Past research on anaphor resolution has investigated the availability of discourse information related to characters that part and reunite. Upon reunion, distant concepts that have been associated with the characters before they separated become more accessible than they were just before the reunion. This phenomenon has been referred to as the "reunion effect" and has been replicated across several studies. In an effort to explain the causes of this effect, this study argues that causal relations between sentences determine the availability of discourse information, and tests this explanation by performing a causal analysis of materials and 5 simulations of the data from 2 sets of published studies. This article then presents the results of 2 new experiments that manipulate the causal connection between the reunion sentence and prior discourse concepts. Results show that predictions based on causal relations provide a better fit for observed patterns of data than predictions based in semantic and lexical associations. The results are consistent with a model of text comprehension where causal connections to earlier information are instantiated into an ongoing representation of the discourse, and the accessibility of a concept is a function of its status in this causal model at a given point in time. (Contains 3 tables and 7 figures.)