One explanation of the correlation often observed between working-memory span scores and reading comprehension is that individuals differ in level of activation available for long-term memory units. Two experiments used the fan manipulation to test this idea. In Experiment 1, high- and low-working-memory Ss learned a set of unrelated sentences varying in the number of shared concepts (fan) and then performed speeded recognition for those sentences. Low-working-memory Ss showed a larger increase in recognition time as fan increased. When the slope of the fan effect was partialed out of the relationship between working-memory span and verbal abilities, the relationship was reduced to nonsignificance. In Experiment 2, Ss learned thematically related sentences that varied in fan. Low-span Ss showed the positive fan effect typically found with thematically unrelated sentences, whereas high-span Ss showed a negative fan effect. The results are discussed in terms of a general capacity theory.