[[This is a Stage 2 Registered Report manuscript now accepted for publication at eNeuro. The accepted Stage 1 manuscript is posted here: https://psyarxiv.com/s7dft, and the pre-registration for the project is available here (https://osf.io/fqh8j, 9/11/2019). A link to the final Stage 2 manuscript will be posted after peer review and publication.]] There is fundamental debate about the nature of forgetting: some have argued that it represents the decay of the memory trace, others that the memory trace persists but becomes inaccessible due to retrieval failure. These different accounts of forgetting lead to different predictions about savings memory, the rapid re-learning of seemingly forgotten information. If forgetting is due to decay, then savings requires re-encoding and should thus involve the same mechanisms as initial learning. If forgetting is due to retrieval failure, then savings should be mechanistically distinct from encoding. In this registered report we conducted a pre-registered and rigorous test between these accounts of forgetting. Specifically, we used microarray to characterize the transcriptional correlates of a new memory (1 day after training), a forgotten memory (8 days after training), and a savings memory (8 days after training but with a reminder on day 7 to evoke a long-term savings memory) for sensitization in Aplysia californica (n = 8 samples/group). We found that the re-activation of sensitization during savings does not involve a substantial transcriptional response. Thus, savings is transcriptionally distinct relative to a newer (1-day old) memory, with no co-regulated transcripts, negligible similarity in regulation-ranked ordering of transcripts, and a negligible correlation in training-induced changes in gene expression (r = .04 95% CI [-.12, .20]). Overall, our results suggest that forgetting of sensitization memory represents retrieval failure.