2,259
Views
6
CrossRef citations to date
0
Altmetric
Original Articles

Pathways from College Students’ Cognitive Scripts for Consensual Sex to Sexual Victimization: A Three-Wave Longitudinal Study

ORCID Icon &

ABSTRACT

Sexual scripts serve as cognitive representations of typical elements of sexual interactions that guide sexual behavior. To the extent that cognitive scripts for consensual sex comprise elements associated with a risk of experiencing nonconsensual sex, they may be indirectly linked to sexual victimization via risky sexual behavior. A longitudinal study with 2,425 college students in Germany (58% female) examined pathways from sexual scripts for consensual sex, sexual behavior, and sexual victimization over three data waves separated by 12-month intervals. Sexual scripts and behavior were defined as risky to the extent that they include known vulnerability factors for sexual victimization (casual sex, alcohol consumption, ambiguous communication of sexual intentions). Path analyses confirmed that more risky sexual scripts prospectively predicted more risky sexual behavior, which predicted higher odds of sexual victimization. The findings held for men and women and participants with exclusively opposite-sex and both same- and opposite-sex contacts. Moreover, reciprocal influences between risky scripts and risky sexual behavior were found over time, confirming the proposed mutual reinforcement of scripts and behavior. The findings have implications for conceptualizing the role of scripts for consensual sex as vulnerability factors for sexual victimization among women and men and may inform intervention efforts.

Scripts are cognitive representations of typical sequences of events that organize thinking, feeling, and behavior in a wide range of contexts. They contain an individual’s generalized knowledge about the characteristic elements of situations, which include typical sequences of events, normative beliefs, and emotions (Schank & Abelson, Citation1977). One area of social psychological theorizing and research to which the script concept has been applied widely is sexuality. Simon and Gagnon (Citation1986) introduced the construct of sexual scripts that reflect cultural constructions of sexuality and function to shape sexual interactions at the interpersonal level as well as regulate an individual’s sexual behavior. A large body of evidence has been generated about the role of sexual scripts for understanding various aspects of sexual interactions, such as the typical elements of the hookup script (Olmstead et al., Citation2019), the construction of consent (Newstrom et al., Citation2021), and the conceptualization of sexual assault (Yeater et al., Citation2020). Sexual scripts can be conceptualized at the cultural level as socially shared constructions of sexual interactions in a society and at the individual level as persons’ endorsement of these cultural constructions. Individuals hold personal scripts as mental representations about their own sexual behavior that are generalized across situations. Past research has shown that these individual sexual scripts are more closely linked to sexual behavior than a person’s knowledge of the general sexual script representing knowledge about the cultural construction of consensual scripts that generalized across both situations and individuals (Krahé et al., Citation2007a).

In the current study, we examined the role of individuals’ cognitive scripts for consensual sexual interactions in explaining women’s and men’s vulnerability to sexual victimization. The core assumption was that cognitive scripts for consensual sexual interactions that contain features associated with an increased probability of sexual victimization will prospectively predict the odds of victimization via the enactment of the script in sexual behavior. This proposition was examined in a large sample of undergraduate students in Germany based on a three-wave longitudinal design. Sexual victimization, defined as sexual experiences without consent, is widespread among college students, as documented in many large-scale studies from a broad range of countries (e.g., Cantor et al., Citation2020; Dworkin et al., Citation2021; Fedina et al., Citation2018). Therefore, understanding vulnerability factors that are related to higher odds of experiencing sexual victimization is significant not only from a theoretical perspective, but also as a basis for an applied approach to improve student sexual health and well-being. Moreover, longitudinal studies demonstrating the role of sexual scripts as predictors of sexual behavior and adverse outcomes are rare in basic social psychological research using the script concept. Therefore, a further aim of the current study was to contribute to the validation of the script concept by examining long-term effects of scripts in the field of sexuality and demonstrating the reciprocal influence of scripts and behavior over time.

Specific Sexual Behaviors as Vulnerability Factors of Sexual Victimization

Establishing a knowledge base of behavioral vulnerability factors is seen as critical for empowering women’s risk management to avoid sexual violence. As suggested by routine activity theory, individuals’ patterns of everyday behavior may create opportunities for other persons’ criminal behavior. For example, the higher rate of violent crime in hotter compared to cooler months of the year has been explained by a higher likelihood of encountering violent others when people spend more time outdoors (Cohen & Felson, Citation1979). In the same vein, individuals’ regular patterns of consensual sexual behavior may create opportunities that perpetrators can exploit to commit acts of sexual aggression against them (e.g., Culatta et al., Citation2020; McGraw et al., Citation2020).

In the context of the present study, we use the term “risky sexual behavior” with a specific meaning to denote behaviors in the context of sexual interactions that are linked to an increased probability of experiencing sexual victimization. Past research has consistently identified the following aspects of sexual behavior to be related to increased odds of experiencing sexual victimization: (1) alcohol consumption in the context of sexual encounters (Abbey et al., Citation2004; Krahé et al., Citation2015; Lorenz & Ullman, Citation2016); (2) casual sex with unfamiliar partners, as captured in the construct of “hooking up” (Garcia et al., Citation2019; Testa et al., Citation2020), and (3) the ambiguous communication of sexual intentions, leaving room for misinterpreting a person’s consent cues and allowing potential perpetrators to deny that victims indicated non-consent to their sexual advances (Muehlenhard et al., Citation2016; Newstrom et al., Citation2021). These factors not only distinguish between victims and non-victims of sexual aggression, but also between individuals who were victimized repeatedly and those victimized only once (Walsh et al., Citation2020). Moreover, they often come together, for example, such that individuals engaging in casual sexual interactions are also more likely to drink alcohol in the context of sexual interactions (Claxton et al., Citation2015).

Sexual Scripts as Guidelines for Sexual Behavior

Sexual behavior in general and sexual behavior that is risky with respect to sexual victimization in particular is shaped by cognitive representations of sexual interactions in the form of sexual scripts (Simon & Gagnon, Citation1986). Sexual scripts denote cognitive representations of the typical elements and events of sexual interactions and serve as guidelines for individuals’ sexual behavior. Cognitive scripts for consensual sex may indicate a higher victimization risk if they contain behaviors linked to an increased probability of sexual victimization, such as drinking in the context of sex, unclear communication of sexual intentions, and engaging in sex with casual partners. Thus, risky sexual scripts for consensual sex may inform risky sexual behavior, which – in accordance with the propositions of routine activity theory – may in turn predict a higher vulnerability to sexual victimization. These associations have been demonstrated in several two-wave longitudinal studies with adolescents and young adults (D’Abreu & Krahé, Citation2016; Krahé et al., Citation2007b; Schuster & Krahé, Citation2019; Tomaszewska & Krahé, Citation2018a). In another two-wave study, adherence to the traditional heterosexual script, defined by the construction of men as initiators and women as gatekeepers of sexual interactions, was linked to a lower ability to resist unwanted hookups (Gamble, Citation2019). In the current study, a three-wave design was employed to yield more conclusive findings regarding the temporal paths from cognitive scripts for consensual sex via sexual behavior to sexual victimization.

The Role of Gender and Sexual Minority Status

A broad research literature has shown consistently that male college students experience sexual victimization, but at a lower rate compared to female students (e.g., Cantor et al., Citation2020; D’Abreu & Krahé, Citation2016; Hines et al., Citation2012; Mumford et al., Citation2020; Schuster & Krahé, Citation2019; Tomaszewska & Krahé, Citation2018b). Despite this difference in prevalence, the risk factors of sexual victimization were found to be similar for women and men. For example, hooking up and heavy drinking were linked to higher odds of victimization in female and male college students (Mellins et al., Citation2017), and the reciprocal associations between sexual self-esteem, depression, and sexual victimization over time were also found to hold for both women and men (Krahé & Berger, Citation2017). This pattern fits the conceptual framework of routine activity theory, which suggests that differences in vulnerability depend on differences in behavior rather than fixed person variables, such as gender (Hines et al., Citation2012).

Another consistent body of research has shown that members of sexual minorities (LGBTQI) have a higher probability of experiencing sexual victimization than heterosexual individuals (e.g., Canan et al., Citation2019; Chen et al., Citation2020; Peterson et al., Citation2011). Again, despite differences in the prevalence rates of sexual victimization, associations with vulnerability factors, such as heavy drinking or childhood sexual abuse, were found to be similar (López & Yeater, Citation2021; Ray et al., Citation2021). In the present study, differences in patterns of sexual behavior, defined by having had sex with a male partner, a female partner, or both, were examined in relation to the pathway from risky sexual scripts to sexual victimization. This approach reflects the key role assigned to sexual behavior in the present study rather than sexual orientation as a stable person characteristic. For example, a substantial number of women engage in sexual contact with men and with other women without identifying as bisexual (see the construct of “mostly straight girls”; Thompson & Morgan, Citation2008).

Overall, the existing literature shows that women experience sexual victimization at a higher rate compared to men, and members of sexual minorities are more vulnerable than heterosexual individuals. A more limited knowledge base suggests that despite these differences in prevalence, the factors predicting an increased vulnerability to sexual victimization may be similar across gender and sexual orientation groups.

The Current Study

Many recent studies on vulnerability factors of sexual victimization are based on cross-sectional designs (e.g., McGraw et al., Citation2020; Ray et al., Citation2021; Tyler et al., Citation2017). These designs are limited in that they cannot go beyond correlational associations, ruling out conclusions about the impact of one variable as a vulnerability factor for the other. To yield more conclusive evidence, longitudinal designs are needed in which the proposed vulnerability factors are assessed prior to the outcomes. Mediational models require at least three data waves in which the predictor is assessed at Time 1, the mediator at Time 2, and the outcome at Time 3, controlling for the stability of the predictor, mediator, and outcome variables over time. The current study adopted such a three-wave longitudinal design. The large sample size with low dropout rates, the use of a validated instrument for collecting reports of sexual victimization with behaviorally specific items, and the inclusion of both female and male college students were further distinctive features of the study. Beyond the field of sexual aggression, the design of the study contributes to the conceptual validation of the script construct by providing longitudinal data on the reciprocal reinforcement of scripts and behavior.

Based on the theoretical arguments and empirical findings outlined above, the study tested the following predictions:

Hypothesis 1: More risky sexual scripts at Time 1 (T1), defined by the extent to which established vulnerability factors for sexual victimization (casual sex, alcohol use, ambiguous communication) are part of a person’s script for consensual sex, predict more risky sexual behavior, defined by the same factors, at Time 2 (T2).

Hypothesis 2: More risky sexual behavior at T2 predicts a higher probability of sexual victimization at T3.

Hypothesis 3: The path from more risky sexual scripts at T1 to a higher probability of sexual victimization at T3 is mediated by more risky sexual behavior at T2.

Hypothesis 4: Risky sexual scripts and risky sexual behavior reinforce each other over time, as reflected in significant indirect paths from risky sexual scripts at T1 via risky sexual behavior at T2 to risky sexual scripts at T3 (Hyp. 4.1), and from risky sexual behavior at T1 via risky sexual scripts at T2 to risky sexual behavior at T3 (Hyp. 4.2).

All proposed associations were tested controlling for the temporal stability of the respective constructs from T1 to T3 and were expected to apply equally to female and male participants as well as participants with different sexual experience backgrounds. Because questions were only asked about male and female partners and only participants who identified as male and female were included, we refer to opposite-sex and same-sex partners to describe these differences in sexual experience background.

Method

Sample

The sample consisted of 2,425 college students (1,415 women, 1,010 men) from different state-funded, tuition-free universities in the Federal States of Berlin and Brandenburg, Germany, who were in their first year at T1 (2011). The mean age of the sample at T1 was 21.23 years (SD = 2.31; range: 18–30 years). Participants were enrolled in a wide range of academic degree courses. The second and third data waves (T2, 2012, and T3, 2013) were conducted 12 and 24 months after T1. The sample size was 1,685 (1,033 women and 652 men) at T2 and 1,619 (1,000 women and 619 men) at T3. This corresponds to dropout rates of 30.5% from T1 to T2 and 3.98% from T2 to T3. Participants who dropped out after T1 had more sexual partners, were younger at first sexual intercourse, and scored higher on the risky scripts and risky behavior measures than did those who remained in the sample at T2. As explained below, all T1 participants were included in the analyses, and missing data were handled using Full Information Maximum Likelihood estimation.

At T1, most participants (91.7% of men and 95.1% of women) reported having had consensual sexual contacts. Of these, 85.1% of men and 77.7% of women reported exclusively opposite-sex contacts, 10.3% of men and 21.0% of women reported sexual contacts with both opposite- and same-sex partners, and 4.6% of men and 1.3% of women reported exclusively same-sex contacts. By T3, all participants had consensual sexual experiences, with 84.0% of men and 73.7% of women reporting exclusively heterosexual contacts, 12.1% of men and 25.6% of women reporting both heterosexual and same-sex contacts, and 3.9% of men and 0.7% of women reporting exclusively same-sex contacts.

At T1, 85.9% of men and 90.9% of women reported they were currently in a steady relationship (i.e., a relationship in which both partners acknowledge a commitment to each other) or had been in a steady relationship in the past. At T3, the corresponding rates were 93.7% for men and 95.5% for women. The mean age at first sexual intercourse, as reported at T1, was 16.99 years for men (SD = 1.94), and 16.47 years (SD = 1.83) for women. Men had a mean number of 5.42 coital partners (SD = 8.02), the mean for women was 4.27 (SD = 4.00). The gender difference was significant on both variables, multivariate F (df = 2,2020) = 37.78, p < .001.

Instruments

Sexual Experience Background and Demographics

At the beginning of the questionnaire, participants were asked to indicate their sex, age, nationality, home university, and subject of study, whether they were currently in a steady relationship and whether they had been in a steady relationship in the past. In terms of sexual experience background, they were asked whether or not they had ever engaged in sexual contact with a member of the same sex and a member of the opposite sex (categories: male/female; response options: no, yes without sexual intercourse; yes with sexual intercourse). Those who reported coital experience were asked to indicate their age at first intercourse and number of coital partners.

Sexual Victimization

Reports of sexual victimization were obtained with the Sexual Aggression and Victimization Scale (SAV-S) developed in Germany by Krahé and Berger (Citation2013). Based on the use of behaviorally-specific items pioneered in the Sexual Experiences Survey by Koss et al. (Citation1987) and Koss et al. (Citation2007), the SAV-S combines (a) three coercive strategies (threat or use of physical force; exploitation of the inability of the victim to resist, e.g., due to alcohol or drug consumption; use of verbal pressure, e.g., calling the victim a failure) with (b) three victim-perpetrator relationships (current or former partner; acquaintance; stranger) and (c) four sexual activities (sexual touch; attempted sexual intercourse; completed sexual intercourse; other sexual acts, e.g., oral sex). This combination results in a total of 36 items. At T1, a four-point scale of 0 (never), 1 (once), 2 (twice) and 3 (three or more times) was used for each item. Because the number of responses in the categories > 2 was very low, the format was changed to a dichotomous response scale of 1 (once) and 2 (more than once) at T2 and T3, with a summary response option (I did not experience any of these actions) replacing the 0 category for each item to reduce the time needed to complete the survey. Based on filter questions about past sexual experiences (opposite-sex partners only, same-sex partners only; both), participants were assigned to tailored versions representing different gender constellations between victims and perpetrators. For example, women who reported exclusively opposite-sex contacts received the questions about a male perpetrator, as did men who reported exclusively same-sex contacts. At T1, participants were asked to complete the items for the time period since their 14th birthday, the legal age of consent in Germany. At T2 and T3, they were asked to complete the items for the preceding 12 months. The reliability and validity of the SAV-S have been demonstrated in previous research (Krahé et al., Citation2016; Schuster et al., Citation2021). A copy of the measure can be obtained from the first author.

In addition to yielding overall prevalence rates based on the number of participants who endorsed at least one victimization item, responses to the SAV-S items can be translated into an ordinal score of sexual victimization by classifying each participant according to her or his most severe form of sexual victimization. Following the procedure adopted by Koss et al. (Citation2008), participants who did not endorse any of the victimization items were assigned to the nonvictim (1) category. Participants who reported at least one experience of unwanted sexual contact without penetration of the body through the use of verbal pressure, exploitation of victim’s intoxicated state, threat or use of physical force, but no attempted sexual coercion, sexual coercion, attempted rape, and rape, were classified as victims of unwanted sexual contact (2). Participants who reported at least one experience of attempted (but not completed) penetration using verbal pressure, but no attempted and completed rape were classified as victims of attempted sexual coercion (3). Those who endorsed at least one item of completed penetration using verbal pressure, but no attempted or completed rape were categorized as victims of sexual coercion (4). Those who reported attempted, but not completed, penetration through exploitation of their inability to resist or threat or use of physical force were classified as victims of attempted rape (5), and those who endorsed at least one item of completed penetration through exploitation of their inability to resist or threat or use of physical force were categorized as victims of completed rape (6). Because of low frequencies for attempted and completed sexual coercion, these two categories were combined to yield a final five-level score (see Johnson et al., Citation2017; Yeater et al., Citation2020, for a similar approach).

Risky Sexual Scripts

This construct was measured by a scenario-based approach developed by Krahé et al. (Citation2007a). Participants were asked to imagine a typical situation in which they had sexual intercourse with a new partner for the first time and to rate how likely the following features would be part of that situation: (1) consumption of alcohol and degree of intoxication (six items, e.g., ”How likely is it that alcohol is consumed by you; by the man/woman?”), (2) ambiguous communication of sexual intentions (four items, e.g., “How likely is it that you first say ‘no’ even though you also want to have sex with her/him?”), and (3) length of acquaintanceship and engaging in casual sex (three items, e.g., “How long have the two of you known each other?,” reverse coded). A five-point response scale was presented with options matched to the content of the items: 1 (very unlikely) to 5 (very likely) for alcohol consumption, ambiguous communication, and casual sex, 1 (not at all) to 5 (totally) for the level of intoxication, and 1 (not at all) to 5 (a few months or longer) for the length of the relationship. Responses were aggregated across the 13 items to yield a total script score. Internal consistencies were α = .69 at T1, α =.74 at T2, and α = .71 at T3.

Risky Sexual Behavior

Behavior reflecting risky sexual scripts was measured by seven items based on previous research by Krahé et al. (Citation2007b). Four items referred to alcohol use in the context of sexual interactions (e.g., “How often did you/did the other person drink alcohol in situations in which you had sexual intercourse”; “How drunk were you/was the other person in these situations?”), two items referred to ambiguous communication of sexual intentions (saying ‘no’ when meaning ‘yes’ and saying “yes” when meaning “no”), and one item referred to having sexual intercourse with a partner they did not know well (“When you had sex: how often was it with someone you knew hardly or not at all”). Five-point response scales were used, ranging from 1 (never/not at all drunk) to 5 (almost every time/totally drunk). At T1, participants were asked to think about situations in which they had sex “in the past.” At T2 and T3, they were instructed to think about situations in which they had sex “in the past 12 months.” Internal consistencies were α = .74 at T1, α = .72 at T2, and α = .65 at T3.

Procedure

Approval for the study and all materials was obtained from the Ethics Committee of the authors’ university. The study was conducted as an online survey. Invitations to participate were sent out via e-mail to first-year students of the participating higher education institutions through the respective student offices or student associations. Interested students registered in a data bank created for the purposes of this study and were sent the link to the online questionnaire upon registration. Participants were required to give active consent before being able to proceed to the items. At each data wave, all participants received a 10-Euro shopping voucher for their participation. On each page of the SAV-S, participants could press a „help button“, in case they felt the need for professional support. Help button was pressed only once, and this turned out to be accidental.

Plan of Analysis

The hypotheses were tested using a cross-lagged panel model (CLPM) approach. This approach examines the hypothesis that individual differences on the predictor variables can predict individual differences on the outcome variables, controlling for the stability of constructs over time, and was found to yield robust findings when applied to multiple data sets examining the same associations (Orth et al., Citation2021). Before testing the proposed longitudinal associations, we examined whether measurement invariance across the three data waves was given for the sexual script and sexual behavior measures, as described below. No measurement invariance was assumed for the frequency reports of sexual victimization.

To test our hypotheses, we first estimated a multi-group model in which all paths were constrained to be equal for men and women. Next, we compared this model to a model in which all paths were allowed to vary between men and women. Based on the finding that the unconstrained model did not fit the data significantly better than the constrained model, a single-group model was estimated and adopted as the final model. The single-group model included gender as a covariate to account for gender differences in mean scores, as described below. To examine potential differences in relation to participants’ sexual experience background, we used the same two-step approach, comparing participants with exclusively opposite-sex contacts and participants with both same-sex and opposite-sex contacts in an unconstrained vs. constrained multi-group model. The group with exclusively same-sex contacts was too small to be considered separately.

Missing data as well as non-normality of the distributions were handled by using a robust Full Information Maximum Likelihood (FIML) estimator (Muthén and Muthén (Citation1998–2012)). Indirect paths were tested through examining 95% and 99% confidence intervals based on 10,000 bootstraps. Since bootstrapping is not available in combination with the MLR estimator, the ML estimator was used for these analyses.

Access to the data on which the analyses are based can be obtained from the first author on request.

Results

Measurement Invariance of Risky Sexual Scripts and Risky Sexual Behavior

Measurement invariance for the script and behavior items conceptualized of indicators as latent constructs was tested following the standard procedure as outlined, for example, by Van de Schoot et al. (Citation2012). First, a measurement model was estimated for each construct in which the factor loadings were freely estimated across the three data waves. To account for item-level correlations over time, item-specific measurement factors were added to the models based on the inspection of the modification indices (9 items for risky sexual scripts; 3 items for risky sexual behavior). Moreover, inter-item correlations within data waves were allowed based on the modification indices (7 items for sexual scripts; 3 items for sexual behavior). Second, this unconstrained model was compared to a model in which both the factor loadings and intercepts were constrained to be equal across time, representing the assumption of strong measurement invariance. Following Cheung and Rensvold (Citation2002), a reduction in CFI of ≤ .01 was used as an indicator of invariance.

For sexual scripts, the unconstrained model showed an acceptable fit with the data, Chi2 (df = 588) = 2,542.53, p < .001; RMSEA = 0.037 (C.I. .036; .039); CFI = 0.929; SRMR = 0.055. The model in which both the factor loadings and the intercepts were constrained to be equal across the three waves also showed an acceptable fit, Chi2 (df = 638) = 2,734.58, p < .001; RMSEA = 0.037 (C.I. .035; .038); CFI = 0.923; SRMR = 0.056. The value of ΔCFI of −.006 supports the assumption of strong measurement invariance.

For the 7-item measure of risky sexual behavior, the model fit of the unconstrained model was acceptable, Chi2 (df = 157) = 1,023.18, p < .001; RMSEA = 0.049 (C.I. .046; .052); CFI = 0.941; SRMR = 0.047, but not significantly better than the model in which the factor loadings were constrained to be equal, Chi2 (df = 169) = 1,110,29, p < .001; RMSEA = 0.049 (C.I. .046; .052); CFI = 0.936; SRMR = 0.051, ΔCFI = −.005. However, the model in which both the factor loadings and the intercepts were constrained to be equal had a worse fit than the unconstrained model, Chi2 (df = 183) = 1,332.99, p < .001; RMSEA = 0.052 (C.I. .049; .055); CFI = 0.922; SRMR = 0.054, ΔCFI = −.019. Therefore, the assumption of weak, but not strong measurement invariance could be supported for the measure of risky sexual behavior.

Descriptive Statistics and Correlations

Across all 36 items, 28.8% of female participants at T1 endorsed at least one victimization item for the period since their 14th birthday, 26.0% of female participants at T2 endorsed at least one item for the period of the last 12 months, and 25.4% of female participants at T3 reported at least one experience of sexual victimization during the past 12 months. For male participants, the corresponding rates were 14.2% at T1, 24.0% at T2, and 22.6% at T3. The gender difference was significant at T1, χ2 (df = 1) = 71.24, p < .001, but not at T2 and T3. The percentage of new victims at T2 (who reported no victimization at T1) was lower for women (54.4%) than for men (73.7%), χ2 (df = 1) = 15.16, p < .001. Similarly, the percentage of victims at T3 who did not report victimization at T1 was higher for men (76.4%) than for women (50.0%), χ2 (df = 1) = 25.95, p < .001. This means that more women than men experienced their first sexual victimization before starting university.

The comparison by sexual experience background yielded significantly higher prevalence rates for women who had sex with both men and women than for women who only had sex with men at each data wave (T1: 40.6% vs. 28.3%, χ2 (df = 1) = 12.79, p < .001; T2: 34.8% vs. 23.8%, χ2 (df = 1) = 10.67, p < .01; T3: 35.4% vs. 22.1%, χ2 (df = 1) = 17.06, p < .001). For men, the corresponding rates for men with both female and male partners and men with only female partners were significantly different only at T2 (T1: 11.0% vs. 12.5%, n.s.; T2: 35.8% vs. 21.9%, χ2 (df = 1) = 6.27, p < .05; T3: 24.4% vs. 21.8%, n.s.).

The internal consistencies for all constructs as well as means and standard deviations for female and male participants at each of the three data waves are presented in . On the five-level severity score of sexual victimization, the mean for women was significantly higher than the mean for men at the first data wave, but not at the second and third data waves, reflecting the higher rate of first victimization prior to university among women reported above. Men had more risky sexual scripts than women at each data wave. No gender differences were found with regard to risky sexual behavior at the three data waves. Participants with both same-sex and opposite-sex contacts scored higher than did those with exclusively opposite-sex contacts on all constructs at each data wave (see Table S1 in the Supplementary Materials).

Table 1. Means and gender differences of all study variables

The bivariate correlations between the model variables for the total sample are presented in , based on the full sample and using FIML estimation to handle missing data at T2 and T3. All correlations were significant. The stability coefficients over the three waves reflect the conceptualization of sexual scripts as relatively stable cognitive representations, whereas sexual behavior and sexual victimization are not conceptually assumed to show high stability over time. The correlation matrix comparing participants with exclusively opposite-sex partners and those with both same-sex and opposite-sex partners is shown in Table S2 of the Supplementary Materials.

Table 2. Bivariate correlations of all study variables

Hypothesis Testing

To test the proposed longitudinal associations between sexual scripts, sexual behavior, and sexual victimization, a cross-lagged panel model was specified for the three data waves. We began by estimating an unconstrained multi-group model, in which all paths were allowed to vary between men and women. This model fitted the data well, Chi2 (df =12) = 21.03, p = .050, CFI = .997, RMSEA = .025 (C.I. .000; .042), SRMR = .013. In the next step, a multi-group model was estimated in which all paths were constrained to be equal for men and women. This model also fitted the data well, Chi2 (df = 39) = 44.59, p = .248, CFI = .998, RMSEA = .011 (C.I. .000; .024), SRMR = .031. The Chi2 difference test indicated that the constrained model did not fit significantly worse than the unconstrained model, Chi2 diff (df = 27) = 23.56, p = .665, suggesting that the paths in the model did not vary significantly by gender. To account for gender differences in the means, we then estimated a single-group model in which gender was included as a covariate for all model variables. Based on the good fit, Chi2 (df = 6) = 13.16, p = .040, CFI = .998, RMSEA = .022 (C.I. .004; .039), SRMR = .009, this model was adopted as the final model, shown in .

Figure 1. Longitudinal pathways from risky sexual scripts and risky sexual behavior to sexual victimization

Note. Single-group model with gender as a covariate. Standardized coefficients for the total sample. Cross-sectional correlations in the model at each data wave are reported in the text. Dashed lines denote nonsignificant associations. *p < .05; ***p < .001.
Figure 1. Longitudinal pathways from risky sexual scripts and risky sexual behavior to sexual victimization

Cross-sectional correlations in the model are different from the bivariate correlations in because they were estimated in combination with all other model variables. Omitted from for the sake of clarity, the cross-sectional correlations between risky scripts and risky behavior were .47 at T1, .30 at T2, and .25 at T3. The correlations between scripts and sexual victimization, were .14 at T1, .04 at T2, and −.01 at T3, and the correlations between risky sexual behavior and sexual victimization were .27 at T1, .18 at T2, and .12 at T3. With the exception of the two values of .04 and −.01, all correlations were significant at p < .001.

Consistent with Hypothesis 1, risky sexual scripts at T1 significantly predicted risky sexual behavior at T2. As proposed in Hypothesis 2, a significant path was also found from risky sexual behavior at T2 to sexual victimization at T3. Bootstrapped confidence intervals were inspected to evaluate the indirect paths. A significant indirect path was found from risky sexual scripts at T1 via risky sexual behavior at T2 to sexual victimization at T3, consistent with Hypothesis 3, ß = .020, 99% C.I. [.004; .035]. Moreover, the indirect path from risky sexual scripts at T1 to risky sexual scripts at T3 via risky sexual behavior at T2, proposed in Hypothesis 4.1, was significant, ß = .028, 99% C.I. [.013; .043]. The indirect path from risky sexual behavior at T1 to risky sexual behavior at T3 via risky sexual scripts at T2, proposed in Hypothesis 4.2, was also significant, ß = .024, 95% C. I. [.010; .038]. These pathways support the assumption of a mutual reinforcement between scripts and behavior. All associations were controlled for the stability of the constructs over time. A full table of all significant indirect paths is provided in Table S3 in the Supplementary Materials.

To compare participants with exclusively opposite-sex and both opposite- and same-sex contacts, we first specified a multi-group model in which the paths were allowed to vary. Gender was included as a covariate. This model showed a good fit, Chi2 (df = 12) = 25.46, p = .013, CFI = .997, RMSEA = .032 (C.I. .014; .049), SRMR = .011. Next, a multi-group model was estimated in which all paths were constrained to be equal for the two groups. This model also fitted the data well, Chi2 (df = 48) = 72.48, p = .013, CFI = .994, RMSEA = .022 (C.I. .010; .031), SRMR = .036. It did not fit significantly worse than the unconstrained model, Chi2 diff (df = 36) = 47.02, p = .103, suggesting that the paths in the model did not vary significantly by sexual experience background. This final model is presented in the Supplementary Materials, Figure S1. The pattern of findings was almost identical to the model for the total sample. The only exception was that the path from risky sexual behavior at T1 to sexual victimization at T2 was nonsignificant in this analysis. Thus, the proposed associations between sexual scripts, sexual behavior, and sexual victimization were supported for participants differing in sexual experience background.

Discussion

The present study applied the construct of cognitive scripts to the analysis of individual differences in the vulnerability to sexual victimization, contributing both to the understanding of sexual victimization and to the validation of the script construct in a longitudinal design. The core proposition underlying the study was that cognitive scripts for consensual sexual interactions may hold a clue to understanding vulnerabilities to sexual victimization based on sexual scripts. Scripts for consensual sex may be more or less risky with regard to experiencing nonconsensual sex depending on whether they contain behavioral features associated with a higher probability of sexual victimization. Three such features have been consistently identified in past research and were addressed in our study: (1) engaging in casual sex with partners one does not know well, (2) alcohol consumption in sexual interactions by self and partner, and (3) ambiguous communication of sexual intentions (e.g., saying “no,” but meaning ’yes’; saying “yes,” but meaning “no”). Based on the theoretical conceptualization of cognitive scripts as generalized mental representations of situations that serve to inform behavior (Abelson, Citation1981), we proposed that the more the risk elements were part of participants’ cognitive representations of a consensual sexual encounter, the more likely they would be to engage in the scripted behavior in actual sexual interactions. Based on routine activity theory (Cohen & Felson, Citation1979), we further proposed that these routine patterns of sexual behavior would create opportunities for potential perpetrators to engage in sexual aggression, resulting in a higher likelihood of experiencing sexual victimization.

Our findings fully support this line of reasoning. As hypothesized, sexual scripts for consensual sex that contained features linked to increased odds of sexual victimization predicted more risky sexual behavior, assessed 12 months later and defined by the same features. This finding underlines the proposed significance of sexual scripts as guides of behavior. Consistent with the perspective of routine activity theory, risky sexual behavior predicted sexual victimization at the following data wave. Moreover, a significant indirect path was found from risky sexual scripts at T1 via risky sexual behavior at T2 to sexual victimization at T3. The associations were found controlling for the temporal stability of risky scripts and risky sexual behavior and repeated experience of sexual victimization over the three data waves, using a cross-lagged panel design. These results based on college students in Germany join findings from student samples in other countries such as Brazil, Chile, Poland, Turkey, and the U.S. (D’Abreu & Krahé, Citation2016; Schuster & Krahé, Citation2019; Tomaszewska & Krahé, Citation2018a; Turchik et al., Citation2009) as well as adolescents in Germany (Krahé et al., Citation2007b) based on two-wave designs. The present three-wave design facilitated a more stringent test of the proposed causal chain from scripts to victimization by measuring the predictor, the mediator, and the outcome at all three data waves.

The findings fully support the hypothesized path model for both women and men and for individuals with exclusively opposite-sex and with both opposite-sex and same-sex contacts. They are consistent with earlier studies that also found that despite differences in the overall prevalence of victimization as defined by gender and sexual orientation, the same mechanisms may explain an increased vulnerability to sexual victimization (Krahé & Berger, Citation2017; Ray et al., Citation2021).

Gender differences were found in scripts for consensual sex, with women holding less risky sexual scripts. This finding ties in with previous studies showing gender differences in the contents of sexual scripts (Jozkowski & Peterson, Citation2013; Olmstead et al., Citation2019), including scripts about situations that pose a risk of sexual assault (Yeater et al., Citation2020). However, the strength of the association between sexual scripts and sexual behavior did not vary between men and women in the current study, underlining the functional significance of scripts as guidelines for behavior irrespective of differences in the content of the script elements.

The overall prevalence rates of sexual victimization of around 25% for women at each of the three data waves were similar to the rates commonly found in studies based on North American college samples (Muehlenhard et al., Citation2017). The rate for men was lower for the period from 14 years to the T1 assessment, but similar to women’s for the two waves while they were in college. The latter finding is consistent with a recent review of male sexual victimization that identified several studies with equal or higher rates for men (Depraetere et al., Citation2020). Moreover, the finding that the one-year rates in the first two college years were of similar magnitude as the rate for the period between the age of 14 years and the beginning of college is consistent with a broad literature identifying the first college years as a “red zone” with regard to the risk of sexual victimization (Cranney, Citation2015). This conclusion holds particularly for men, who had a higher rate of first-time victimization at T2 and T3 compared to women. A tentative explanation could be that men were significantly older than women at their first sexual intercourse, yet had a significantly higher number of partners, which could suggest that their sexual activity started closer to their arrival at university, but then involved a higher number of opportunities for victimization.

With respect to the conceptual validation of the script construct, the present findings contribute longitudinal data for evaluating two critical predictions. The first is that cognitive scripts function to organize inferences about the typical and expected behavior in specific situations. The second is that cognitive scripts and social behavior inform each other in a reciprocal fashion. Scripts are seen as learned generalizations extracted from repeated experiences of a situation, which proposes a path from behavior to scripts, but scripts also influence behavior by providing a knowledge base on which behavioral decision are made. This reciprocal process is conceptualized, for example, as a core mechanism in developmental theories of aggressive behavior (Huesmann, Citation1988, Citation2018). The current study supported the reciprocal influence between scripts and behavior by showing significant paths from scripts via behavior to scripts and from behavior via scripts to behavior over a period of 24 months, controlling for the temporal stability of both constructs.

Strengths, Limitations, and Conclusion

Our study had both strengths and limitations. Strengths include the large sample size, the three-wave design covering a period of 24 months, and the inclusion of both female and male participants with different sexual experience backgrounds. Moreover, the study can provide novel insights not only to the field of sexual violence research but also to the conceptualization of cognitive scripts. Limitations are the reliance on a convenience sample of college students and the small number of participants with exclusively same-sex contacts, which precluded a comparison with participants with exclusively opposite-sex contacts. Moreover, participants with a nonbinary gender identity were not included in the study. These aspects mean that the extent to which the findings can be generalized to college students in Germany covering a broad spectrum of gender diversity remains a question for future research. Nevertheless, the findings suggest that sexual scripts as generalized cognitive representations of consensual sexual interactions may provide critical information for understanding sexual behavior that is associated with a vulnerability to sexual victimization. Because both sexual scripts and sexual behavior are amenable to change (Schuster et al., Citation2020), the findings provide a starting point for developing interventions addressing risky scripts for consensual sex to reduce the vulnerability to sexual victimization. Such interventions should aim at reducing the prominence of risk elements of sexual victimization in individuals’ scripts for consensual sex, for example, the perception that alcohol use is both a typical and a normative part of sexual encounters. Because scripts provide a basis for sexual behavior across multiple situations, altering the content of sexual scripts to make them less risky with regard to the experience of sexual victimization can be an effective part of promoting risk management at the level of sexuality-related cognitions.

Ethics Approval

The study was conducted with the approval of the ethics committee at the authors’ university.

Supplemental material

Supplemental Material

Download MS Word (136.5 KB)

Disclosure Statement

No potential conflict of interest was reported by the authors.

Data Availability Statement

The authors are willing to share the data on which this manuscript is based on request.

Supplementary Material

Supplemental data for this article can be accessed on the publisher’s website

Additional information

Funding

The research reported in this paper was supported by a grant from the German Research Foundation to the first author [Kr 972/11-1]. Anja Berger is now at the Berlin School of Economics and Law, Berlin, Germany.

References