Abstract: Using the Deese-Roediger-McDermott (DRM) paradigm, two experiments were conducted to investigate the forgetting trends of false memory after short time delay. In within-subject design, false-alarm rates of both critical lures and medium related lures increased with time (immediate, 1/2 hour, and one hour). However, in between-subject design, false-alarm rates of critical lures didn’t change with time. The results reveal that false memory could be steady at least if only forgetting takes place, but variable if some other disturbing processes like repeating false recognition in within-subject experiment are mixed with forgetting. Furthermore, the effect of test delay was combined with the effect of association and the test context. The results are discussed under source-monitoring framework and activation-spread theory

Key words: false memory, DRM paradigm, forgetting, association