Roediger and Karpicke's 2006 experiment is one of the most cited findings in learning science. Students who studied a passage once and then took a test remembered 61% of it a week later. Students who studied the passage twice remembered 40%. The test — not the extra study session — was what made memory durable.
This finding has been replicated across ages, subjects, and formats. It is one of the strongest interventions in the entire learning-science literature.
Yet consumer study products often optimize for the opposite: comfortable re-reading, video re-watching, highlight collecting. These interventions feel productive but produce transient memory. Users who spend hours in such products often report feeling less prepared, not more.
Brainback bakes retrieval practice into the daily loop. The recall queue is a series of forced-generation tests, not re-reading opportunities. Explain-back is a forced-production test. The attempt-gate turns every problem into a test-before-you-teach-me event.
The design choice is not novel. The testing effect has been known for decades. The novelty is a consumer AI product that actually implements it.
Written by Brainback Research. Published Mar 9, 2026. Filed under Learning science.