Abstract: Countries that share rivers have a higher risk of military disputes, even when controlling for a range of standard variables from studies of interstate conflict. A study incorporating the length of the land boundary showed that the shared river variable is not just a proxy for a higher degree of interaction opportunity. A weakness of earlier work is that the existing shared rivers data do not distinguish properly between dyads where the rivers run mainly across the boundary and dyads where the shared river runs along the boundary. Dyads with fivers running across the boundary would be expected to give rise to resource scarcity-related conflict. while in dyads where the fiver forms the boundary conflict may arise because fiver boundaries are fluid and fuzzy. Using a new dataset on shared water basins and two measures of water scarcity, we test for the relevance of these two scenarios. Shared basins do predict an increased propensity for conflict in a multivariate analysis. However, we find little support for the fuzzy boundary scenario. Support for a scarcity theory of water conflict is somewhat ambiguous. Neither the number of river crossings nor the share of the basin upstream is significant. Dry countries have more conflict, but less so when the basin is large. Drought has no influence. The size of the basin, however, is significantly associated with conflict. Modernization theory receives some support in that development interacted with basin size predicts less conflict, and we find some evidence here for an environmental Kuznets curve. The importance of basin size suggests a possible 'resource curse' effect for water resources.