We offer an account of reading comprehension that we believe will help clarify some common conceptual confusions in the relevant literature, as well as contribute to existing functional accounts. We argue that defining texts qua texts as stimulus classes, on the one hand, and equating "comprehension" with behavior (covert or otherwise), on the other, are not useful conceptual moves, especially when behavioral settings go beyond basic literacy skills acquisition. We then analyze the structure of the contingencies that usually evoke talk of "comprehension" using techniques from analytic philosophy. We show how keeping the results of this analysis in mind can help avoid the conceptual bafflement that often arises, even among behavior analysts, when defining or assessing behavioral phenomena related to reading comprehension. Using two contrasting cases (legal texts and stories), we argue that what counts as comprehension depends, not peripherally but crucially, on the shared social practices of which texts are a part. Finally, we propose a new framework for classifying reader-text contingencies by combining two dimensions: openness of setting and embeddedness of reinforcement.