Analytic Plan ------------- **Lagged Analysis of Confluence Model Relationships.** If pornography use causes sexual aggression among men who are high in other risk factors for sexual aggression (e.g. hostile masculinity and impersonal sexuality), then self-reported sexual aggression at any given wave should be function of a positive three-way interaction between pornography use, hostile masculinity and impersonal sexuality at the preceding wave—while controlling for self-reported sexual aggression at the preceding wave. In this model, a significant effect for pornography use would indicate a relationship between pornography use and subsequent change in the probability of self-reported sexual aggression across time. To this end, general linear mixed modeling will be used to predict the probability of self-reported sexual aggression at waves t + 1 with pornography use, hostile masculinity, and impersonal sexuality, all two-way and three-way interactions between these constructs, and self-reported sexual aggression at waves t. Dependencies between measurement occasions will initially be modeled by imposing compound symmetry on the covariance matrix for the residuals and this model will employ a logit link function to predict dichotomized sexual aggression. Because there is likely going to be a large drop in the size of the sample with addition of the control variables (due to missing information in one or more variables), the first model (without control variables) will be run twice. The first run will include the full sample of participants who have complete data on the key measures and the second will use a restricted sample that has complete data on all control variables. In a following step, additional time variant controls will be added to the model (masturbation frequency, testosterone, pubertal status, impulsiveness, sensation seeking and social desirability). The addition of a random intercept for individuals, random individual-level slope effects for the fixed pornography use components of the model (the main effect and all two and three-way interactions with hostile masculinity and impersonal sexuality), as well as alternative residual covariance matrix structures (e.g. ARH1) will be considered independently in following steps. **Average Time-Varying Confluence Model Relationships.** Previous cross-sectional research has found that pornography use is associated with self-reported sexual aggression, but only among men who are high in both hostile masculinity and impersonal sexuality. To test for this pattern of contemporaneous correlations in the current study, general linear mixed modeling will be used to predict the probability of self-reported sexual aggression at each wave with time varying assessments of pornography use, hostile masculinity, impersonal sexuality and all two-way and three-way interactions between these constructs. Dependencies between measurement occasions will be modeled by imposing compound symmetry on the covariance matrix for the residuals and this model will employ a logit link function to predict the dependent variable (dichotomized sexual aggression). Because there is likely going to be a large drop in the size of the sample with addition of the control variables (due to incomplete measures on one or more variables), the first model (without control variables) will be run twice. The first run will include the full sample of participants who have complete data on the primary measures of interest and the second will use a restricted sample that has complete data on all control variables as well. In a following step, additional time invariant control variables will be added to the model to account for the influence of variables that lie outside the Confluence model (masturbation frequency, testosterone, pubertal status, impulsiveness, sensation seeking, and social desirability). The addition of a random intercept for individuals, random individual-level slope effects for the fixed pornography use components of the model (the main effect and all two and three-way interactions with hostile masculinity and impersonal sexuality), as well as alternative residual covariance matrix structures (e.g., ARH1) will be considered independently in following steps. **Time Varying Use of Violent Pornography Use.** Following Baer et al. (2015)’s cross-sectional analysis, the probability of self-reported use of violent pornography at each wave is expected to covary with the hostile masculinity and impersonal sexuality interaction term at that wave. Thus, time varying hostile masculinity, impersonal sexuality and their interaction term will be used to predict self-reported use of coercive and violent pornography. Dependencies between measurement occasions will be modeled by imposing compound symmetry on the covariance matrix for the residuals and this model will employ a logit link function to predict use of violent pornography. Additional covariance structures for the residuals will also be explored in subsequent steps.