Abstract
If memories of similar things are stored in the brain at about the same location or using a similar group of neurons and with similar neural networks, then a newly acquired memory will interfere with or even erase an old similar memory and lead to forgetting. We designed two experiments to test this hypothesis: a verbal experiment (for declarative semantic memory) and a gymnastic experiment (for nondeclarative procedural-related memory). In the verbal experiment, the participants first memorized two poem stanzas, then memorized two new stanzas that are very similar to the first two, and finally, took a test on the stanzas they memorized first. In the gymnastic experiment, the participants first learned a set of gymnastics, then learned another (very similar) set, and finally, took a test on the gymnastics set they learned first. The results showed that the participants who memorized two new poem stanzas made more mistakes than did the control participants in the last test, and the participants who learned a new set of gymnastics performed the first set much poorly compared with the control participants in the last test. We included an additional group in each experiment. The participants in this group were required to review the previously learned poem stanzas or gymnastics when they were learning the new ones (to compare the new ones with the old ones). The results showed that this group performed far better in the last test (on what they learned first). In summary, our experiments showed that newly acquired memories significantly interfered with or even erased similar old (semantic and procedural long-term) memories; nevertheless, if new things were learned in comparison with old ones, the interference could be noticeably reduced. Our study provided support for similarity-based retroactive interference in long-term semantic and procedural memories and offered ways to reduce such interference.