This paper develops an incomplete information model of extended deterr
ence relationships. It postulates players who are fully informed about
the costs of war and all other relevant variables, save for the value
s their opponents place on the issues at stake, i.e., the pawn. We pro
vide consistent and intuitively satisfying parallel definitions for tw
o types of players, Hard and Soft, in terms of the parameters of our m
odel. We also answer several particular questions about the strategy c
hoices of players in an extended deterrence relationship and, by ident
ifying all the Perfect Bayesian Equilibria of the game model we constr
uct, specify typical behavior patterns. Our most general finding is th
at an Extended Deterrence Game always has a unique Perfect Bayesian Eq
uilibrium with a rather simple form. A challenger initiates for certai
n if the pawn is valuable enough to it and never challenges otherwise,
Likewise, a defender always resists if the pawn is valuable enough an
d never resists otherwise. In addition to identifying threshold values
for optimal strategic choice, we provide existence conditions for the
two distinct types of Perfect Bayesian Equilibria. This permits us to
identify the exact set of conditions associated with bluffing strateg
ies and to develop a theory of the necessary and sufficient conditions
for extended deterrence crisis initiation and its resolution.