Syntes: Kornell (2025) — does memory accessibility affect how much we learn from studying?

Källa: Kornell, N. (2025). Does memory accessibility affect how much we learn from studying? Behavioral Sciences, 15(6), 760. https://doi.org/10.3390/bs15060760

Metadata

• ID: LEARN-012
• Tier: RCT (two preregistered within-participant experiments; OSF st4dx, p. 3)
• N / design: Prolific adults. Exp 1 N = 77 after exclusions (p. 4); Exp 2 N = 145 (p. 7). 48 target nouns (from Nelson et al. 1998 norms) per participant. Phase 1: pleasantness rating of each word. Phase 2: a sentence that either primed the target (it contained two associates, e.g. "The sea wave was huge" → OCEAN) or was unrelated, then either a second pleasantness rating (Exp 1, restudy) or a two-letter-stem recall attempt with 2 s feedback (Exp 2). Phase 3: final two-letter-stem recall, immediately after phase 2, same session (pp. 4–7).
• Population: adults (mean age 25 in Exp 1 and 30 in Exp 2; pp. 4, 7).
• Full text: MDPI publisher PDF, read in full 2026-09-29 (page numbers below are the article's "x of 12").

Finding

1. Exp 1 (priming before restudy): final accuracy was 0.476 (SD 0.238) primed vs 0.462 (SD 0.240) not primed. BF10 = 0.243, and 0.060 for the directional not-primed > primed hypothesis, which supports the null. Priming main effect F(1, 76) = 1.07, p = 0.30 (p. 6).
2. Exp 2 (priming before retrieval with feedback): priming roughly doubled retrieval success during practice, 0.613 (SD 0.182) primed vs 0.312 (SD 0.174), d = 1.58 (p. 8). Final test showed no effect: 0.501 vs 0.486, BF10 = 0.22, d = 0.11; directional BF10 = 0.041 (p. 9).
3. Conditional analysis (Exp 2): items recalled correctly during practice scored 0.68 on the final test vs 0.33 for items not recalled. The author attributes this to item selection, not to a causal effect of success (p. 9).
4. Conclusion: momentary accessibility, retrieval difficulty and retrieval success had no causal effect on learning. The known correlation between lower accessibility and more learning (spacing) may come from forgetting processes, which were equal across conditions here (pp. 9–10). The author concludes that retrieval difficulty per se should not be counted as a desirable difficulty (p. 10).

Limits (author and reader)

• Accessibility was measured with one test (two-letter stem), and this may not capture pure semantic activation (p. 9).
• The final test was immediate, in the same session. Other materials and longer retention intervals are untested (p. 10).
• There was no restudy-only or no-practice control in Exp 2, so the design says nothing about retrieval vs restudy. Initial encoding (phase 1) was identical across conditions.
• Word lists; online adult sample; exclusion rule corrected from a preregistration typo (p. 4).

Relevance to the graph

• C1: the primed boost is access under extra cues (related sentence + stem) at the moment of practice. It did not show up on the later matched-cue test (stem only). This is consistent with C1's criterion: later access under cue conditions that match the criterion, not in-the-moment access with added cues. It is not a test of C1's kill-condition, because there is no restudy contrast and no manipulation of encoding. Rejected as an attack on C1 on 2026-09-29 (see reports/attack-2026-09-29-kornell.md).
• C0 (queued challenge, not a verdict): retrieval success at practice, when it was manipulated this way, did not change later recall after retrieval with feedback. This bears on C0's clause "items correctly retrieved repeatedly in practice". The design has no restudy arm, a single practice trial and an immediate test. See Chan et al. 2024 (DOI 10.1037/xge0001593) for the direct test.

Swedish implication

Getting it right in practice with a hint, or with a related cue just before, is not evidence that it has been learned. Check later, with the prompt it will actually be needed for.

Links

• DOI: https://doi.org/10.3390/bs15060760
• Related: tulving-pearlstone-1966-availability-accessibility, rowland-2014-testing-effect-meta, bjork-bjork-2020-desirable-difficulties, cepeda-etal-2006-distributed-practice