Two experiments were carried out to investigate the generation effect by manipulating frequency of rehearsal (once, three, or self paced), material (word vs. nonword), and task condition (generate vs. read). Experiment 1 tested the generation effect based on a within-subjects design. Experiment 2 tested it on a between-subjects design. The results of these experiments showed that the generation effect occurred only meaningful material (word), and that it was not affected by frequency of rehearsal and experimental designs. These results are discussed in terms of enhanced activation of semantic memory.