Corrected misinformation keeps shaping inference after the correction has been accepted.
Johnson and Seifert's finding: told a fire was caused by volatile chemicals and then told the report was mistaken, subjects still reason about the fire as a chemical fire. The retraction is accepted and the inference survives.
In practice
- A recall notice is accepted and the product still feels unsafe two years later, because the inference outlived the retraction.
- A candidate cleared of an allegation is not selected, and no one on the panel would say the allegation was the reason.