In the Anglo-American legal system, legal practitioners have long called upon expert witnesses to assist them in resolving disputes. However, this alliance has been an uneasy one, with each side expressing reservations. Legal practitioners have long complained of partisanship and experts’ inability to agree whereas experts have criticized the corrupting influence of the adversarial trial process. It is this longstanding tension that motivates my consideration of groundwater models in the courtroom. I begin with a brief history of the mutual enmity between lawyers and expert witnesses. I then outline the solution articulated by the U.S. Supreme Court in the 1993 landmark case of Daubert v. Merrell Dow. Many legal commentators have judged Daubert a failure, noting the persistence of controversy surrounding expert witnesses. To explain this disappointment, I argue that Daubert is poorly suited to address one of the most vexing problems posed by scientific reasoning: the logical invalidity of inferences that bridge the gap between facts and conclusions. The problem of inference has been analyzed by a number of philosophers, but it has received little systematic attention in the context of expert witness testimony. With respect to groundwater models, both the literature and my experience suggest that other factors exacerbate the problem, including model complexity and data demands. Taking into account the general problem of inference as well as problems specific to groundwater modeling, I conclude by offering some recommendations to facilitate the use of groundwater models in the courtroom.