The existing studies on determinants of alliance reliability focus exclusively on alliance types and alliance attributes. One big weakness of this approach is that it depicts the decision making of upholding or disregarding alliance commitments as largely externally determined, and thus downplays the role of individual states. The present study assumes that states are rational utility maximiers and contends that the decision of whether or not to uphold alliance commitments is determined by the utility calculations at the time when an alliance is formed and the time when an ally is called upon. An ally will assist its defense pact partner under attack only when the expected costs of upholding commitment are lower than the expectations held at the time of alliance formation. Empirical testing, however, fails to lend strong support for the expectation. A discussion of the performance of the model leads to some interesting conclusions about the existing literature and possible future studies.