Explaining forgetting at different timescales requires a time-variant forgetting function

Employing memory models in practical applications provides an opportunity to test and refine theories on a large scale. The better models work in real-world scenarios, the more they can benefit educational practice. Using recall data spanning timescales from minutes to weeks, we find that a single forgetting curve cannot simultaneously capture short- and long-term performance, and show that an accurate model therefore requires a forgetting function that changes with time.