Abstract: This article questions, and in some contexts disproves, the validity of the efficiency justifications for the comparative negligence rule. One argument in the literature suggests that comparative negligence is the superior rule in the presence of court errors. The analysis here shows the analytical flaw in this claim and conducts numerical simulations -- a form of synthetic "empirical" tests -- that prove the potential superiority of other rules. The second argument in the literature in favor of the comparative negligence rule is based on its alleged superior ability to deal with private information. This article develops a general approach to liability rules as mechanisms that induce self-selection among actors. It then shows that self-selection can occur, not only under comparative negligence, but also under every other negligence rule. These conclusions weaken the efficiency explanation for the growing appeal of the "division-of-liability" principle within tort law and beyond. Copyright 2003, Oxford University Press.