Syntes: Karpicke & Roediger (2008) — after the first correct recall, more testing (not more study) decided recall a week later

Källa: Karpicke, J. D.; Roediger, H. L., III (2008 (Science 319(5865), 966–968)). The Critical Importance of Retrieval for Learning. https://doi.org/10.1126/science.1152408

Verified numbers (full text read 2026-10-11, authors' lab copy)

• Design: 40 Swahili–English pairs (e.g. mashua–boat); four study–test cycles. Once a pair was recalled correctly it was, by condition, kept in both study and test periods (ST), dropped from study but still tested (SNT), dropped from testing but still studied (STN), or dropped from both (SNTN).
• Learning phase: cumulative learning curves did not differ between conditions; nearly all 40 words were recalled by the end. Total trials: ST 320, SNT 236.8, STN 243.0, SNTN 154.8.
• One week later: about 80 % recalled in the conditions with repeated testing (ST, SNT) versus 36 % (STN) and 33 % (SNTN) when pairs were dropped from testing; combined, d = 4.03, and the distributions did not overlap (10–60 % vs 63–95 %). Repeated study after the first correct recall produced no measurable gain.
• Metamemory: students predicted about 50 % recall in all conditions (means 20.3–22.0 of 40 words; F < 1), so predictions were unrelated to actual recall.

Why it is in the library

Foundational for C0 and for its bound that a single correct retrieval is not enough: once an item had been recalled once, extra study did nothing for one-week retention while extra retrieval more than doubled it. Also bears on C1 (resting): being able to recall an item once, the C1 criterion of encoding, did not protect it from forgetting without further retrieval. Word pairs, no feedback beyond the study trials, university students.