Introduction: Collective punishment, i.e., punishing an entire group for some members’ wrongdoing, is morally controversial yet widely applied, particularly when sanctioning individuals is infeasible. Although often regarded as unfair, collective punishment is commonly defended on consequentialist grounds: By holding groups jointly responsible, it may deter individual transgressions and generate overall societal benefits. Yet deterrence is far from guaranteed: In practice, collective punishment often creates a threshold social dilemma, which in theory can encourage non-compliance. Nonetheless, deterrence may still arise through at least two mechanisms: People may transgress less under collective punishment (relative to no punishment) because they fear triggering a collective cost. Such punishment may also promote compliance through communicating what is disapproved and altering social norms. This research disentangles these mechanisms to clarify how collective punishment can generate deterrence and when it may fail to do so.

Method: We examined how collective punishment affects transgression, particularly dishonesty, using a variant of ‘cheating paradigms’. Specifically, 1,556 online participants played a Mind Game, where they could lie for personal gain without being identified. The Game permits inference of only group-level dishonesty, thus mirroring real-world contexts where collective punishment is often applied. We compared three between-participants conditions: (i) no punishment, (ii) an individually non-deterrent (i.e., not offsetting the private gains from transgression) collective punishment that would reduce payoffs for all when group-level dishonesty exceeded a threshold, and (iii) communication of an injunctive norm about a maximal level of dishonesty in the group—equal to the punishment threshold in (ii)—without punishment. Comparing (ii) and (iii) isolates the effect of the punitive component of collective punishment from its function of communicating normative disapproval. We assessed personal and social norms across conditions, incentivising norm measures where appropriate.

Results: Collective punishment reduced dishonesty in the absence of individual identifiability. It lowered personal appropriateness of lying and reduced expectations of others’ dishonesty, which in turn predicted participants’ own honesty. However, merely communicating behavioural expectations led to comparable normative shifts and even less dishonesty. These findings suggest that the punitive component of collective punishment can crowd out its communicative effects.

Discussion: Our results point to competing forces behind the deterrence of collective punishment: norm activation that curbs transgression versus rational calculation that can undermine deterrence. By disentangling them, this research clarifies when and why collective punishment is (in)effective, informing both normative and policy debates about its use.