This research presents a computational model that simulates inference generation during reading comprehension. Inferences refer to information that readers generate from their background knowledge in order to clarify, connect, and elaborate textual information. The computational model integrates Latent Semantic Analysis (LSA), which simulates general knowledge by computing the strength of semantic association between concepts (Landauer & Dumais, 1997), with the Landscape Model, a dynamic model of reading comprehension that simulates fluctuations of concepts' activation and the emergence of episodic connections between them (Yeari and van den Broek, 2011). The extended model was used to simulate behavioral data from a large number of studies on inference generation. Successful simulations of the various findings demonstrate the unique roles of semantic associations, episodic inter-textual relations, and working memory (limitation of concepts' activation sum) in the activation of different types of inferences (i.e., elaborative and bridging inferences).