This Article's claim is that deterrence in criminal law is composed of four oft-competing goals that ought to be taken into account when calibrating sanctions. Consider an offender, such as an armed robber or kidnapper, who threatens to commit a yet more serious crime unless the victim surrenders. The first goal, or dimension, of deterrence is to deter the threat, or crime in general, and this suggests a high sanction for the armed robbery or other threatening crime. A second goal is to deter escalation, in this case the execution of the threat once issued. This goal normally suggests a large gap between the sanctions for lesser and greater crimes, in order to discourage escalation. Given a natural or practical ceiling to criminal penalties, the suggested gap often implies relatively low sanction for the lesser crime, or threat, and some sacrifice of direct deterrence. Two other goals, or dimensions of deterrence, have been unrecognized by academic commentators and legislatures. By extending the analysis to include behavior by victims, it becomes apparent that law ought to make the offender’s threat incredible to victims. If the threat can be weakened, victims will be less inclined to submit, and then offenders will have less reason to threaten in the first place. This third dimension suggests a large gap between the sanctions for the lesser (threat) and greater (execution) crimes; if the gap is small, victims will recognize that the offender has little to lose from continuing on and executing his threat, and these victims will submit and make the criminal’s threat profitable, thereby encouraging yet more criminal activity. A fourth, equally novel, dimension of deterrence is the ability of law to make incredible the implicit promise by a threatener not to escalate if the victim submits. When this implicit promise is incredible, victims will again tend not to submit to the threat, and offenders will in turn issue fewer threats. To achieve this goal, the sanction for repeating a threat following submission by the victim should be low; victims will be less apt to submit to a threat if they realize that submission will often bring on another crime, or threat. The idea — which should be applied with caution — is familiar from blackmail; a victim who thinks the blackmailer will return for further payments will not pay the first time, and the blackmailer, in turn, will be discouraged.Optimal deterrence requires that criminal law should be attentive to the four dimensions described above. This Article presents a model for how law can do so and better accomplish its deterrent function.