In previous research, rereading after a long lag versus a short lag led to greater performance on delayed tests but not on immediate tests. The current study tested two accounts of why the effects of rereading lag depend on test delay. The "levels of representation" ("LOR") "hypothesis" states that the effects reflect differential emphasis on encoding of higher-level versus lower-level text information during long-lag versus short-lag rereading. The "integration hypothesis" states that the effect reflects a greater degree of integration of encoded information during long-lag versus short-lag rereading. Across two experiments, 341 students read a text either once, twice with a short lag, or twice with a long lag. Final free recall and cued recall tests were administered either immediately or 2 days after study. Results disconfirmed the LOR hypothesis (neither the decreases in reading times across trials nor the changes in free recall across timepoints differed as a function of information level in either rereading group) and confirmed the integration hypothesis (performance on cued recall measures of within-sentence and between-sentence integration was greater with long-lag versus short-lag rereading). (Contains 1 table and 8 figures.)