When theoreticians think about nuclear deterrence, often they focus on a nasty choice between two rival package deals. The two have gone by various names over the years, but let me take the paired epithets: MAD (Mutual Assured Destruction) versus NUTS (Nuclear Use Theorists). Each package is a bundle of policies: centrally, policies for the procurement of strategic nuclear forces and conditional intentions about how to use those forces in case of war. I think we can break up the packages and keep only part of each. What we get may be in some sense MAD and in some sense NUTS – for the terms are elastic – but I hope it is the better half of both.

In a debate between MAD and NUTS, each side may say that the other's policies involve a twofold risk: a grave moral risk of committing massacres and a grave prudential risk of inviting and undergoing massacres. If they say so, they are right. The contest between these two repugnant alternatives gives nuclear deterrence itself a bad name. How does the very idea of nuclear deterrence turn into the nasty choice between MAD and NUTS? Does it have to happen? Is there no way around it?

MAD: IF YOU CAN'T BE CREDIBLE, BE DREADFUL

To trace the reasoning that drives us MAD, start with a simple conception of nuclear deterrence. We deter the enemy from doing X by threatening that if he does, then we will punish him by doing Y.