A grain of truth in the grain size effect: Retrieval practice is more effective when interspersed during learning.
Open accessThe DOI matches its Crossref record; 11 of the 11 references it cites were found in a registry or an index; 1 of the 45 sources its claims cite were checked for support.
Abstract
Retrieval practice is a powerful method for consolidating long-term learning. When learning takes place over an extended period, how should tests be scheduled to obtain the maximal benefit? In an end-test schedule, all material is studied prior to a large practice test on all studied material, whereas in an interim test schedule, learning is divided into multiple study/test cycles in which each test is smaller and only assesses material from the preceding study block. Past investigations have generally found a difference between these schedules during practice but not during a final assessment, although they may have been underpowered. Five experiments confirmed that final assessment performance was better in students taught using interim than end tests in list (Experiments 1, 2, and 5) and paired associate (Experiments 3 and 4) learning, with a meta-analysis of all available studies (k = 19) yielding a small- to medium-sized effect, g = 0.25, 95% confidence interval [0.09, 0.42]. Experiment 5 finds that the higher level of practice retrieval success in interim tests contributes to the grain size effect, but the effect is eliminated if these tests are too easy. Additional analyses also suggest that the forward testing effect, in which tests promote subsequent learning, may be a major cause of the grain size effect. The practical and theoretical implications of these demonstrations of robust grain size effects are discussed. (PsycInfo Database Record (c) 2024 APA, all rights reserved).
Overview
CoChat Summary
TL;DR
Across five experiments, interim tests interspersed during learning produced better immediate final recall than end tests, and a meta-analysis of all available studies (k = 19) yielded a small-to-medium grain size effect, g = 0.25, 95% CI [0.09, 0.42]. The effect depended on test difficulty: adding stems to practice tests eliminated it, and forward testing effects appeared to contribute.
Key findings
- Meta-analysis of k = 19 studies found a significant grain size effect in immediate tests, g = 0.25, 95% CI [0.09, 0.42], and a larger schedule effect in practice tests, g = 1.11, 95% CI [0.87, 1.36]; 15/19 experiments showed a numerical benefit of interim over end tests.
- Interim tests beat end tests in immediate recall for word lists (Experiments 1, 2, 5) and paired associates (Experiments 3, 4); e.g., Experiment 1 whole-format immediate test: F(1,342) = 27.50, p < .001, ηp² = .074; Experiment 2: t(98) = 2.62, p = .01, d = 0.52; Experiment 4: F(1,74) = 7.04, p = .010, ηp² = .087.
- Experiment 5 showed the effect is eliminated if practice tests are too easy: no-stem condition had a grain size effect, F(1,139) = 12.42, p < .001, ηp² = .082, but the stem condition did not, F(1,140) = 0.55, p = .459.
- Combined analyses indicated interim tests boosted immediate recall by over 40% overall and over 70% for List 4; delayed-test boost was approximately 30%. Experiment 2 found a robust forward testing effect: List 4 interim M = 7.71 vs. FTE control M = 5.04, t(96) = 3.97, p < .001, d = 0.80.
Methods
- Five between-subjects experiments compared interim test schedules (test after each study list) with end test schedules (test all material after all lists); Experiment 1 also included restudy and whole vs. list final-test formats, Experiment 2 included an FTE control, Experiments 3–4 used paired associates with/without feedback, and Experiment 5 crossed grain size with stem vs. no-stem practice tests.
- Participants were undergraduate students in classroom tutorials (Experiments 1, 3, 4) or Prolific online participants (Experiments 2, 5); sample sizes were 678/426 delayed (Exp 1), 149/137 delayed (Exp 2), 120 (Exp 3), 116 (Exp 4), and 283/275 delayed (Exp 5).
- Analyses used ANOVAs and t-tests with Bayes factors, free recall scored with the amatch function in R, and a meta-analysis using Hedges’ g across k = 19 effect sizes.
Limitations
- The paper states constraints on generality: Experiments 1, 3, and 4 used relatively homogenous undergraduate samples, and all experiments used laboratory word lists or word pairs; it is unclear whether results generalize to more complex real-world materials, especially since prior research with complex materials found weaker effects.
- The paper notes that relatedness conclusions rest on between-experiment comparisons, item-selection effects are difficult to rule out, feedback was included in only one experiment, and future research is needed on feedback, complex materials, and optimal difficulty.
- Evident from the text: delayed grain size effects were not significant in Experiments 2 and 5, only Experiment 1 showed persistence after one week, and the meta-analysis had significant between-study heterogeneity (immediate tests: I² = 67.6%; practice tests: I² = 80.2%).
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
1 Valid · 1 Not valid · 36 Not checked · 200 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“trial sequences that increase training performance are only beneficial for transfer of relational rules if the rule is sufficiently difficult (Don et al., 2020).”
What the sources say
- Supportedcites Don et al., 2020 source
The Trial-sequencing section reports that blocked trials improved training accuracy for both rules, but benefited transfer only for the more difficult biconditional rule; they were detrimental to transfer for the simpler patterning rule. This supports the claim as a description of blocked trial sequencing. (LLM, conf=0.93)
plainstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Experiment 5 finds that the higher level of practice retrieval success in interim tests contributes to the grain size effect”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Five experiments confirmed that final assessment performance was better in students taught using interim than end tests in list (Experiments 1, 2, and 5) and paired-associate (Experiments 3 and 4) learning”
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Five experiments confirmed that final assessment performance was better in students taught using interim than end tests in list (Experiments 1, 2, and 5) and paired-associate (Experiments 3 and 4) learning”
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“with a metaanalysis of all available studies ( k = 19) yielding a small-to-medium-sized effect, g = 0.25, 95% CI [0.09, 0.42].”
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
123 more claims
Create a free account to see all 238 claims and to verify the 1 reference still open.
- It may state two separate claims as one. 75
- The quote only partly states this claim. 2
- Probably not a substantial claim. 48
Statistics that don’t add up
Each reported p-value recomputed from its test statistic and degrees of freedom.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
“F (3,963) = 0.99, p = .393”
Reported p = .393 · recomputed p = .397
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
“t (74) = 2.65, p = .004”
Reported p = .004 · recomputed p = .010
Statistics recomputed
Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.
- Doesn’t add upt(74) = 2.65reported p = .004 · recomputed p = .010
“t (74) = 2.65, p = .004”
- Doesn’t add upF(3, 963) = 0.99reported p = .393 · recomputed p = .397
“F (3,963) = 0.99, p = .393”
- ConsistentF(3, 60) = 0.39reported p = .76 · recomputed p = .761
“F (3, 60) = 0.39, p = .76”
- ConsistentF(3, 60) = 0.2reported p = .9 · recomputed p = .896
“F (3, 60) = 0.20, p = .90”
- Consistentt(358) = 14.66reported p < .001 · recomputed p < .001
“t (358) = 14.66, p < .001”
85 more recomputed tests
Create a free account to see every test the Stats check recomputed.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
“In both formats, participants were allowed to proceed once half the allotted time had elapsed (i.e., 128s in the whole format and 34s in the list format).”
Reported 34 · recomputed 32
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Immediate test
“There was better recall in the interim ( M = 17.38, SD = 8.76) than end ( M = 14.08, SD = 8.27) test group, t (77) = 1.72, p = .045 one-tailed, BF10 = 1.58, d = 0.39. A two-tailed t -test showed no significant difference between restudy ( M = 17.34, SD = 7.97) and end test groups, t (78) = 1.80”
§ THE GRAIN SIZE EFFECT
“with 41 in the interim test group, 39 in the end test group, and 40 in the restudy group.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ THE GRAIN SIZE EFFECT
“g = 0.64, 95% CI [-0.36, 0.92].”
§ THE GRAIN SIZE EFFECT
“| Exp. 1 - Whole format | | 0.64[0.36; 0.92] | 28.2% | |”
Citations
Cites
The publisher hasn’t deposited this paper’s reference list.
Cited by
Rivas, B. A. A., Mamani, J. C. M., Lipa, R. C., & Nugent, M. A. L. M. (2025) · Sapienza International Journal of Interdisciplinary Studies · doi:10.51798/sijis.v6i4.1066 · Cited by 2
“…controlled by LED systems to minimize shadows and color variations, and adequate processing units such as industrial computers with GPUs or optimized microcontrollers (NVIDIA Jetson Nano, Raspberry Pi 4 with TPU accelerators) (Bordas, Bordas, et al., 2022; Don et al., 2024; Zerrouki et al., 2024).”
Metadata
- Authors
- Don, H. J., Boustani, S., Yang, C., & Shanks, D. R.
- Affiliations
- University College London (GB)
- Beijing Normal University (CN)
- Venue
- Journal of Experimental Psychology: Learning, Memory, and Cognition
- Published
- 2024-11
- Volume
- 50
- Issue
- 11
- Pages
- 1791-1810
- ISSN
- 0278-7393 (ISSN-L)
- 1939-1285
- OpenAlex
- W4404473653
- Cited by
- 0
- References
- 0
- FWCI
- 0.00
- Type
- journal article
- Language
- en
- Update notices
- correction (2025-01-13)
- Open access
- hybrid
- doi.org
- Licence
- Locations
- Journal of Experimental Psychology Learning Memory and Cognition (publishedVersion, cc-by)
- PubMed (publishedVersion)
- UCL Discovery (University College London) (submittedVersion)
- Funders
- United Kingdom Economic and Social Research Council (ES/S014616/1)
- Topics
- Life Sciences › Neuroscience › Cognitive Neuroscience › Memory Processes and Influences
- Social Sciences › Psychology › Experimental and Cognitive Psychology › Visual and Cognitive Learning Processes
- Social Sciences › Psychology › Developmental and Educational Psychology › Innovative Teaching and Learning Methods
- Keywords
- grain size effects, retrieval practice, test scheduling, spaced learning, forward testing effect, long-term retention, list learning, paired-associate learning, test difficulty
- MeSH terms
- Adult, Educational Measurement, Female, Humans, Learning, Male, Practice, Psychological, Mental Recall, Young Adult
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-09-29, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-09-29 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: A grain of truth in the grain size effect: Retrieval practice is more effective when interspersed during learning. https://app.cochat.ai/papers/10.1037/xlm0001382
- The paper
- Don, H. J., Boustani, S., Yang, C., & Shanks, D. R. (2024). A grain of truth in the grain size effect: Retrieval practice is more effective when interspersed during learning. Journal of Experimental Psychology: Learning, Memory, and Cognition. https://doi.org/10.1037/xlm0001382
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.