Memory failure predicts belief regression after the correction of misinformation

After misinformation has been corrected, people initially update their belief extremely well. However, this change is rarely sustained over time, with belief returning towards pre-correction levels. This is called belief regression. The current study aimed to examine the association between memory for the correction and belief regression, and whether corrected misinformation suffers from belief regression more than affirmed facts. Participants from Prolific Academic (N = 612) rated the veracity of 16 misinformation and 16 factual items and were randomly assigned to a correction condition or test-retest control. Immediately after misinformation was corrected and facts affirmed, participants re-rated their belief and were asked whether they could remember the items' presented veracity. Participants repeated this post-test one month later. We found that belief and memory were highly associated, both immediately (⍴ = 0.51), and after one month (⍴ = 0.82), and that memory explained 66% of the variance in belief regression after correcting for measurement reliability. We found the rate of dissenting (accurately remembering that misinformation was presented as false but still believing it) remained stable between the immediate and delayed post-test, while the rate of forgetting quadrupled. After one month, 57% of participants who believed in the misinformation thought that the items were presented to them as true. Belief regression was more pronounced for misinformation than facts, but this was greatly attenuated once pre-test belief was equated. Together, these results clearly indicate that memory plays a fundamental role in belief regression, and that repeated corrections could be an effective method to counteract this phenomenon.