In rendering a decision in a particular case, judges are not limited to finding simply for the appellant or for the respondent. Rather, in many cases, they have the option to find for the former on one or more issues and for the latter on one or more other issues. By thus “splitting the difference,” judges can render a judgment that favors both litigants to some degree. What accounts for such mixed outcomes? Several theoretical perspectives provide potential explanations for this phenomenon. First, Galanter (1974) suggests that litigants with greater resources will achieve more favorable outcomes in the courts. Where two high-resource, repeat-player litigants meet in the appeals courts, these more sophisticated and successful parties may be able to persuade the court to render decisions with mixed outcomes that at least partially favor each party. Second, split outcomes may result from strategic interactions among the appeals court judges on the decisionmaking panel. Where majority opinion writers seek to accommodate other judges on the panel, split outcomes have the potential to serve as an inducement for more ideologically extreme judges to join the majority opinion. Finally, Shapiro and Stone Sweet (Stone Sweet 2000; Shapiro & Stone Sweet 2002) propose that courts will sometimes split the difference in order to enhance their legitimacy (and ultimately enhance compliance by losing parties). For example, in highly salient cases, where noncompliance would more clearly threaten court legitimacy, judges may be more likely to split the difference in order to mollify even the losing party. We develop an empirical model of mixed outcomes to test these propositions using data available from the U. S. Courts of Appeals Database and find evidence supportive of all three theoretical perspectives.