Does Testing Help You Learn What You Weren't Tested On?

The question, and why it matters

There is an awkward gap between what students do to study and what actually works. Most people reread, highlight, and review — strategies that feel productive and mostly are not. The single most robust finding in the psychology of learning is that retrieval practice — closing the book and trying to recall the material, rather than looking at it again — produces far more durable memory than restudy [1,2]. This is why flashcards, low-stakes quizzes, and "brain dumps" are pushed so hard in classrooms and learning advice.

But the reason people study is rarely to recite facts. The point is usually to use knowledge in a situation that is not identical to the one where it was learned: to solve a new problem, answer a new question, write in a new genre, or reason about a new case. That is transfer, and it is where learning either pays off or does not. So the question this brief addresses is narrower and more consequential than "does testing work":

Does retrieval practice produce learning that transfers to tasks, materials, and contexts the learner was never tested on — and if so, how far does that transfer reach?

This matters to anyone who teaches, studies, trains, or designs learning technology, and to the wider public argument about whether schools should spend scarce time on practice testing. The popular claim — "testing improves learning" — is true and well-evidenced. The stronger claim that testing builds general, flexible understanding you can apply anywhere is where the evidence thins out, and conflating the two leads to real mistakes in classrooms, in study advice, and in the multi-billion-dollar "science of learning" industry.

What testing does, and what "transfer" means

Four effects hide inside the phrase "the testing effect," and they are not interchangeable [9]. Distinguishing them is the key to reading this literature honestly.

• The direct testing effect. Taking a test improves later memory for the tested material itself, relative to restudying it. This is the oldest and strongest result [1,2].
• The forward testing effect (test-potentiated learning). A test on earlier material also improves encoding of material studied afterward [18]. This can be mistaken for transfer if a study does not isolate it.
• Transfer. Performance on a new test that differs from the practice test — a different question format, a new item, an inference, an application, or a new context entirely [6,7,8,20].
• Classroom achievement. Gains on real course assessments — exams, quizzes, grades — which mix all of the above plus motivation, time-on-task, and test anxiety [4,5].

Within transfer, a further distinction recurs: near transfer applies the learned knowledge in a context clearly related to where it was learned (e.g., a new problem using the same rule in the same domain), while far transfer requires the learner to recognise that the knowledge is relevant in a context that shares few surface features [6,14,15]. The gap between these two is the heart of the matter.

The meta-analytic verdict: real but moderate, and smaller under correction

The most careful synthesis of the retention literature is Rowland's [2]: across 61 experiments, testing beat restudy on retention by about half a standard deviation (g ≈ 0.50), with initial recall tests outperforming recognition and with effortful processing doing the causal work. Larger pools reach similar conclusions: Adesope and colleagues found an overall advantage of practice tests over all non-testing conditions of g ≈ 0.70 [3], and Yang and colleagues integrated 48,478 students across 222 classroom studies and found a medium effect of quizzing on achievement, g ≈ 0.50 [4]. Agarwal and colleagues reviewed 50 genuine classroom experiments and found that 57% produced medium or large benefits — but only 6% were conducted outside Western, educated, industrialised samples, a real limit on generality [5].

The decisive question is transfer, and the decisive synthesis is Pan and Rickard's, which assembled 192 transfer effect sizes from 122 experiments (N ≈ 10,382) spanning 40 years and found an average transfer advantage over a re-exposure control of d = 0.40 (95% CI 0.31–0.50) [6]. That is a moderate effect — meaningful, but well below the reputation the technique enjoys. Then comes the correction that the popular account almost always omits: when the authors applied publication-bias methods (PET-PEESE and selection models), the core moderator effects barely moved, but the intercept — the baseline transfer benefit when none of the favourable moderators are present — fell substantially, "often indicating no positive transfer" at all [6]. In other words, the average hides a distribution in which the beneficial cases are those with particular features, and the ordinary case can be null.

Where transfer works: near, inferential, and feedback-supported

The same meta-analysis maps where transfer succeeds [6]. It is greatest across test formats (practising recall helps even when the final test is multiple-choice), and — importantly for the "does it help you think?" question — to application and inference questions, to medical-diagnosis problems, and to related word cues. Butler's classic experiments showed the pattern directly: repeated testing produced better performance than repeated studying on new inferential questions, both within the same knowledge domain and, more weakly, across domains [7]. Karpicke and Blunt found that retrieval practice beat elaborative concept mapping on inference and comprehension questions, not just verbatim recall, by a large margin (d ≈ 1.0–1.5 in their texts) [8].

The most recent primary work sharpens the conditions. Corral and Carpenter found that retrieval practice improved both retention and the application of complex, educationally realistic concepts — but crucially, the transfer advantage appeared only after three rounds of retrieval practice and a one-week delay, and it was clearest on the recognition component of transfer: the gain was in noticing that a concept applied, not only in retrieving its definition [14]. Franzoi and colleagues found retrieval practice improved learning in real primary-school classrooms, whether the practice was massed or distributed [21]. Rohrer, Taylor, and Sholar found, in fourth and fifth graders, that the testing advantage on a transfer test was no smaller — slightly larger — than on a repeated test [20]. And Opitz and Kubik reported far transfer (a grammaticality-judgement task) a week after learning, arguing the benefit came from acquiring the underlying rule rather than memorising instances [16].

Three moderators consistently predict whether transfer will materialise [6]: response congruency (how much the practice task and the transfer task require the same kind of response), elaborated retrieval (retrieving richer semantic detail, not just a bare word), and initial retrieval success. A newer experiment challenges the last of these: Chan and colleagues manipulated retrieval difficulty directly and found the magnitude of the testing effect to be independent of how well people performed during practice [10]. That is a live disagreement, and it matters because "only count the successful retrievals" is a common methodological move.

Where it breaks down: far transfer and untested material

The weaknesses are equally clear. In Pan and Rickard's synthesis, transfer was weakest to rearranged stimulus–response items, to material that was studied but never tested, and to worked-example problems [6]. Brunyé and colleagues, using map learning, found that retrieval practice improved near-to-medium transfer (reconstructing the map, judging directions from new orientations) but did not help with the medium-to-far transfer tasks that required taking an alternate perspective or navigating from memory [15]. This is the recurring pattern: testing helps you use the knowledge you practised, in ways that overlap with how you practised it; it does not automatically install a portable, context-free skill.

The measurement problem compounds the optimism. Gereau and Mulligan's 2026 meta-analysis of free-recall testing effects found that 66% of the studies labelled as testing-effect studies were "mixed" designs that confound the direct testing effect with the forward testing effect, and only 34% isolated the direct effect — and the mixed designs produced larger effects [17]. Much of the literature's apparent size is therefore inflated by designs that credit retrieval for benefits that partly come from testing having changed how people studied what came later.

The comparison that matters: retrieval against other good strategies

The "testing beats restudying" result is robust, but restudying is a weak benchmark. The practically relevant comparison is retrieval against other effective strategies — elaborative encoding, generation, concept mapping, worked-example study. Here the picture changes sharply. A 2025 meta-analysis of 44 studies and 142 direct comparisons found the overall advantage of retrieval over elaborative conditions was small (g = 0.14), was conditional on feedback (g = 0.50 when feedback was given; elaborative encoding was better than retrieval when it was not), and was essentially confined to comparisons against concept mapping and group discussion [11]. Transfer, moreover, was tested in only 29 of the 142 comparisons [11]. Jaeger and colleagues found that for some associative-learning materials, simple elaborative encoding (imagining scenes, generating sentences) beat retrieval practice outright [12]. Murray and colleagues' meta-analysis of mathematics learning found the testing-versus-restudy effect was g = 0.18 with a confidence interval crossing zero — not robust in that domain — while spaced practice showed a more reliable small-to-medium effect, g = 0.28 [13].

Put together: retrieval practice is not a general-purpose cognitive enhancer that dominates all other study methods. It is one strong tool whose edge over restudy is large and whose edge over other good, effortful strategies is modest and feedback-dependent. That is a very different message from "just test yourself and you'll learn anything better."

Why it works when it works: mechanisms

Several accounts are compatible and do not need a single winner [2,6,9].

• Effortful retrieval changes the memory itself. The act of pulling an item from memory is not neutral replay; it strengthens and modifies the trace, and the more effortful the successful retrieval, the larger the later benefit [2].
• Mediator and elaboration effects. Retrieval prompts learners to generate the semantic links (mediators) that later support flexible access; bare cued recall generates fewer, which is why elaborated retrieval transfers better [6,11].
• Transfer-appropriate processing and encoding variability. Retrieving in varied contexts and formats produces a more flexible memory that is less tied to the original cues — which is exactly why transfer is best across formats [6,9].
• Test-potentiated learning (forward effect). Testing also improves how the learner processes subsequent material, a partly separate mechanism [17,18].

The unifying theme is that transfer depends less on the strength of a memory than on whether the learner recognises that prior knowledge is relevant in the new situation [14]. Retrieval practice can build that recognition when the practice is varied and the knowledge is applied, but it cannot manufacture it from rote rehearsal.

Costs and counterevidence

Testing is not free of downsides. Retrieval can cause retrieval-induced forgetting: practising some items can temporarily impair memory for related, unpractised competitors [19]. Errors during retrieval, if not corrected by feedback, can be entrenched rather than repaired — one reason feedback is a near-necessary condition in the elaborative-encoding comparison [11]. Testing can also raise anxiety and consume class time, though the classroom evidence suggests quizzing does not systematically increase test anxiety [4,9]. And the entire body of evidence is dominated by short laboratory experiments with undergraduate participants and verbal materials; the smaller, more ecologically valid classroom literature is more encouraging but thinner and heavily WEIRD [5].

Evidence quality

• Strong: the direct testing effect for practised material (many experiments, multiple meta-analyses, converging estimates around g = 0.5) [1,2,3,4].
• Moderate: moderate transfer to near, inferential, and application tasks, with feedback and sufficient retrieval [6,7,8,14,20].
• Weak / contested: far transfer to genuinely new contexts [6,15], the size of the advantage over good elaborative strategies [11,12], and whether retrieval success moderates the effect [6,10].
• Measurement caveat: the headline numbers are inflated by designs that confound direct and forward effects [17] and by a literature that rarely tests transfer at all [11].

Calibrated confidence

• ~90% retrieval practice improves retention of the practised material more than restudy, across ages, materials, and settings.
• ~75% it produces a real but moderate advantage on near and inferential transfer, concentrated where feedback is given and retrieval succeeds.
• ~70% far transfer to contexts sharing few surface features is weak or absent as a general effect; the positive far-transfer demonstrations are real but conditional.
• ~70% the advantage over other good study strategies (elaborative encoding, generation, worked-example study) is small once feedback and exposure are equated, and can reverse without feedback.
• ~60% the field's average transfer estimate is inflated by confounded designs and publication bias; the true baseline effect without favourable moderators is close to zero.

Reversal conditions

I would materially revise these numbers if: (a) preregistered, adequately powered experiments repeatedly produced far transfer after controlling exposure and feedback; (b) the retrieval-versus-elaboration comparison reversed under standardised feedback and materials (some work already points that way [11,12]); (c) large, diverse classroom trials found durable transfer to standardised, near-independent assessments; or (d) replication showed the direct testing effect to be largely a forward-effect artefact [17]. Conversely, evidence that transfer scales with the variety of retrieval cues and the explicitness of application practice would strengthen the "retrieval builds flexible knowledge" account.

The gap that would settle it

The literature is not exhausted, but the next material advance is not more small lab experiments; it is a single well-powered, preregistered experiment that isolates the direct testing effect from the forward effect while measuring transfer to a genuinely novel task, with feedback and exposure held constant. Existing designs almost never do all four at once [11,17]. Until such a study exists, the best-supported answer to the question in the title is: yes, if "learned" means you can use it in a recognisably related way; much less, and only under conditions, if it means you can apply it somewhere new.

References

1. Roediger HL, Karpicke JD. Test-enhanced learning: taking memory tests improves long-term retention. Psychological Science. 2006;17(3):249-255. doi:10.1111/j.1467-9280.2006.01693.x
2. Rowland CA. The effect of testing versus restudy on retention: a meta-analytic review of the testing effect. Psychological Bulletin. 2014;140(6):1432-1463. doi:10.1037/a0037559
3. Adesope OO, Trevisan DA, Sundararajan N. Rethinking the use of tests: a meta-analysis of practice testing. Review of Educational Research. 2017;87(3):659-701. doi:10.3102/0034654316689306
4. Yang C, Luo L, Vadillo MA, Yu R, Shanks DR. Testing (quizzing) boosts classroom learning: a systematic and meta-analytic review. Psychological Bulletin. 2021;147(4):399-435. doi:10.1037/bul0000309
5. Agarwal PK, Nunes LD, Blunt JR. Retrieval practice consistently benefits student learning: a systematic review of applied research in schools and classrooms. Educational Psychology Review. 2021;33(4):1409-1453. doi:10.1007/s10648-021-09595-9
6. Pan SC, Rickard TC. Transfer of test-enhanced learning: meta-analytic review and synthesis. Psychological Bulletin. 2018;144(7):710-756. doi:10.1037/bul0000151
7. Butler AC. Repeated testing produces superior transfer of learning relative to repeated studying. Journal of Experimental Psychology: Learning, Memory, and Cognition. 2010;36(5):1118-1133. doi:10.1037/a0019902
8. Karpicke JD, Blunt JR. Retrieval practice produces more learning than elaborative studying with concept mapping. Science. 2011;331(6018):772-775. doi:10.1126/science.1199327
9. Carpenter SK, Pan SC, Butler AC. The science of effective learning with spacing and retrieval practice. Nature Reviews Psychology. 2022;1(9):496-511. doi:10.1038/s44159-022-00089-1
10. Chan JCK, Davis SD, Yurtsever A, Myers SJ. The magnitude of the testing effect is independent of retrieval practice performance. Journal of Experimental Psychology: General. 2024;153(7):1816-1837. doi:10.1037/xge0001593
11. Gonçalves AO, Muniz BFB, Jaeger A. Retrieval practice versus elaborative encoding: a systematic and meta-analytic review. Educational Psychology Review. 2025;37(4):100. doi:10.1007/s10648-025-10076-6
12. Jaeger A, Martins THPG, Rodrigues JPP, Muniz BFB, Fonseca ALSS, Gonçalves AO. The benefits of elaborative encoding over retrieval practice for associative learning. Memory & Cognition. 2024. doi:10.3758/s13421-024-01671-z
13. Murray E, Horner AJ, Göbel SM. A meta-analytic review of the effectiveness of spacing and retrieval practice for mathematics learning. Educational Psychology Review. 2025;37(3):75. doi:10.1007/s10648-025-10035-1
14. Corral D, Carpenter SK. Effects of retrieval practice on retention and application of complex educational concepts. Learning and Instruction. 2025;100:102219. doi:10.1016/j.learninstruc.2025.102219
15. Brunyé TT, Smith AM, Hendel D, Gardony AL, Martis SB, Taylor HA. Retrieval practice enhances near but not far transfer of spatial memory. Journal of Experimental Psychology: Learning, Memory, and Cognition. 2020;46(1):24-45. doi:10.1037/xlm0000710
16. Opitz B, Kubik V. Far transfer of retrieval-practice benefits: rule-based learning as the underlying mechanism. Cognitive Research: Principles and Implications. 2024;9(1):65. doi:10.1186/s41235-024-00598-y
17. Gereau AD, Mulligan NW. 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. 2026 (online ahead of print). doi:10.1037/xlm0001634
18. Pastötter B, Bäuml KHT. Retrieval practice enhances new learning: the forward effect of testing. Frontiers in Psychology. 2014;5:286. doi:10.3389/fpsyg.2014.00286
19. Anderson MC, Bjork RA, Bjork EL. Remembering can cause forgetting: retrieval dynamics in long-term memory. Journal of Experimental Psychology: Learning, Memory, and Cognition. 1994;20(5):1063-1087. doi:10.1037/0278-7393.20.5.1063
20. Rohrer D, Taylor K, Sholar B. Tests enhance the transfer of learning. Journal of Experimental Psychology: Learning, Memory, and Cognition. 2010;36(1):233-239. doi:10.1037/a0017678
21. Franzoi L, Cembrani V, Mulatti C, Treccani B. Retrieval practice enhances learning in real primary school settings, whether distributed or not. Frontiers in Psychology. 2025;16:1632206. doi:10.3389/fpsyg.2025.1632206