In two experiments, we examined transfer to the reading of a normal text from a prior reading of that intact text or from a prior reading of a scrambled word version of the passage. In Experiment 1, we studied good and poor readers in Grade 4; in Experiment 2, high- and low-ability undergraduate readers. Good readers at both ages showed rereading benefits only when the prior reading was of the intact text, with no reliable benefit from experience with words only. The poorer readers showed reliable rereading benefits even when only the words, in a scrambled order, were read on the first encounter. The results are discussed in terms of two forms of transfer nonfluent reading transfer when attention must be focused on word recognition, and fluent rereading transfer when word recognition is skilled so that attention can be focused on text processing.