven process. Our study focused on teen selections of films, but there are numerous additional ways in which media uses may reflect and contribute to gender expectations and performances. To return again to 1995, Steele and Brown (1995) described the presence of gendered media content in adolescent room culture: “The stern looks on the athletes’ faces reflected the tough, determined attitude he liked to imitate when playing baseball or basketball” (p. 568). Arnett (1995) argued that teens use media to self-socialize and to create a shared youth culture; media presentations of emotions valued as “masculine” or “feminine” might inform the values that youths share with each other, perpetuating broader cycles of cultural socialization. Aggression, assertiveness, and dominance don’t have to be masculine emotional attributes, nor should affection, tenderness, and warmth be read as feminine. Thus, a larger goal of this research program is to understand the gendering of these emotions through mediated experiences and, ultimately, to examine ways in which teens might come to recognize these emotions not as feminine or masculine, but as fundamental to being human.

Supplementary material

Supplementary material is available at Human Communication Research online.

Data availability

The data and syntax underlying this article are available at https://osf.io/7aynj/.

Funding

The project did not receive any external funding.

Conflicts of interest

None declared.

Acknowledgements

The authors wish to thank Zhongdang Pan, Joanne Cantor, and Janet Hyde for providing feedback on an earlier version of this paper, Y. Anthony Chen for his assistance with the path analysis, as well as the editors, Steven Wilson and Riva Tukachinsky Forster, and our three anonymous reviewers for their thoughtful questions, suggestions, and encouragement.

References