Abstract. Why punish a whole group for the actions of only some members? Conventional wisdom explains away collective punishment by an inability to identify and sanction the offenders individually, yet people are often punished for the behavior of their peers even when it is perfectly observable. I argue that rulers can use collective punishment to enforce a structure of collective responsibility, and study the incentives that this structure generates with a general model featuring a ruler and a set of people. There is a trade-off between "vertical" repression, how a ruler can directly deter disobedience by the threat of punishment, and ``horizontal” repression, how a ruler can incentivize people to enforce each other's obedience. Holding people responsible for others creates incentives to self-police but dilutes vertical deterrence, since obedience no longer guarantees safety from the punishment. When people have altruistic concerns for each other, collective punishments can be effective as an extension of individual punishment, but they may backfire as people are less willing to police others. Even if the ruler can directly incentivize policing, the basic trade-off between horizontal and vertical deterrence persists.