Some literary scholars have claimed that predictive models can measure the strength of the boundaries that separate different cultural categories—genres, for instance, or market segments. But interpreting textual models as evidence about the strength of a cultural distinction has seemed a questionable move to many readers. We use book reviews to test this inference. Are the similarities between fictional texts purely verbal phenomena, or do they reflect social categories that are also legible (although expressed differently) in readers’ responses to those texts? We find that the subject and genre categories most strongly marked in fictional texts are also the categories most strongly marked in reviews of fiction. The correlation is strong; r > .8.