Arms Races: A Test of Two Models

This is an investigation of two models of arms races: Lewis F. Richardson's linear, additive, symmetric model and a curvilinear, asymmetric alternative based on the same interaction paradigm but with empirical equations from psychophysical experiments on perception as premises. Two hypotheses derived from each model are tested using data from seven arms races. In general, the data give stronger support to the curvilinear, asymmetric model which predicts the follower in the arms race will increase its armaments through time as an exponential function of the leader's armaments, and the leader will increase its armaments as a power function of the follower's efforts at the previous time period. The conclusions are based on the relative fit of the models to the seven sets of data, parsimony, reasonableness of the parameter estimates as adjudged by their signs and magnitudes, and predictions about the outcomes of the races--war or armed truce.