Two experiments examined how cognitive resources are allocated to comprehension processes across two readings of the same scientific texts. In Experiment 1, readers read and later reread texts describing scientific topics. The results indicated that across readings, readers decreased resources allocated to proposition assembly, increased resources allocated to text-level integration, and expended a similar amount of resources to lexical access. Subjects who reread the texts after a week delay showed a similar pattern, except that they did not show the increase for text-level integration. Experiment 2 revealed a similar pattern of results with a moving window procedure, except that there was a significant decrease in resources allocated to lexical access across exposures. This experiment also indicated that the rereading speedup was greatest at sentence boundaries, suggesting that the prior exposure enabled readers to immediately process each word. Overall, the results are consistent with the claim that readers allocate proportionally more available resources to text-level integration during rereading because proposition assembly, which enables text-level integration, can be completed with fewer resources.