Even if we don’t like it, we often face counterexamples to the inferences we have made or would like to make. With the exception of the SimProb model (Blok, Medin & Osherson, 2007), models of inductions to date have predominantly focused on the relevance of positive evidence to the inference process. Here we provide data from single and double premise arguments in a category-based property induction task using positive and negative evidence. A simple similarity model, the Similarity-Coverage model (Osherson et al., 1990) and the SimProb model are tested on negative and mixed evidence arguments.