Retrieval practice of previously studied information seems to be more effective in the long run than restudying the information--a phenomenon called the "testing effect". In the present study, we investigated whether individual differences in the testing effect can be attributed to variation in gist trace processing. One-hundred-thirty-one participants (7-13 years old children) studied twelve DRM word lists in a within-subject design with learning (restudying vs. taking an intervening free recall test) as a factor. Each of the participants took a final yes/no recognition test 1 week after the study phase. A latent class analysis on the final-test data revealed three classes. One class of children did not show a testing effect. In the other two classes strong testing effects emerged, but the magnitude of the effect differed in these two classes. Furthermore, the three classes differed in false recognition of semantically related distractors, suggesting that the testing effect is related to differences in gist processing. We interpreted our findings in terms of fuzzy trace theory.