Retrieval practice in the classroom

1. Vignette: the weekly test nobody used for learning

The vignette documents neither a school nor a figure. Whoever teaches in this way confuses a summative examination without return of the work with retrieval practice: attempting to bring back what has already been encountered, with correction and with repetitions set apart in time. Grading without review pushes students to reread the night before, the strategy with which retrieval is usually compared. Feeling that the text is familiar is not producing it from a blank page.

The student’s block is not clinical evidence. If the benefit depends on working memory, on encoding and on format, a single high-stakes regime is not neutral. To say that retrieval is already being done because there is a quiz describes the surface. What follows separates the effect, its translation into the classroom, the moderators and adoption.

2. Problem: a robust effect is not a school practice

Trying to recall usually retains better than restudying: hence the reputation of retrieval practice, also called the testing effect or test-enhanced learning. From that reputation it does not follow that any test captures the effect. A laboratory contrast does not say who designs the week, what is done with the error, or what happens if the demand is not sustained. The gap is the distance between a memory phenomenon and a teaching regime.

The question has a population, a concept and a context. Population: students from primary school through university, and teachers who decide whether retrieval enters the week. Concept: not the isolated examination, but a regime of attempts, with or without spacing, interleaving, relearning to a criterion and feedback. Context: real classrooms and, as a contrast, settings in which the effect is sought and does not appear. When does that regime improve learning, and which moderators of knowledge, of profile and of adoption limit it?

Wilschut et al. (2026) diagnose the field. In a perspective on 23850 publications they see a turn toward educational applications and little attention to learning difficulties. They use dyslexia as a theoretical case against universal generalization and call for research with neurodiverse learners. It is surveillance of the field, not an effect size, and it does not include a Mexican classroom. To say that retrieval “works” remains, on that account, incomplete.

The Friday examination falls into four gaps. There is no regime: one summative moment, without return of the work and without a later return to the content. There is no type of knowledge: a correct answer is taken for knowing how to use what was taught. There is no profile: the same sheet and the same clock for any working memory. There is no considered adoption: the name circulates without a habit that would replace rereading. Closing those gaps does not require more examinations. It requires saying what counts as practice.

3. Framework: direct effect, forward effect and desirable difficulty

This section is a conceptual framework. It fixes three distinctions: the direct effect as against the forward effect, desirable difficulty as a cost that sometimes pays for itself, and successive relearning as against the single session. Without them, setting a test looks like a single decision.

The direct effect improves memory for what was tested, relative to restudy. The forward effect is different: a practice test improves what is learned afterwards. Gereau and Mulligan (2026), in a meta-analysis of free recall, show that the literature calls “the” testing effect direct and nevertheless conflates the two. An empirical finding about that literature, not about a classroom: 66% of the effects come from mixed studies and 34% from studies that isolate the direct effect; the mixed studies produce larger effects, and the forward component sometimes raises the outcome, sometimes lowers it and sometimes leaves it halfway. It is not a single phenomenon, and it is not confused here with transfer to a new problem.

Desirable difficulty is anchored in Lyle et al. (2022). In a calculus course, spacing the problems increased end-of-semester retention and reduced performance during practice. The authors read this as slower learning and, in the end, more solid learning: spacing imposes a desirable form of difficulty. An immediate worsening does not refute the practice; it may be the cost of delayed retention. Desirable does not mean cost-free. Section 6 specifies the design — how many problems, and on which questions the cost fell — and Section 8, when that cost ceases to be paid.

Successive relearning, in Rawson and Dunlosky (2022), is practicing until a correct response and returning to that criterion in spaced sessions. Outside school it is common; few students use it to maintain curricular knowledge. The agenda weighs as much as the definition: to leave the single session of fixed duration and to treat time as an outcome. Set against the examination in the vignette, which neither fixes a criterion nor reserves a return, that unit is the framework. It is not yet a prescription.

4. Method: narrative review of classroom studies and moderators

The method is a critical narrative review, not a meta-analysis and not a systematic review of effectiveness. Effect sizes were not pooled, and twenty-four works are not claimed to be a random sample. Selection was manual, from a single index. The empirical sections are not a ranking. They can show patterns, nuances and silences in a recent corpus assembled under explicit rules.

The search used the Crossref REST API and was logged on 19 October 2026: 16 queries of up to 60 records, journal articles from 2021 whose title joins a retrieval or testing term with an educational term, excluding uses unrelated to school learning and the clinical literature. Table 1 fixes the rest: 171 unique records, 24 included, 19 with an abstract and 5 by title only.

Table 1. Flow of the narrative review. Figures from the selection record, not effects.

Phase	Criterion	Record
Search	16 queries to Crossref, cap of 60 records	Search window, not a census
Filters	Journal article from 2021; title with retrieval or testing and an educational term; excluded, uses unrelated to learning and the clinical literature	171 unique records
Inclusion	Manual selection: classroom, moderators or adoption	24 works
Usable abstract	Crossref or OpenAlex	19
Title only	No abstract; the citation is limited to the object of the title	5

Left without an abstract are Sumeracki and Castillo (2022), Carpenter (2023), Pan et al. (2024), Lehman (2025) and Weissgerber and Rummer (2023). They are cited for the object of the title — covert and overt retrieval, interventions to encourage use, future directions, optional review and anxiety, metacognitive beliefs — and no designs or magnitudes are attributed to them. If the title states a result, it is presented as a statement, not as a figure taken from an abstract. For the other nineteen, contrasts, delays and, only when they are on record, samples or effects are reported.

Three labels organize the reading. Empirical finding: a result described in an abstract. Conceptual framework: a distinction that organizes, not a recipe. Design inference: a consequence for the school, above all a Mexican or Latin American one, which the corpus does not study. There are no sources beyond the twenty-four. What precedes 2021 is left out; this review does not re-estimate the classic effect.

5. From laboratory to classroom: what holds and what fades

Wilschut et al. (2026) describe the move from laboratory to classroom as a trend in the field, not as a guarantee that the contrast will survive. Here the same name has three destinations: a primary classroom in which retrieval is compared with a real practice; an English primary school in which the technique is already policy and is confused with assessment; and an online setting of weak participation, in which the effect does not appear. There is no room for the single sentence “it works in the classroom.”

It holds, with course materials, in Ophuis-Cox et al. (2023). They compared retrieval cards with singing the multiplication tables, in an authentic primary school, with 48 second-grade pupils: one lesson and three spaced sessions per strategy, counterbalanced across two sets, a pretest, a posttest at five minutes and a posttest at one week. Empirical finding: retrieval increased fluency more, in the short term and in the long term. It is not a weekly examination, nor a reported improvement in the global mathematics grade. It is a small regime.

It does not hold in Sigayret et al. (2026), and that study is not a school study. Two preliminary experiments on Prolific, with validated materials, delayed posttests, feedback and factual and application measures, found no significant differences between retrieving and restudying, despite filters, attention checks and fair compensation. The authors read a limit of the setting: sustained engagement suffers in open-collaboration samples. An online quiz does not reproduce the laboratory merely by being online.

Bates (2025) shows a dilution of meaning. In a primary school in England, where policy has consolidated the technique, a case study combined questionnaires, observation and interviews. Qualitative empirical finding: reported use and observed use align, but the teachers confused retrieval with assessment for learning, although the research frames it as an occasion for learning and not as assessment. There is no effect size and no count of teachers. The sentence from the staff room is, then, a described pattern: one can regularly do something called retrieval and still be assessing.

Sumeracki and Castillo (2022) add no results: there is no abstract. The title announces covert and overt retrieval practice in the classroom. Superiority is not attributed to either modality. Only the fork that the visible examination does not exhaust is taken up: part of the attempt may not be public, and choosing that is deciding the regime. Taken together, the translation into the classroom holds for the spaced tables, is lost if engagement is thin and changes meaning if the name is used in order to assess.

6. Designing practice: spacing, interleaving and successive relearning

The object of design is not the loose item. It is the regime: how many times one returns, whether concepts are mixed, whether there is a criterion of correctness and correction, and whether the attempt carries a grade. No study in this section is a Mexican secondary school. Several are real courses.

Sana and Yan (2022) interleaved retrieval in science. Over four weeks, students in grades 9 to 12 (N = 155) answered a weekly quiz on half of the concepts, blocked or interleaved. At one month, what had been quizzed in blocks (M = 54%, SD = 28%) exceeded what had not been quizzed (M = 47%, SD = 20%; d = 0.30), and what had been interleaved (M = 63%, SD = 26%) exceeded the blocked material (d = 0.35). An empirical finding of low cost, with wide dispersion. Interleaving does not remove the need to have taught the material.

Lyle et al. (2022) compared three problems on a single quiz with the same three spread across three quizzes. Spacing improved end-of-semester retention and worsened practice only on the first two questions; the third did not change. Whoever looks only at that afternoon will say that spacing failed before the cost becomes retention. The abstract gives neither the number of students nor the magnitude of the final gain.

Higham et al. (2022) applied the successive relearning of Rawson and Dunlosky (2022) in introductory psychology. After each class, three sessions separated by two days covered the remaining ten weeks, alternating relearning and restudy. Twenty completion questions carried feedback: what was correct dropped out and what was missed returned up to twice; restudy reread the same sentences twice, with exposure equated. At the close, recall was better after relearning. That in-session recall was associated with better metacognition, greater self-reported mastery, greater attentional control and less anxiety, and the individual differences examined were not associated with the advantage. Time was equated; Rawson and Dunlosky (2022) would also want it as each learner’s own variable.

Lehman (2025) enters only by the title: optional review with retrieval practice improves learning and reduces test anxiety. There is no sample and no effect size, and the statement is not equated with Higham et al. (2022). It does place low stakes among the objects of the recent literature. The Friday of the vignette is not optional.

Table 2. Components of the regime. A reading scheme, not a ranking of effects.

Component	What it adds	Anchor	What it does not authorize
Spacing	Separates attempts; it may lower practice performance and raise delayed retention	Lyle et al. (2022)	The same cost outside that course
Interleaving	Mixes concepts instead of blocking them	Sana and Yan (2022)	Replacing the initial teaching
Successive relearning	Until correct, and again, with correction	Rawson and Dunlosky (2022); Higham et al. (2022)	Treating time as known
Feedback	Returns the error within the session	Higham et al. (2022); Cairney et al. (2026), adults	Turning a summative examination into a formative one
Low stakes	Separates the attempt from the grade	Lehman (2025), title only	A figure for anxiety

Against the vignette there is a test and a grade, and return, interleaving, a criterion, correction and low stakes are missing.

7. Beyond recall: transfer and procedural knowledge

The reputation of the effect rests above all on facts. School asks, in addition, for rules that travel to an unseen item. The corpus answers with a dissociation: retaining better is not applying better, and transfer, when it appears, depends on how much retrieval there was and on the delay.

Opitz and Kubik (2024) used an artificial language, not a school subject. One group restudied correct sentences; the other did completion. There was a grammaticality judgment at five minutes and at one week, and at one week a memory test. Empirical finding: a benefit for long-term memory and for far transfer after the week, which the authors attribute to the rule and not to the exemplar. They do not report the number of participants. It is a possible mechanism, not a grammar lesson.

Corral and Carpenter (2025) use concepts from research methods. In Experiment 1 (N = 309), at eight minutes, retrieval exceeded restudy — not the study of the questions together with their answers, and not recognition — on repeated items, and not on application. With three rounds and one week, Experiment 2 (N = 158) gained on the repeated material relative to restudy and to study of the test, and Experiment 3 (N = 255) gained on application. The conditional analysis, application given a correct answer on the repeated material, points to recognition within transfer. Sufficient practice and a delay of days are required, and studying the question together with the answer is a serious rival.

Ophuis-Cox et al. (2026) compared worked examples alone with retrieval plus examples, in verb spelling. The participants were 105 fourth-grade pupils, seven classes and three schools in the Netherlands (M age = 9.4 years, SD = 0.67). At one week, the combination had a significant effect on retention of the rules and did not differ significantly on application. A mixed empirical finding: one can remember a procedural rule better and not use it better.

Gereau and Mulligan (2026) prevent three extensions from being fused: remembering what was asked, learning better what is studied afterwards — the forward effect, often confused with the direct effect — and using a rule in a new case. The third has support at one week in an artificial language (Opitz and Kubik, 2024) and, with several rounds, in complex concepts (Corral and Carpenter, 2025). It does not automatically accompany verb spelling.

8. Who benefits: working memory, ADHD and special educational needs

Wilschut et al. (2026) left the debt toward those who are not in the typical sample. This section brings together working memory, attention-deficit/hyperactivity disorder, special education and, as a population boundary, autistic adults. The question is not whether a diagnosis is served by the effect, but under what demand, encoding, format and correction, and what gap remains even when the benefit appears.

Zheng et al. (2023) state that working memory matters only if the demand challenges capacity. After pretraining 30 undergraduates on visual search, they contrasted retrieval and restudy with familiar or unfamiliar stimuli. With higher capacity there was a consistent effect; with lower capacity, a benefit only for familiar stimuli. A dual-process model suggested that the effect appears when capacity is sufficient for both processes. The cost in Lyle et al. (2022) may be unaffordable if that capacity does not cover the demand.

Minear et al. (2023) compared 36 people with attention-deficit/hyperactivity disorder and 36 controls, with Swahili–English pairs, half restudied and half tested. All showed a benefit of retrieval and status did not moderate it. Unmedicated participants performed worse at encoding and on the final test, with fewer deep strategies; medicated participants did not differ significantly from the controls. Retrieval does not compensate for poor encoding, and the contrast is not a pharmacological indication. Carrying it into secondary school is an inference.

Stern and Halamish (2023) compare free recall by sections or of the whole text, with and without the disorder, and a test at two days. During practice there were more units under recall by sections; on the criterion the whole text won, and no retrieval condition exceeded restudy, in both groups. The whole text preserved order better, and recall with the disorder was less ordered. If retrieval is used, they recommend the whole together with mental organization. What feels productive may be worse two days later.

Tempel and Sollich (2023) worked in a special-needs school — learning and mental development — and in a regular group. Within subjects, one session was restudy only or cycles of test and restudy, and the next reversed the form; the test followed a brief distractor. With pictures and with picture–word pairs there was a testing effect. Lower capacity, or little experience with examinations, did not prevent the benefit. The delay is short and the material is not a curriculum: the datum does not authorize the high-stakes examination of the vignette.

Cairney et al. (2026) are a population boundary: autistic and non-autistic adults in the United States, not school pupils. Experiment 1 was auditory (40 per group). Experiment 2 was visual, with restudy, between-subjects feedback (80 per group) and four-alternative recognition during practice; the test was cued recall. Subsequent feedback benefited all groups in both experiments, and in the second, retrieval exceeded restudy even without it. Correction enters the regime; autism among school pupils is not represented, and the dyslexia to which Wilschut et al. (2026) point remains without a trial.

9. Adoption: beliefs, teacher use and student self-regulation

A regime that the articles describe and that schools do not sustain is not yet a practice. The teacher decides whether Friday holds an examination or a session with correction. The student decides, without supervision, whether to self-test or to reread. Those decisions are not settled by demonstrating that the effect exists.

Bates and Shea (2024) describe the use that teachers in England report outside a controlled study. A finding about use, not about learning: it is employed regularly even when it is not a school requirement; quizzes and short answers predominate; motivation goes beyond the direct benefit of the testing effect. There is no count of teachers. Together with Bates (2025), that regularity does not prove that the regime is in place: one can ask questions often and still be assessing.

Carpenter (2023) announces in the title five types of interventions so that students will use retrieval. Without an abstract they are not enumerated. If explaining the effect were enough, that literature would not be needed: adoption is behavior, not a corollary of memory. Weissgerber and Rummer (2023), by the title alone, maintain that metacognitive beliefs about the effectiveness of practicing retrieval are more accurate than has been assumed. There is no margin and no instrument. It is not claimed here that students avoid retrieval because of a crude belief, nor is the contrary accuracy quantified.

Broeren et al. (2021) instructed one group (n = 58) about the strategy and another (n = 58) with a neutral prompt, using concepts from marketing communication and an online environment, in sessions 1 and 2. There were no significant differences while the instruction was in place. In session 3, without instruction, the experimental group self-tested more and made more attempts, including correct ones, per concept, and there was no correlation with performance. The habit can shift when the scaffold is withdrawn without the grade accompanying it.

Pan et al. (2024) title emerging and future directions in test-enhanced learning. Without an abstract, those directions are neither listed nor attributed to this page. The title only authorizes treating the field as unfinished. Regime, type of knowledge, profile and adoption fit inside that unfinished state; they are not a summary of what Pan and colleagues proposed.

10. Discussion: from technique to practice regime

The reputation of retrieval practice is not enough to found a school practice. In free recall, most of the effects mix two phenomena. In real courses the benefit requires a regime and splits in the face of application, thin engagement or a demand that working memory does not cover. It appears, moreover, without erasing encoding or format. It is not adopted because the effect exists.

Figure 1 orders that thesis as a chain, not as a datum. Nothing in it is a mean or a Mexican school. The weekly examination and the regime share the gesture of asking and not the architecture.

Link	Toward durable learning	Weekly graded test	Retrieval regime
Laboratory effect	Retrieving may retain better than restudying; it is not a single phenomenon	The name justifies the sheet	It is a starting point, not the practice accomplished
Regime	Spacing, interleaving, feedback, low stakes	One moment, without return of the work or a later return	Distributed and corrected attempts, apart from the grade
Knowledge	Fact, rule or new case	A correct answer is taken for knowing how to use it	Retention and application are measured separately
Learner	Working memory, encoding, format	The same sheet and the same clock	Demand and format are adjusted
Adoption	Teacher use, beliefs, self-regulation	There is a test, therefore “it is already being done”	The habit is designed and is distinguished from assessing
Durable	Recall and, if the regime is sufficient, delayed use	It is not the aim of the instrument	It is the criterion; the grade does not replace it

Two tensions do not fit into a single directive. Lyle et al. (2022) show that worse performance when spacing can precede better end-of-semester retention; Zheng et al. (2023), that with less working memory the benefit is absent if demand is high. Ophuis-Cox et al. (2026) improve recall of rules without improving their application, and Corral and Carpenter (2025) see an application advantage only with several rounds and days in between.

The six implications that follow are design inferences for Mexican and Latin American schools. They are not findings of this corpus: there is no Mexican classroom in it. They are formulated because the graded Friday and the diversity of the group make the vignette recognizable.

1. Separate learning and grading. The design inference is to use brief attempts that are corrected and not summative, and to leave the grade for another moment. A number without return of the work is not feedback.
2. Warn that the first spaced attempt may come out worse, as Lyle et al. (2022) show, and mix concepts, as in Sana and Yan (2022). Without that warning, the cost is read as failure.
3. Do not replace application with recall of the rule. Ophuis-Cox et al. (2026) and Corral and Carpenter (2025) require that the rule be produced and used on an unseen item.
4. Adjust the demand before saying that it does not work for a group. Zheng et al. (2023), Minear et al. (2023) and Stern and Halamish (2023) change the result according to capacity, encoding and format. Freezing on a high-stakes examination does not refute the practice, and the medication datum is not clinical guidance.
5. Train in the regime, not in the name. Bates (2025) shows that what is said and what is done can align and still be confused with assessing. Broeren et al. (2021) increase the attempts without improving the grade. The inference is to look at the next day’s error.
6. A digital quiz is not the regime. Higham et al. (2022) sustain a semester with correction; Sigayret et al. (2026) do not obtain the effect if engagement is thin.

None of this asks for more examinations. It asks that the certifying function not be disguised as a technique. The low-stakes regime demands more design and less threat.

11. Limitations

The review is of Crossref alone: no other databases, no books, no theses and nothing earlier than 2021. It does not re-estimate the classic effect and does not present twenty-four works as the history of the phenomenon. Manual selection, from 171 records to 24, is a judgment. Without a meta-analysis there is neither a single school effect nor a formal comparison between spacing and interleaving. The 66% and the 34% in Gereau and Mulligan (2026) describe their meta-analysis of free recall, not this corpus.

Five sources have no abstract: Sumeracki and Castillo (2022), Carpenter (2023), Pan et al. (2024), Lehman (2025) and Weissgerber and Rummer (2023). The set leans toward higher education, as against few primary studies and one science study in grades 9 to 12. Where the abstract locates the place, English schools, Dutch primary schools and adults in the United States appear. Several experiments do not declare a country. There is no direct Mexican evidence; to speak of an Anglophone and European predominance describes what these abstracts make visible, not a census. The vignette illustrates. It does not prove.

The null results restrain the pamphlet. Sigayret et al. (2026) do not find the effect in two online experiments. Stern and Halamish (2023) do not find that retrieval exceeds restudy of texts. Ophuis-Cox et al. (2026) do not find a significant difference in application. Broeren et al. (2021) do not find a correlation between greater use and performance. Zheng et al. (2023) tie the benefit to capacity and demand. Cairney et al. (2026) speak of adults. Dyslexia remains, in Wilschut et al. (2026), without a trial in this corpus. Changing the format of Friday does not guarantee the outcome of a given school.

12. Conclusions

Retrieval practice deserves its place in the psychology of learning and not the use made of it in the vignette. Trying to recall, when the phenomenon is isolated, usually favors retention relative to restudy. The recent corpus qualifies that statement: the free-recall literature mixes two effects, the contrast can be null if engagement is thin, and an examination without return of the work is not the phenomenon. A laboratory effect does not by itself constitute a school practice. It is, at most, a reason to design one.

What holds is tied to a regime, and it is an empirical finding: greater fluency on the tables with cards than by singing them; in secondary school, an advantage of quizzing and, on top of that, of interleaving; in calculus, greater end-of-semester retention when problems are distributed, with a cost in initial practice; relearning with correction above restudy; greater recall of what was tested in special education; and an advantage also with attention-deficit disorder, without offsetting poor encoding.

What is not automatic keeps the pamphlet at bay. Spelling application did not, in a significant way, accompany the better retention of the rule. Application of complex concepts was absent after one round measured at eight minutes and was more likely with three rounds and one week. Free recall of texts did not exceed restudy at two days. A sample platform did not detect the effect. Greater self-regulation did not correlate with performance. Working memory moderated the benefit when demand was high.

Without adoption, the regime is a bibliography. There is teacher use that may be assessment under another name, interventions that this review does not detail, beliefs that are perhaps less naive and still unquantified, and an open field. For whoever recognizes the vignette, the conclusion is a design inference: a low-stakes regime — spacing, correction, type of knowledge and the profile of the person who learns — not more examinations. Measuring it in Mexican and Latin American classrooms remains an empirical debt.

NEXTECH.IA Editorial Lab / Ingeniero Mitre

References

1. Bates, G. (2025). ‘What’s in a name?’ Retrieval practice in the primary classroom: a case study. Education 3-13. https://doi.org/10.1080/03004279.2025.2504492
2. Bates, G., & Shea, J. (2024). Retrieval Practice “in the Wild”: Teachers' Reported Use of Retrieval Practice in the Classroom. Mind, Brain, and Education, 18(3), 249–257. https://doi.org/10.1111/mbe.12420
3. Broeren, M., Heijltjes, A., Verkoeijen, P., Smeets, G., & Arends, L. (2021). Supporting the self-regulated use of retrieval practice: A higher education classroom experiment. Contemporary Educational Psychology, 64, 101939. https://doi.org/10.1016/j.cedpsych.2020.101939
4. Cairney, B. E., Haebig, E., & Lucas, H. D. (2026). Examining the Roles of Retrieval Practice and Post-Retrieval Feedback in Learning Among Autistic Adults. Autism, 30(9), 2272–2287. https://doi.org/10.1177/13623613261464659
5. Carpenter, S. K. (2023). Encouraging Students to Use Retrieval Practice: a Review of Emerging Research from Five Types of Interventions. Educational Psychology Review, 35(4), Article 96. https://doi.org/10.1007/s10648-023-09811-8
6. Corral, D., & Carpenter, S. K. (2025). Effects of retrieval practice on retention and application of complex educational concepts. Learning and Instruction, 100, 102219. https://doi.org/10.1016/j.learninstruc.2025.102219
7. Gereau, A. D., & Mulligan, N. W. (2026). Research on “the testing effect” routinely conflates direct and forward testing effects: A meta-analysis of testing effects with free recall. Journal of Experimental Psychology: Learning, Memory, and Cognition. https://doi.org/10.1037/xlm0001634
8. Higham, P. A., Zengel, B., Bartlett, L. K., & Hadwin, J. A. (2022). The benefits of successive relearning on multiple learning outcomes. Journal of Educational Psychology, 114(5), 928–944. https://doi.org/10.1037/edu0000693
9. Lehman, M. (2025). Optional review activities with retrieval practice improve learning and reduce test anxiety. Scholarship of Teaching and Learning in Psychology. https://doi.org/10.1037/stl0000451
10. Lyle, K. B., Bego, C. R., Ralston, P. A. S., & Immekus, J. C. (2022). Spaced Retrieval Practice Imposes Desirable Difficulty in Calculus Learning. Educational Psychology Review, 34(3), 1799–1812. https://doi.org/10.1007/s10648-022-09677-2
11. Minear, M. E., Coane, J. H., Cooney, L. H., Boland, S. C., & Serrano, J. W. (2023). Is practice good enough? Retrieval benefits students with ADHD but does not compensate for poor encoding in unmedicated students. Frontiers in Psychology, 14, Article 1186566. https://doi.org/10.3389/fpsyg.2023.1186566
12. Ophuis-Cox, F. H. A., Catrysse, L., & Camp, G. (2023). The effect of retrieval practice on fluently retrieving multiplication facts in an authentic elementary school setting. Applied Cognitive Psychology, 37(6), 1463–1469. https://doi.org/10.1002/acp.4141
13. Ophuis-Cox, F. H. A., Catrysse, L., & Camp, G. (2026). Mixed effects of retrieval practice on the acquisition of procedural knowledge in verb spelling education. Learning and Instruction, 104, 102377. https://doi.org/10.1016/j.learninstruc.2026.102377
14. Opitz, B., & Kubik, V. (2024). Far transfer of retrieval-practice benefits: rule-based learning as the underlying mechanism. Cognitive Research: Principles and Implications, 9(1), Article 65. https://doi.org/10.1186/s41235-024-00598-y
15. Pan, S. C., Dunlosky, J., Xu, K. M., & Ouwehand, K. (2024). Emerging and Future Directions in Test-Enhanced Learning Research. Educational Psychology Review, 36(1), Article 20. https://doi.org/10.1007/s10648-024-09857-2
16. Rawson, K. A., & Dunlosky, J. (2022). Successive Relearning: An Underexplored but Potent Technique for Obtaining and Maintaining Knowledge. Current Directions in Psychological Science, 31(4), 362–368. https://doi.org/10.1177/09637214221100484
17. Sana, F., & Yan, V. X. (2022). Interleaving Retrieval Practice Promotes Science Learning. Psychological Science, 33(5), 782–788. https://doi.org/10.1177/09567976211057507
18. Sigayret, K., Parmentier, J.-F., & Silvestre, F. (2026). Testing the testing effect on prolific: when retrieval practice fails to boost learning. Frontiers in Psychology, 17, Article 1727423. https://doi.org/10.3389/fpsyg.2026.1727423
19. Stern, P., & Halamish, V. (2023). Free-recall retrieval practice tasks for students with ADHD: whole-text versus section recall. Frontiers in Psychology, 14, Article 1301726. https://doi.org/10.3389/fpsyg.2023.1301726
20. Sumeracki, M. A., & Castillo, J. (2022). Covert and overt retrieval practice in the classroom. Translational Issues in Psychological Science, 8(2), 282–293. https://doi.org/10.1037/tps0000332
21. Tempel, T., & Sollich, S. (2023). Retrieval-based learning in special education. Journal of Research in Special Educational Needs, 23(3), 244–250. https://doi.org/10.1111/1471-3802.12594
22. Weissgerber, S. C., & Rummer, R. (2023). More accurate than assumed: Learners’ metacognitive beliefs about the effectiveness of retrieval practice. Learning and Instruction, 83, 101679. https://doi.org/10.1016/j.learninstruc.2022.101679
23. Wilschut, T., Sense, F., & van Rijn, H. (2026). Trends in testing effect research: from lab to classroom, but not yet for all learners. npj Science of Learning, 11(1), Article 25. https://doi.org/10.1038/s41539-026-00400-2
24. Zheng, Y., Sun, P., & Liu, X. L. (2023). Retrieval practice is costly and is beneficial only when working memory capacity is abundant. npj Science of Learning, 8(1), Article 8. https://doi.org/10.1038/s41539-023-00159-w