“Anchoring” results from insufficient adjustment up or down from an original— often arbitrary—starting value. Six sets of surveys were designed to assess the effects of anchoring on subjective likelihood estimates of a nuclear war. Based on responses from 1600 students, results indicated that: (a) likelihood estimates were strongly susceptible to anchoring; (b) neither likelihood estimates nor the effects of anchoring were significantly influenced by the ease with which respondents could imagine a nuclear war (outcome availability), by instructions to list the most likely path to nuclear war (path availability), or by casting the problem in terms of the avoidance, rather than the occurrence, of nuclear war; (c) the effects of anchoring extended to estimates concerning the efficacy of strategic defenses; and (d) likelihood estimates were affected by anchoring even after correcting for social demand biases. In estimating the likelihood of nuclear war and otherwise attempting to “think the unthinkable”, many students responded in a manner consistent with denial; the paper concludes with a discussion of these individuals.