Empirical studies of post-ADA employment effects foreground a phenomenon that is puzzling. Although analyses suggest that employing workers with disabilities can be cost effective, and despite a burgeoning economy in which the unemployment rate for most categories of workers has plummeted, unemployment of working age individuals with disabilities appears not to have similarly diminished. From the point of view defined by scholars applying the neoclassical labor market paradigm to Title I, the clearest explanation of this phenomenon would seem to be that the studies reporting the cost effectiveness of employing the disabled are incorrect (even if only overstated). Following from this explication is the conclusion that selecting workers with disabilities over nondisabled workers is an inefficient practice. In what follows, I examine and assess the arguments made by proponents of the view that the inefficiency of employing workers with disabilities is a deterrent to their inclusion in the labor market. If these arguments are sound, then rational market forces appear to be inexorably at work to attenuate the strategy embodied by Title I of the ADA. To the contrary, however, I will identify a market failure that prevents certain employers from reaching rational labor market decisions by creating a "taste for discrimination" in which the costs of including people with disabilities in a workforce are perceived as being greater than they really are. Further, I will propose an improved manner for assessing the efficiency of employing workers with disabilities and consider what this method implies regarding the rationality of Title I's strategy. Finally, I will show that the failure of the existing neoclassical economic model, as well as the Title I critiques that rely on it, is attributable at least in part to societal misconceptions about people with disabilities being built into the model's assumptions. That is, far from being neutral or objective, these critiques sanction and perpetuate the very irrational biases the ADA was designed to correct.