Prediction is arguably the most fundamental problem that people face. Having discovered that some object possesses a particular feature, how is it that people are able to accurately infer that another object exhibits the property? Psychologists have actively studied this reasoning process; yet, current models of induction cannot provide an explanation for the entirety of the related phenomena. One reason may be that current models fail to account for people's ability to assess multiple categories when making an inference. Building on previous research (Shafto et al., 2006), we present a model of inductive reasoning based on cross-cutting knowledge representation. We present an experiment that investigates the ability of this model to account for known inductive phenomena. We show that a model which assesses multiple kinds of knowledge explains the flexibility of human inference, better than models relying on a single kind of knowledge.