Participants and procedure
In March and April 2005, an on-line survey was conducted among 745 Dutch adolescents (48% boys, 52% girls) between 13 and 18 years of age (M = 15.5, SD = 1.69). Ninety-two percent of the respondents were Dutch, the remaining 8% belonged to other ethnic groups. For the study of sensitive issues, on-line surveys or, more generally, computer-mediated surveys have generally proven superior to other modes of interviewing (e.g., Mustanski, 2001). Respondents were recruited from an existing online panel managed by Intomart GfK, an established audience and media research institute in the Netherlands. Sampling and fieldwork were done by Intomart GfK. The institute had sampled the respondents in all parts of the Netherlands, partly through random telephone interviews, partly through respondents’ social networks, and requested from each respondent informed consent and, for the minors, parental consent before the adolescents took part in research. The response rate was 60%. Analyses showed that the gender, age, and formal education of our respondents did not deviate from official statistics. Prior to the implementation of the survey, institutional approval was obtained from our university.
Adolescents were notified that the study would be about sexuality and the internet and, if they wished, they could stop participation at any time. We took several steps to improve the confidentiality, anonymity, and privacy of the response process (Mustanski, 2001). On the introduction screen of the on-line questionnaire, we emphasized that the answers would be analyzed only by us, the principal investigators. Moreover, respondents were asked to make certain that they filled in the questionnaire in private. Finally, we assured the respondents that their answers would remain anonymous. That is, we explained explicitly that there was no possibility for the principal investigators to identify who had filled in the questionnaire and that, on the other hand, Intomart GfK could not see what the respondents answered. Intomart GfK did not link respondents’ answers in our questionnaire to their names and contact information, and only provided us with the background variables plus the answers to our questionnaire. This procedure has proven successful in various other studies on sensitive issues and ensures the protection of respondents’ anonymity. Completing the questionnaire took about 15 min.
For the regression analyses presented in this article, we had complete data from 674 of the 745 respondents who had started the questionnaire. Further analyses indicated that, in terms of age, gender, ethnicity, and formal education, the 674 respondents for whom we had complete data did not meaningfully deviate from those respondents for whom we lacked complete data.
Exposure to sexually non-explicit content in magazines
We operationalized this concept with adolescents’ exposure to three Dutch magazines that typically include some sexual content (e.g., Actueel). These magazines have a recreational orientation. They report about sports, crime, and erotica, typically in a slightly sensationalist way. Adolescents were asked how many issues of a particular magazine they usually read; response categories ranging from 1 (none) to 7 (all issues). The factor structure was uni-dimensional (explained variance 80%), and resulted in a Cronbach’s Alpha of .87 (M = 1.27, SD = .82).
Exposure to sexually non-explicit content on television
We operationalized this measure by drawing on a question that asked adolescents to what extent they were interested in various types of television programming. Based partly on results from content analyses (e.g., Kunkel et al., 2005; Pardun et al., 2005; for a review, see Ward, 2003), we included television genres that present at least some sexually non-explicit content: soap operas (e.g., Good Times, Bad Times), music shows (e.g., on MTV), comedy series (e.g., Friends, Sex in the City), romantic movies (e.g., When Harry Met Sally), romantic shows (e.g., All You Need Is Love), and action series (e.g., 24, JAG). Response categories ranged from 1 (not interested at all) to 4 (very much interested). The factor structure of the scale was unidimensional (explained variance 42%), Cronbach’s alpha was .72 (M = 2.82, SD = .69).
Exposure to sexually semi-explicit content in magazines
This measure was operationalized with two items—adolescents’ exposure to Playboy and Penthouse. Adolescents were asked to indicate how many issues of Playboy and Penthouse they usually read, and the response categories ranged from 1 (none) to 7 (all issues). The two items correlated at .80, Cronbach’s alpha was .89 (M = 1.16, SD = .71).
Exposure to sexually semi-explicit content on television
This measure was operationalized with adolescents’ frequency of exposure to three sexually semi-explicit television programs (Sexcetera, Sex Court, Latin Lover). The respondents were asked how often, in the past year, they had on average watched the three television programs. Response categories ranged from 1 (never) to 5 (several times a week). When the three items were entered into a factor analysis, they formed a unidimensional scale (explained variance 78%). Cronbach’s alpha was .85 (M = 1.28, SD = .59).
To test whether exposure to sexually semi-explicit content on television was also empirically distinguishable from exposure to sexually non-explicit content on television, we z-transformed the items used for the two scales and subjected them to a factor analysis with varimax rotation. The resulting two factors exactly mirrored the operationalization of the two scales, which indicates that the two constructs are empirically independent.
Exposure to sexually explicit material in magazines
Adolescents were asked to indicate how often they had, on average, read erotic magazines in the past year. Response categories ranged from 1 (never) to 5 (several times a week) (M = 1.35, SD = .76). In Dutch, the term erotic magazines is often used as a euphemism for sexually explicit Dutch magazines, which present vaginal, oral, and anal sex in unconcealed, uncensored ways. To test whether exposure to sexually explicit magazines (i.e., erotic magazines) was different from what we defined as exposure to sexually semi-explicit magazines (i.e., Playboy and Penthouse), we correlated the three items. Exposure to erotic magazines was only moderately correlated with exposure to Playboy, r = .24, p < .001, and Penthouse, r = .15, p < .001. The relatively low correlations support the expected difference between exposure to sexually semi-explicit material (as displayed in the Dutch versions of Playboy and Penthouse) and the more explicit material available in Dutch erotic magazines.
Exposure to sexually explicit material on video/DVD
The respondents were asked how often, in the past year, they had, on average, watched a pornographic movie. Response categories ranged again from 1 (never) to 5 (several times a week) (M = 1.43, SD = .90).
Exposure to sexually explicit pictures on the internet
Adolescents were asked how often, in the past 6 months, they had, on average, looked at on-line pictures in which people are having sex. The response categories were 1 (never), 2 (less than once a month), 3 (1-3 times a month), 4 (once a week), 5 (several times a week), and 6 (every day), (M = 1.87, SD = 1.29).
Exposure to sexually explicit movies on the internet
We asked adolescents how often, in the past 6 months, they had, on average, watched on-line movies or movie clips in which people are having sex. The response categories were the same as for exposure to sexually explicit pictures on the internet (M = 1.82, SD = 1.28).
Pre-tests revealed that adolescents did not need more elaborate explanations as to the content of the two items that we used to measure exposure to sexually explicit on-line pictures and movies. The respondents were aware that the two items were about sexually explicit content and their purposeful exposure to them.
Women as sex objects
We largely followed an operationalization by Ward (2002), but adjusted it slightly for the use among Dutch adolescents. Furthermore, we replaced two items of Ward’s original scale (i.e., whistling at shapely women, attractive women give men prestige) with two items that more strongly refer to sex (i.e., “Unconsciously, girls always want to be persuaded to have sex” and “Sexually active girls are more attractive partners”). Response categories ranged from 1 (disagree completely) to 5 (agree completely). In a subsequent factor analysis with varimax rotation, the three items of Ward’s scale that dealt with face and body care and the importance of women’s appearance to attract men loaded on a separate factor. As a result, we eventually measured the concept of women as sex objects with the remaining three items from Ward’s scale (i.e., “An attractive woman should expect sexual advances;” “It bothers me when a man is interested in a women only if she is pretty;” “There is nothing wrong with men being primarily interested in a woman’s body”) plus the two items we had added. These five items formed a uni-dimensional scale (explained variance 50%), with a Cronbach’s alpha of .75 (M = 2.81, SD = .74).
Age and gender
The measurement of age and gender was straightforward. Boys were coded with 0, girls with 1.
We operationalized respondents’ race/ethnicity as a dichotomy where 0 meant Non-Dutch, and 1 meant Dutch.
We operationalized sexual experience with three items: mutual masturbation, oral sex, and coital sex. Pre-tests revealed that adolescents had no problems understanding the terms. Respondents were asked whether they had performed one or more of the three behaviors. To avoid problems with the log-transformation of the resulting scale, experience with a particular sexual behavior was coded as 2; lacking experience with a particular behavior was coded as 1. The three items loaded on one factor (explained variance 81%). We first summed these items and then divided them by the number of items to form a scale. The resulting alpha was .88 (M = 1.30, SD = .41).
Education was measured on a 5-point scale that represented the different educational levels at which Dutch adolescents can be (M = 2.75, SD = 1.22). The response categories were 1 (Elementary education, lower vocational education), 2 (Lower general secondary education), 3 (Intermediate vocational education), 4 (Higher general secondary education, pre-university education) and 5 (Higher vocational education, university). It should be noted that, in the Netherlands, adolescents of the same age may have different formal levels of education. This also shows in a modest correlation of r = .23 between formal education and age.
Adolescents’ socio-economic resources were operationalized as a combination of two measures: the profession and the educational level of the family’s primary breadwinner (i.e., the person that earns most of the money in a family). For example, if the family’s breadwinner has a low formal education and does unskilled work, a low socio-economic status results. In contrast, somebody with a university degree and in a leading professional position would be assigned a high socio-economic status. The two measures were combined so that a 5-point scale resulted. The anchors of the resulting scale were 1 (low socio-economic status) and 5 (high socio-economic status) (M = 2.97, SD = 1.28).
Whether adolescents are religious was measured with the item “I am religious.” Response categories ranged from 1 (does not apply at all) to 5 (applies completely) (M = 2.23, SD = 1.33).
Pubertal status was operationalized with the Pubertal Status Scale developed by Petersen, Crockett, Richards, and Boxer (1988). The scale contains five items for boys—body hair, voice change, skin change, growth spurt, and facial hair—and five for girls—body hair, breast change, skin change, growth spurt, and menstruation. We removed the skin change item because Petersen et al. noted that it was the least reliable and least valid of the various items. Adolescents could indicate on a 4-point scale that ranged from 1 (has not started yet) to 4 (has already finished) whether each bodily change had already begun or had already finished. For validity reasons, we did not provide girls with the response category has already finished for the menstruation item. The internal consistency of the scale was .89 for boys (M = 2.91, SD = .83) and .82 for girls (M = 3.19, SD = .56).
Adolescents’ relationship status was measured with the question “Are you currently in a romantic relationship?” Adolescents who were single were coded 0 (67.9%); adolescents who had a relationship were coded 1 (32.1%).
Adolescents were asked whether they were gay/lesbian, bisexual, or heterosexual. In the present study, we included sexual orientation only to account for potential differences in heterosexual and non-heterosexual adolescents in their notions of women as sex objects. We therefore dichotomized the variable into non-heterosexual adolescents (coded 0, 6.8%) and heterosexual adolescents (coded 1, 93.2 %).
Inadvertent exposure to sexually explicit material on the internet
We asked adolescents how often, on average, they had by chance encountered explicit sexual content on the internet in the last 6 months. The sexual content we referred to was (a) pictures with clearly exposed genitals; (b) movies with clearly exposed genitals; (c) pictures in which people are having sex; (d) movies in which people are having sex; (e) erotic contact sites. On erotic contact sites, people can get in touch with other people for sexual purposes, for example by posting visually and/or textually sexually explicit profiles, which may also appear in a context of sexually explicit advertising or links. The response categories were 1 (never), 2 (less than once a month), 3 (1–3 times a month), 4 (once a week), 5 (several times a week), and 6 (every day). The items loaded on one factor (explained variance 67%), and resulted in a Cronbach’s alpha of .87 (M = 2.10, SD = 1.11).
We conducted hierarchical multiple regression analyses to test our research questions. Multiple regression analysis assumes that the variables have normal distributions, but sexual measures are typically positively skewed. Prior to the multiple regression analysis, we conducted Shapiro-Wilk tests for normality to determine whether the metric variables were normally distributed. As a result of the test, we had to log-transform the measures of religiosity, pubertal status, sexual experience, and all exposure measures. Because some of our measures might be strongly correlated, we checked whether there was evidence of multicollinearity between the variables. This was not the case; all variance inflation factors were clearly below the critical value of 4.0. The Cook-Weisberg test confirmed that our model met the assumption of homoskedasticity. For the investigation of the interaction terms between adolescents’ gender and their exposure to various sexual content, we centered the exposure variables around their means to avoid multicollinearity problems (Aiken & West, 1991).