Objective: Paternity is uncertain, so if paternal feelings evolved to promote fitness, we might expect them to vary in response to variables indicative of paternity probability. We therefore hypothesized that the risk of lapses of paternal affection, including abusive assaults on children, will be exacerbated by cues of non-paternity. Methods: Cross-sectional study of 331 Brazilian mothers, interviewed about 1 focal child (age 1-12) residing with her and the putative father. Child physical abuse was assessed using the "Conflict Tactic Scales: Parent Child" (CTSPC). Two potential cues of (non) paternity were (1) whether the parents co-resided when the child was conceived, and (2) whether third parties allegedly commented on father-child resemblance. Data were analyzed through multiple logistic regressions. Results: Mothers reported child physical abuse by 15.9% (95% CI 4.6-27.1) of fathers who had not cohabited with them at conception, compared to 5.9% (95% CI 3.1-8.7) of those who had. The odds ratio for abuse by fathers who had not cohabited at conception in relation to those who had-adjusted for income, education, age, sex of child, whether child was first born, household size, time father spent with child, and alcohol abuse and drug use by father-was 4.3 (95% CI 1.4-13.8). Mothers reported abuse of 7.0% (95% CI 4.0-10.0) of children who purportedly resembled their fathers, versus 8.7% (95% CI 0.2-17.1) of those who did not. Conclusion: According to maternal reports, not having co-resided at conception quadrupled the chance of child physical abuse by currently co-residing Brazilian fathers. The reported prevalence of abuse was unrelated to reported allegations of father-child resemblance. (Contains 3 tables.)