Skip to main content

Testing the Confluence Model of the Association Between Pornography Use and Male Sexual Aggression: A Longitudinal Assessment in Two Independent Adolescent Samples from Croatia


According to confluence model theorizing, pornography use contributes to sexual violence, but only among men who are predisposed to sexual aggression. Support for this assertion is limited to cross-sectional research, which cannot speak to the temporal ordering of assumed causes and consequences. To address this issue, we employed generalized linear mixed modeling to determine whether hostile masculinity, impersonal sexuality, and pornography use, and their interactions, predicted change in the odds of subsequently reported sexual aggression in two independent panel samples of male Croatian adolescents (N1 = 936 with 2808 observations; N2 = 743 with 2972 observations). While we observed the link between hostile masculinity and self-reported sexual aggression in both panels, we found no evidence that impersonal sexuality and pornography use increased the odds of subsequently reporting sexual aggression—regardless of participants’ predisposed risk. This study’s findings are difficult to reconcile with the view that pornography use plays a causal role in male sexual violence.

This is a preview of subscription content, access via your institution.

Fig. 1


  1. The PROBIOPS data have been used for several publications concerning adolescent pornography use. Specific topics have included: sexual activity, body surveillance, compulsive pornography use, religiosity, subjective well-being, sexual permissiveness, and sexual risk taking (http:// For further details, see

  2. The full 10-item scale was employed to measure hostile masculinity in the Zagreb panel, with a 3-form planned missing design (Graham, Taylor, Olchowski, & Cumsille, 2006). All participants were randomly assigned to one of three forms of the scale. Each of these forms was composed of 8 items. Missing responses were then imputed with full information maximum likelihood-based regression approach. For comparison purposes, the current study focused on a reduced set of 5 items that were employed in the Rijeka panel.

  3. We had originally pre-registered an operationalization of impersonal sexuality that was based solely on number of partners. The results with respect to impersonal sexuality did not differ substantially from those presented below and can be found here: Upon further consideration, we were concerned that this operationalization departed too far from Malamuth’s formulation of impersonal sexuality and updated our operationalization and analyses accordingly.

  4. In cases when participants lacked data for a specific wave but their reported number of sexual partners was stable across the preceding and following assessments, the missing data were replaced with the number of sexual partners that they reported before and after the gap. For example, if a participant reported one lifetime partner at both T1 and T3, missing data at T2 were replaced with one lifetime partner.

  5. In cases where missing information occurred between an indication of no intercourse and first reported intercourse, the wave of first reported intercourse was modified to reflect the possibility that first intercourse occurred during the period in which data were missing. In this case, the assumed age at first intercourse was calculated by averaging between the wave of first reported intercourse and the first wave of missing data that preceded the report of first intercourse. For example, if a participant reported no intercourse at T1, and first intercourse at T4, but had missing data at T2 and T3, then their assumed wave of first intercourse was defined as (T4 + T2)/2 = T3. A similar approach was used when participants were lost to follow-up without identifying a wave of first intercourse. In this case, their wave of first intercourse was defined as the average between the last wave they reported no intercourse and T6. For example, if a participant reported no intercourse at T1, but supplied no further sexual behavior data, their assumed age of first intercourse was (T1 + T6)/2 = T3.5.

  6. Corrections for multiple comparisons were not employed in these tests to improve sensitivity to detect possible differences between the retained and omitted subsamples (minimizing type II error).

  7. The initial pre-registration involved a planned analysis that was limited to the number of available observations and an earlier version of the paper based on this approach can be found here: The current approach, involving multiple imputation, was only adopted after receiving feedback from the peer review of the original paper. The only inferential difference that emerged after data imputation was that hostile masculinity was associated with an increase in the probability of subsequent sexual aggression in the Rijeka panel.

  8. For details, see

  9. We had originally pre-registered a plan that included a measure of testosterone among the control variables in the Rijeka panel. It was ultimately not included in the analyses presented below. Testosterone measurement was dropped for three reasons. First, it was uncorrelated with all variables of interest in this study. Second, its inclusion would have severely reduced the number of available observations (much more so than the other control variables), due to the fact that only 252 male adolescents provided a saliva sample. Finally, its inclusion would have interfered with the comparability of results across the two panels.

  10. See


Download references


This work was supported by the Croatian Science Foundation (Grant #9221 awarded to the last author) and the University of Zagreb.

Author information

Authors and Affiliations


Corresponding author

Correspondence to Taylor Kohut.

Ethics declarations

Conflict of interest

The authors declare that they have no conflict of interest.

Ethical Approval

All procedures of human participants were in accordance with the ethical standards of the Ethical Research Committee of the University of Zagreb and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.

Informed Consent

Informed consent was obtained from all individual participants included in the study.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Kohut, T., Landripet, I. & Štulhofer, A. Testing the Confluence Model of the Association Between Pornography Use and Male Sexual Aggression: A Longitudinal Assessment in Two Independent Adolescent Samples from Croatia. Arch Sex Behav 50, 647–665 (2021).

Download citation

  • Received:

  • Revised:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI:


  • Pornography
  • Sexual aggression
  • Confluence model
  • Adolescence