Responsive regulation is a general theory of how to steer the flow of events. This article seeks to understand when violence is and is not defensible as an enforcement escalation. It specifies limits on the claim of responsive regulatory theory that a tough enforcement peak to a regulatory pyramid helps drive regulation down to persuasion at the base of the pyramid. Those limits are about the counterproductive effects of violence at the peak of an enforcement pyramid. Erica Chenoweth and her colleagues show that nonviolent civilian resistance to regimes is twice as likely as armed struggle to succeed. Nonviolence complemented by a violent radical flank is less effective than disciplined nonviolence. This refutes the "benign big gun" aspect of responsive regulatory theory as a general theory of the regulation of social action. The theory implies that capacity to escalate to armed struggle at the peak of a regulatory pyramid should empower resistance. Can responsive theory be adapted to this empirical challenge? Can that adaptation show a productive path to an ethics of when to constrain escalation to violence as an option at the peak of all kinds of regulatory pyramids? Lessons are drawn from how Nelson Mandela's struggle against apartheid opened nonviolent paths to transformation without total renunciation of violence.