Abstract
The present study examines differences in cybersex behavior related to gender and sexual orientation. Data were from 4,492 individuals being treated for sexual addiction, including 3,561 heterosexual men, 164 homosexual men, 131 bisexual men, 429 heterosexual women,18 homosexual women, and 61 bisexual women. Measures included the Sexual Dependency Inventor –Revised (SDI-R; Carnes & Delmonico, 1996) and the Sexual Addiction Screening Test-Revised (SAST-R; Carnes, Green & Carnes, 2010). Similarities and differences across genders and sexual orientations are reported. Multiple regression results demonstrated a general pattern in which isolative sex behaviors predicted more Internet use and relational sex behaviors predicted less use.
Notes
Cohen (Citation1988) suggests that when the likely effect size range is not known, a small effect could be represented by an r of .100, but this would only account for 1.0% of the shared variance between two variables. When correlating two variables from survey instruments within-method variance alone should exceed r = .100 routinely. Setting our threshold at r = .250 allows examination of small effects that are likely to be more than just within-method variance and random error, which can on average account for 26.3% and 32.0% of the observed relationship respectively (Cote & Buckley, Citation1987).
As readers sophisticated in multiple regression will observe, the standardized beta weights exceeding 1.00 absolute value suggest either a suppression effect or a problem with multicollinearity, the latter being most likely. Though none of the other groups produced tolerance of variance inflation factor (VIF) estimates to suggest multicollinearity problems, the homosexual women group had tolerance and VIF estimates that were extremely high for several predictors, both in the model predicting with SDI-R Clinical scales, and the model using SDI-R Behavior scales.
Although most of the SDI-R scales correlated with SAST-R Internet in the group of homosexual women, the sample was so small that all findings for that group should be considered preliminary, at best, and likely to diminish in a larger sample.
The exceptions were for the homosexual women group, where the strongest zero-order correlates of the SAST-R Internet scale were the SDI-R Behavior scales Pornography and Pain Exchange, and the SDI-R Clinical scales ER 2-Boundary Violation and Preoccupied Sadomasochism. Again we restate the caveat that this was the smallest sample we examined, and the more extraordinary results for this group may be somewhat artifactual, and may diminish in a larger sample.