Young university students and elderly community-dwelling volunteers read three short narratives word by word on a video display terminal, pacing the text by pressing a key. With this 'moving window' method, reading times for each word were measured. Subjects recalled each narrative immediately after reading. Regression analyses were used to predict word-by-word reading times from text features representing different levels of discourse processing (e.g., word length and frequency, serial position, new concepts introduced, and the presence of syntactic boundaries). The effects of these features on reading time were examined not only for young and old as a whole, but also for good and poor recallers within each age group. For many of these features, younger and older readers responded similarly to the demands of the text in allocating reading time. Interesting differences among these different kinds of readers revolved around how they allocated time to encode and organize new concepts. Overall, older readers allocated less time to process new concepts at all levels of discourse. Most notably, they were less likely to spend extra time at the ends of sentences, a site a which younger readers reliably show evidence of organizational processing. Interestingly, relatively good recallers within the older group did not show any evidence of this 'wrap-up' processing. Rather, older adults who showed high levels of text recall allocated more time to process new concepts at minor syntactic boundaries (e.g., prepositional phrases), suggesting that they pause more frequently to organize new information during reading. Copyright © 1995 Taylor and Francis.