We explore the implications of judicial learning for trade disputes through a model where both the initiation of disputes and the occurrence of rulings are endogenous, governments bargain “in the shadow of the law,” and the efficiency of the court increases with experience. Judicial learning can explain litigation on the equilibrium path, since going to court today implies future payoff gains for the governments. Our model predicts that where learning is present the likelihood of both disputes and rulings should tend to decrease with court experience. Using detailed data on WTO disputes, we find evidence consistent with significant judicial learning at the WTO, but this learning appears to be article-specific and disputant-specific, rather than general, in scope.