Background: This article builds on a companion study using the same national dataset, which found that childhood physical victimisation predicts parents' very severe physical abuse of their own children, and that knowledge of a protective law is associated with a weakening of this transmission. Whereas that study examined behaviour, the present study investigates parents' beliefs about discipline, which were not previously addressed. Prevention models premised on raising awareness of harm implicitly assume that new information about risk translates into changed practice — an assumption that literatures on motivated reasoning and attitude–behaviour gaps increasingly call into question. Objectives: This study examines whether Tunisian parents' recognition of punishment’s harms coexists with their continued endorsement and use of violent disciplinary practices. It also identifies which beliefs — harm recognition, perceived benefits of punishment, or authoritarian parenting norms — are independently associated with acceptance and use of such practices. Method: Using the same nationally representative, population-weighted sample of 1,002 Tunisian parents as the companion study, we constructed six scales measuring perceived benefits of punishment, perceived harms of punishment, authoritarian parenting beliefs, acceptance of physical and psychological disciplinary methods, and past-year use of psychological aggression and moderate physical punishment (all α = .71–.92). Weighted regression analyses examined associations among these constructs, controlling for childhood victimisation, knowledge of the law, and demographic characteristics. Results: Harm recognition was high on average and essentially uncorrelated with perceived benefit of punishment (r = -.04, ns), indicating that most parents hold both beliefs concurrently rather than one displacing the other. Harm recognition was not a significant predictor of acceptance or use once other beliefs were controlled (β = -.05 and -.03, both ns). Instead, two distinct pathways emerged: benefit belief was the strongest predictor of what parents found acceptable (β = .16, p < .001), while authoritarian beliefs specifically predicted what parents actually did (β = .09, p < .05). Childhood victimisation was the strongest predictor overall of both acceptance and use. Acceptance itself strongly predicted use (β = .30, p < .001), and was entered as an additional covariate to test whether the victimisation–use association was robust to potential omitted-variable bias, consistent with a pathway through which beliefs are expressed behaviourally, though victimisation retained a significant direct association with use even after acceptance was controlled. Descriptively, just over one in four parents (26.3%) scored above the sample median on both harm recognition and acceptance of violent discipline. Conclusion: Violent discipline in this sample was not sustained primarily by ignorance of its harms. Consistent with motivated-reasoning and attitude–behaviour-gap accounts, harm recognition coexisted with, rather than displaced, beliefs about the benefits and legitimacy of punishment, and it was these latter beliefs — not harm recognition — that were associated with parents' acceptance and use of violent discipline. Interventions that rely solely on communicating risk may be insufficient; approaches that also engage perceived effectiveness and normative beliefs about parental authority warrant consideration.