Melz, Cheng, Holyoak, and Waldmann (1993) argue that the partial blocking of cue A that I reported (Shanks, 1991a) when subjects were presented with intermixed AB → 1, B → 1 category learning trials is not consistent with the associative Rescorla and Wagner (1972) theory analysis that I offered, given that the theory predicts complete blocking at asymptote. However, this claim assumes that subjects' were trained to asymptote in my experiments, and there is no reason to believe this was the case. Melz et al.'s further argument that there has been no reported evidence of complete blocking in associative learning tasks is incorrect. I show that, on the contrary, there is abundant evidence of it. The Rescorla and Wagner theory analysis of my results is therefore sound. The results I reported were inconsistent with contingency theories as they are normally formulated. Melz et al. propose a revised contingency theory which, they argue, can account for data from a range of learning tasks. In particular, Melz et al. claim that their theory can accommodate my results. I show that the theory can be refuted on the following grounds: (a) It is contradicted by a wealth of data from other associative learning experiments, (b) it does not in fact account for the data I obtained, and (c) for many situations in which the Rescorla and Wagner theory makes clear testable predictions, the revised contingency theory is either undefined and hence makes no predictions, or else has so many degrees of freedom that it can essentially predict any result that might be obtained.