Previously reported empirical evidence suggests that when conflict arises,
military alliances are not reliable; state leaders should only expect their
alliance partners to join them in war about 25% of the time. Yet. theoreti
cal arguments explaining the choices of leaden to form cooperative agreemen
ts are at odds with such empirical evidence. This puzzling gap between theo
ry and evidence motivates a reconsideration of previous measures of allianc
e reliability. Many alliance treaties include specific language regarding t
he circumstances under which the alliance comes into effect, often limiting
obligations to disputes with specific target states or in specific geograp
hic areas, and many treaties do not go so far as to require states to join
in active fighting. Considering the specific obligations included in allian
ce agreements provides an improved estimate of the propensity of states to
honor their commitments. Results show that alliances are reliable 74.5% of
the time.