Secondary analysis of a large German male community sample (N = 8,554) comparing multiple viewing time (VT) scoring algorithms for assessing sexual interest in children, evaluating reliability, convergent validity, confound susceptibility, and classification prevalence rates.
VT validity was robust across scoring changes (AUCs 0.70–0.94), but reliability varied widely. The best overall difference score was **Dlower 75%** (α = 0.89–0.93), which excludes the highest VT per category; **DSD** (difference standardized by SD) showed the highest validity (AUC up to 0.94) and least confounding by general processing speed. Prevalence of sexual interest in children ranged dramatically by method: 0.1% (self-report preference) to 42.9% (progressive peak score), with VT difference indices yielding 5.8–6.8%.
- Exploratory secondary analysis without pre-registration; results need replication in forensic or self-identified pedophilic samples. - Self-report criteria used for validation have known social desirability bias and very low base rates, limiting logistic regression performance. - Sample restricted to German adult males; race/ethnicity not recorded, limiting generalizability.
Clinicians using VT measures should select the scoring algorithm based on their goal: use **Dlower 75%** when reliability is the priority (e.g., repeated assessments), and **DSD** when maximizing predictive validity and minimizing processing-speed confounds. Prevalence estimates from VT tasks vary enormously by scoring method—context and algorithm choice must always be reported alongside any classification rate.
Explore related topics
