¹ See Hart, B. H. Liddell, Deterrent or Defense, New York, 1960Google Scholar, esp. ch. 10. That both sides would have recognized the superiority of prepared defenses in 1900, when most nations were ignoring rather than profiting from the lessons of the U.S. Civil War, is unlikely, but we presume them unusually enlightened.

² The classic development of this finding is in Lanchester, F. W., Aircraft in Warfare, London, 1916.Google Scholar

³ For a valuable historical account of these and other examples, and some intriguing if controversial conclusions, see Huntington, Samuel P., “Arms Races: Prerequisites and Results,” in Friedrich, C. J. and Harris, S. E., eds., Public Policy, Cambridge, Mass., 1958, pp. 41–86.Google Scholar For a more general and abstract treatment of arms races, see Rapoport, Anatol, Fights, Games, and Debates, Ann Arbor, Mich., 1960.Google Scholar

⁴ See Schelling, T. C., The Strategy of Conflict, Cambridge, Mass., 1960Google Scholar, esp. ch. 9.

⁵ To illustrate the principle, once again we can calculate oversimply and precisely where policy-makers never can calculate precisely, even when they can face the calculation psychologically. Suppose a country estimates that its feared enemy is unlikely to attack, but may, with the probability in any single year only one in twenty. If it expects this probability to remain the same per year, and decides that it will never attack, the cumulative probability of peace over a decade is, barring accidents, about three chances in five. Assuming arbitrary value weights, we can get this situation:

Here our country prefers peace to overwhelmingly victorious war 6 to 5, but victorious war over disastrous war 5 to 1, so that it can improve its expected “score” from 4 to 5 by rejecting deterrence in favor of “preventive” war.

Yet this illustration merely sets the stage for a compounding of reciprocal fears of attack. If the enemy estimates mat you will estimate as above, clearly his incentive to pre-empt soars. This means that your original estimate of the probability of his attack was wildly optimistic, as his must have been of your intent, and so on in an explosively interacting sequence.

⁶ The same percentage survival rate applied to a larger force will yield a greater absolute survival. In addition, as the attacker moves from initial parity to pre-attack superiority, he may go beyond the tactic of only one missile aimed at one missile, with resultant lower kills per missile. Higher absolute-force levels, moreover, will make defensive calculations less sensitive to errors in intelligence about enemy capabilities.

⁷ See Brodie, Bernard, Strategy in the Missile Age, Princeton, N.J., 1959, especially pp. 160–72.Google Scholar Dissenters have not yet, to my knowledge, provided the calculations, inter alia, of protection against radioactive fall-out in combat areas that would be consistent with political and military prosecution of organized war.

⁸ With the kill probability of defense set at .3, 70 per cent of arriving missiles should survive. The kill probability of an undefended city was set at .8 to allow for unreliability and inaccuracy. Against a defended city only 70 per cent survival of the 80 per cent that arrive would be expected, yielding a .56 single-shot kill probability. The city survival probability would then be 1–.56, or .44, but if two extra missiles were shot at it, this probability would be cubed, falling to .085, or appreciably less than the assumed .20 single-shot survival probability of an undefended city.

In the interests of simplicity, we here and elsewhere overlook all sophisticated consideration of the added elements of uncertainty in Ruritanian and enemy calculations that would be induced. Our use of expected values makes discussion too simple, but introduces no significant errors for our restricted purposes.

⁹ The military reader will recognize the old sound point that some guide to priorities is better than no guide to allocations at all. I have elsewhere tried to move in turn beyond the inadequacies of blunt military priorities to what, in economists' jargon, are income elasticities. See “Some Complexities in Military Planning,” World Politics, XI, No. 4 (July 1959), pp. 553–77.

¹⁰ The first-order civil defense measures cannot be detailed here. They might be constrained, however, by these criteria: (1) favorable exchange rates against enemy offensive counters; (2) catching up with, rather than outpacing, enemy civil defenses; (3) consistency with strike-second protection of people when the enemy tries to preserve them as hostages, e.g., shelters in cities that protect against the fall-out from an enemy first-strike against military targets. Each of these constraints limits our program markedly, and surely no resultant civil defenses could be provocative in any major sense.

¹¹ An alliance-wide collective retaliatory force, notably for NATO, may be a good compromise, depending upon its design and control.

¹² The first systematic account of this policy, to my knowledge, was that of Kaplan, Morton A., The Strategy of Limited Retaliation, Policy Memorandum No. 19, Center of International Studies, Princeton University, 1959.Google Scholar See especially Schelling, op.cit., and Szilard, Leo, “How to Live with the Bomb and Survive,” Bulletin of the Atomic Scientists, XVI, No. 2 (February 1960), pp. 59–73.Google Scholar

¹³ I should not like the country to adopt as full-fledged and belligerent-appearing a central war preparedness as that advocated by Kahn, Herman in On Thermonuclear War (Princeton, N.J., 1960)Google Scholar, mainly because to go that far toward a “credible first strike” posture jeopardizes the longer-term prospects for arms control. My policy dissent, however, is coupled with great admiration for the book, which is enthusiastically recommended for, among other things, a much fuller treatment of many of the issues discussed in this article.