Two experiments were conducted to test three hypotheses related to comprehension. The hypotheses were: that actions are harder to modify than states; that implications or inferences from modified concepts would also change in memory; and that propagation of modifications would be less likely to states than to actions. The first experiment tested the direct modifiability of actions versus states and the propagation of modifications from actions to states versus modifications from states to actions. Results indicated states are initially easier to modify than actions. Propagation of modification through memory occurs, especially from states to related actions. In the second experiment, all possible pairs of actions and states were filled out. The demands on the subjects' processing were increased in several ways. Results indicated propagation of misleading information from states to related actions and, less strongly, from actions to related states. The least propagation occurred to related states. States were not typically inferred. The experiments confirmed the hypothesis that question answering is an integrated memory search update process, and representation of narratives is "action oriented." Further systematic study should be conducted to determine how questions and memory interact. This knowledge would provide educators with a more effective teaching tool. (DWH)