The Testing Effect: Active Recall vs. Re-reading

Study Schedule

Review interval 4 days

Autoplay speed 1.5 days/s

Live Readouts

Re-reading group Active-recall group

Simulated day

0.0

Testing-effect gap

+0 pts

Re-read retention

100%

Recall retention

100%

How it works

Ten matched facts are assigned to each group, all starting with identical memory strength S (an Ebbinghaus-style forgetting-curve parameter). At any moment, retention decays as:

R(t) = e^(-(t - L) / S)

where t is the simulated day, L is the day of the last review, and S is that item's current memory strength in days. Both groups are reviewed on the exact same schedule (the interval slider), so the schedule itself is never the variable — only how each group reviews:

Re-reading group — every session, each item is simply re-read. Retention resets to 100% and S grows by a modest fixed multiplier (illustrative: ×1.25 per session), the typical effect of passive restudy.
Active-recall group — every session, each item is tested first. The probability of a correct answer equals its current retention R(t) at that moment (a probabilistic quiz). A successful retrieval resets retention to 100% and grows S by a much larger multiplier (illustrative: ×2.6); a failed attempt still reveals the answer but only grows S by a small amount (×1.15) — close to the re-reading case.

Because successful retrievals compound faster than passive rereading, the recall group's average S — and therefore its long-run retention — pulls steadily ahead of the re-reading group even though both study for the same number of sessions. This is the testing effect (Roediger & Karpicke, 2006): actively retrieving information from memory strengthens the memory trace far more than re-exposure to it, which is why self-testing and practice-recall (flashcards, closed-book summarizing, past-paper questions) consistently outperform highlighting and re-reading as study techniques. The multipliers here are stylised for clarity, not a precise empirical fit — the qualitative divergence they produce is the real, well-replicated finding.

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.