In 3 experiments, participants read pairs of sentences requiring a bridging inference between a category in 1 sentence and a typical or atypical exemplar of that category in the following sentence. In Experiments 2 and 3, the critical pair of sentences was preceded by several lines of text in which the exemplar was introduced and then backgrounded by a shift in topic. Accessibility of the earlier mentioned exemplar was manipulated by varying the distance between the first and the subsequent mention of the exemplar, or by varying the elaboration of the exemplar when first introduced into the text. Eye fixations in Experiment 1 revealed that the bridging inference was affected by typicality, that the process began at the earliest opportunity but was not completed when the eye left the exemplar word. In Experiments 2 and 3, both typicality and episodic accessibility influenced reading time. These effects are discussed in terms of several models of the underlying process.