We identify conditions for which perceptual differences diminish the r
isk of war and others for which differences in perceptions increase th
e likelihood of warfare. We show empirically on a data set concerning
469 European disputes between 1816 and 1970 that these game-theoretic
conditions are a significant improvement over the decision-theoretic v
iew suggested by Bueno de Mesquita and Lalman (1986) and help resolve
some concerns raised by Jervis (1989). The expectations from the game
are empirically robust in the face of bipolarity/multipolarity, a vari
able commonly thought to be important by structural realists or neorea
lists. Differences in perceptions need not be thought of as purely idi
osyncratic phenomena with unpredictable consequences for events. Nor i
s it necessary to rely on ex post, thick analysis of specific events t
o portray the relationship between perceptions and outcomes. Rather, w
e identify a formulation that allows ex ante predictions of how decisi
on makers will behave in crisis situations based on their perceptions
of the circumstances they find themselves in.