Learn something once and it drains away on a curve — fast at first, then slower. Ebbinghaus measured it on himself in 1885. The fix isn’t cramming harder; it’s reviewing at the right moments, because each review makes the NEXT forgetting slower. Slide the reviews and watch memory learn to last.
Retention decays as R(t) = e^(−t/S), where S is the memory’s stability. Each review resets R to 1 AND multiplies S (the next decay is gentler) — and the gain is larger when you review just as the memory is fading. The instrument runs this model live. A fail-loud self-check throws unless R(0)=1, R is strictly decreasing in time, and a larger S retains more at a fixed later time — the whole case for spaced repetition, not cramming.
A single-trace exponential is a simplified model (real memory has multiple time-constants; SM-2/FSRS schedulers add more state); the qualitative laws — exponential decay, review-boosts-stability — are the robust, reproduced findings.