Research was conducted to investigate the effects of contextual information on the speed and accuracy with which two general classes of inferences were verified by readers. These types of inferences were based on information in conversations that were or were not topically ambiguous, depending upon the amount of available contextual information. Subjects were 68 undergraduate students who listened to eight dialogues, summarized the content of each dialogue, and responded to test statements about information from each. When the subjects' response latencies were measured, the results indicated that, for inferences/responses related to the contextual information, those that were compatible with the context took less time to verify and were rated truer than those that were not compatible with the context. An explanation based on script comprehension was proposed for these results. For those inferences/responses not related to the context, both true and false types based on implied information were rated closer to "indeterminate" in truth values and took longer to verify than those based on explicit information. This finding was consistent with results from sentence-memory and prose-processing literature. (RL)