Pretesting enhances learning¶

Claim · All claims
Evidence · 3 studies · 2 causal r1, 1 review r3 · q3 · i2–i3
Currency · newest study 2023 · 1 of 3 entries checked against their sources (automated) · not reviewed by a person

Attempting to answer questions about material before it has been taught — even when those attempts fail — improves later retention of that material relative to studying for the same time without a pretest, provided the attempt is followed by the answer or the instruction. The two multi-experiment laboratory studies recorded here (undergraduates; expository text, fictional trivia and word pairs) test retention, not transfer.

The claim covers pre-instruction testing on not-yet-learned material; it is distinct from retrieval practice on already-learned material.

Subclaims¶

q3 i3 Testing learners on to-be-read educational material before they read it (a "pretest," items answered essentially at floor) produces better retention of that material than giving equivalent extra study time, even though only pretest-failed items are analyzed. → Richland Kornell and Kao 2009

q3 i2 Unsuccessful attempts to retrieve an answer, when followed by feedback, enhance subsequent learning of that answer relative to simply studying the same material for an equal amount of time, both for fictional trivia facts and for weak word-associate pairs. → Kornell Hays and Bjork 2009

q3 i2 Failed retrieval attempts on material with no existing memory trace (fictional trivia questions with no true answer) still improved later recall relative to reading the question and answer together for the same time; every trial ended with the correct answer shown, and no condition without the answer was tested, so the study shows the attempt enhancing the encoding that followed it and does not show what an attempt does without one. → Kornell et al. 2009

q3 i3 Testing readers on passage content before they had read it ("pretesting") produced better retention than giving equivalent extra study time; every pretest was followed by reading the full passage, so the study shows the unsuccessful attempt enhancing the reading that followed it, and does not test a pretest without that reading. → Richland et al. 2009

q3 i? Participants who attempted practice questions before reading a text passage outperformed controls on posttest questions targeting that content, both immediately and after one week. → Steven C. Pan 2023

Evidence¶

Richland Kornell and Kao 2009¶

Richland, L. E., Kornell, N., & Kao, L. S. (2009). The pretesting effect: Do unsuccessful retrieval attempts enhance learning? Journal of Experimental Psychology: Applied, 15(3), 243–257. doi:10.1037/a0016496

q3 · peer-reviewed multi-experiment lab study · i3 · large effect, d=1.1 (Experiment 1) · n=63 undergraduates (Experiment 1); 5 experiments, ns 59–158 · causal · r?

Across five experiments, undergraduates read an expository essay on human vision. In the "test" condition they were asked questions about upcoming concepts before reading the passage (nearly all answered incorrectly or left blank at pretest); in the "extended study" condition they instead got proportionally more time to read. On a later test of the same material, the pretested group outperformed the extended-study group in every experiment — in Experiment 1, 75% vs. 56% correct, a large effect (d = 1.1) — even though the analysis excluded any item a participant had actually gotten right on the pretest, so the benefit cannot be attributed to successful retrieval. Later experiments ruled out mere attention-direction (emphasizing the same concepts with italics/bold in the study condition) as the explanation, and Experiment 5 found that simply reading the questions without attempting an answer produced a smaller benefit than actually attempting to answer them.

Kornell Hays and Bjork 2009¶

Kornell, N., Hays, M. J., & Bjork, R. A. (2009). Unsuccessful retrieval attempts enhance subsequent learning. Journal of Experimental Psychology: Learning, Memory, and Cognition, 35(4), 989–998. doi:10.1037/a0015729

q3 · peer-reviewed multi-experiment lab study · i2 · medium effect, d=0.58 (Experiment 1) · n=25 UCLA undergraduates (Experiment 1); 6 experiments, ns 20–32 · causal · r1

Six experiments tested whether failed retrieval attempts help or hurt later learning, using materials engineered so retrieval would fail: fictional trivia questions (Experiments 1–2, e.g. "What peace treaty ended the Calumet War?") and cue–weak-associate word pairs (Experiments 3–6, e.g. cue "whale," target "mammal"), with any rare correct guesses excluded from analysis. In the test condition participants attempted an answer before being shown it; in the read-only condition the question and answer were simply presented together for the same study time. On a later cued-recall test, the test condition beat the read-only condition in every experiment — in Experiment 1, 41% vs. 31% correct, t(24) = 2.97, p < .01, d = 0.58 — and in Experiment 6, initial wrong guesses ("commission errors") were recalled better than initially blank items, arguing against the idea that producing an incorrect answer is itself harmful.

Steven C. Pan 2023¶

Steven C. Pan, Shana K. Carpenter. (2023). Prequestioning and Pretesting Effects: a Review of Empirical Research, Theoretical Perspectives, and Implications for Educational Practice. Educational Psychology Review. https://doi.org/10.1007/s10648-023-09814-5

q3 · i? · review · r3

Narrative review of Richland et al. (2009), in which participants attempted five practice questions on a cerebral achromatopsia passage before reading; the review reports that "participants in the experimental group consistently outperformed the control group" on immediate and one-week delayed posttests.

"When the test questions were re-administered on an immediate posttest (experiments 1–3) or on a one-week delayed posttest (experiment 4), however, participants in the experimental group consistently outperformed the control group"

Studies bearing on this claim¶

Does attempting questions on material before studying it improve later memory of the questioned content, and under what conditions on the answer and the material? This table gathers the studies behind this claim and the 11 narrower claims that bear on it: 8 distinct studies (6 causal, 2 review), q2–q3, 3 of 8 with an effect size coded. One row per study, codes as each entry records them; no effect is averaged and no verdict is drawn. Built by scripts/synthesize_claim.py from eval/synthesis/; a row that looks wrong is fixed on the claim it comes from.

Discussion¶

Mechanism. Pretesting is typically explained through productive failure and search-set activation: an unsuccessful attempt to answer a question makes learners aware of gaps in their knowledge, activates related prior knowledge, and focuses attention on the to-be-learned answer during subsequent instruction. The error itself is not harmful — what matters is that the attempt creates a "search set" that the correct answer can then resolve. This connects to Activation and to Conflict-based instruction improves science conceptual learning, though staged contradictions helped only learners who reported being confused, and no study isolates disequilibrium as the mechanism.

Boundary conditions. The pretesting benefit appears strongest when the pretest questions are conceptually aligned with the subsequent instruction, when learners receive feedback or corrective instruction after the attempt, and when the material is meaningful rather than arbitrary. Pretests on entirely unrelated material, or pretests that consume so much time that instruction is curtailed, show weaker or no benefits. Low-stakes framing matters: pretests are diagnostic, not evaluative, and grading them can undermine the exploratory mindset that makes them effective — see Assessment for learning improves achievement.

Relation to retrieval practice. Pretesting differs from standard retrieval practice: retrieval practice strengthens already-learned material, whereas pretesting operates on not-yet-learned material and works despite failure. Designers should not assume the two effects have identical moderators — in particular, the feedback-and-instruction phase is constitutive of the pretesting effect, not merely an adjunct.

Design implications. Practical deployments include opening a unit with two or three conceptually central questions learners cannot yet answer, using diagnostic "warm-up" quizzes in flipped or blended settings, and framing pretests explicitly as low-stakes so that guessing is encouraged rather than penalized. Because the benefit depends on subsequent instruction resolving the search set, a pretest without timely, well-aligned instruction is likely wasted time.

Open questions. The durability of pretesting benefits over long retention intervals, and their magnitude relative to equivalent time spent studying, remain actively debated in the literature. Two multi-experiment laboratory studies are recorded above (Richland et al. 2009, five experiments with expository text; Kornell et al. 2009, six with trivia and word pairs), both with undergraduates and retention intervals of at most one week (the review entry reports a one-week delayed posttest in Richland et al.'s Experiment 4). A classroom study and a retention-interval comparison up to one week are recorded on narrower claims (Classroom pretesting improves high-stakes exam performance on pretested content; The pretesting effect on text memory grows with retention interval); delays longer than a week and transfer are not recorded.

Merged from “Pretesting Improves Retention” (pretesting-improves-retention): The pretesting (prequestioning) effect. Failed retrieval attempts before instruction appear to prime learners to encode the corrective information presented afterward. The dominant explanations are search-set activation — the pretest activates related prior knowledge and narrows what learners look for in the to-be-learned material — and productive failure, in which the experience of not knowing creates a gap that subsequent instruction fills. Both accounts predict that the pretest need not be answered correctly to confer a benefit, which distinguishes pretesting from activation aimed only at surfacing prior knowledge.

Scope and moderators. The effect has been reported most consistently for verbal, fact-like material (word pairs, trivia, expository text) and for relatively short retention intervals. Whether pretesting benefits transfer or complex-skill learning, and whether gains persist over long delays, remain open questions. Feedback after the pretest is likely important: without corrective information, a failed guess may entrench errors rather than prime correction — a boundary condition shared with claims about when retrieval practice helps and when errors persist.

Practical form. Pretests should be low- or no-stakes, brief, and on the same content as the upcoming instruction. They pair naturally with retrieval practice during and after instruction, forming a test–study–test cycle. In classroom settings, brief prequestions embedded at the start of a lesson or online module are the lowest-cost implementation.

Constraints. Pretesting gains shrink or vanish when the pretest targets content unrelated to the subsequent instruction, when no corrective feedback or instruction follows the failed attempt, and possibly when learners lack enough prior knowledge to generate even plausible guesses. High-stakes pretests can also induce anxiety and discourage the productive guessing the effect depends on.

Single-family limitation. The page documents two peer-reviewed laboratory studies from overlapping authors (Kornell is on both), so the evidence here is one research programme's; independent replications are not yet recorded.

Merged from “Retrieval Fails Without Encoding” (retrieval-fails-without-encoding): (That page's title overstated its evidence: its two studies, now the entries above, show unsuccessful retrieval attempts followed by the answer enhancing learning; neither tests retrieval with no encoding at all. The paragraphs below are its argument, not findings of those studies.) Mechanism. Retrieval practice works by strengthening and making accessible existing memory traces. If the trace is weak, incomplete, or never formed — because the learner skimmed, was distracted, or experienced cognitive overload during initial instruction — retrieval attempts have nothing to consolidate. This places a sequencing constraint on design: meaningful encoding activities (worked examples, explanation, elaboration) must precede or accompany retrieval, not be replaced by it. Where instruction itself exceeds working memory capacity, the resulting trace is too fragile for retrieval to strengthen — see Cognitive overload degrades learning. and Chunking reduces working memory load..

Boundary conditions. The claim does not mean retrieval is useless for novices. Successful retrieval of even partially encoded material can itself serve as a restudy event, and feedback after failed retrieval can support encoding. The failure mode is retrieval without feedback or before any exposure — e.g., quizzing students on content never taught, or using pre-tests as the sole instructional event. Designers should therefore treat retrieval as a consolidation phase layered on top of worked examples and explicit instruction, not as a substitute for them.

Design implications. Sequence instruction encode-first: worked examples, advance organizers, or activation of prior knowledge before the first retrieval attempt; ensure feedback follows every retrieval attempt, especially failed ones; and treat pre-tests as activation devices rather than as the sole instructional event. Retrieval-based formats such as quizzing and flashcards should be scheduled after, not instead of, initial instruction.

Open questions. The precise threshold of encoding strength at which retrieval becomes beneficial remains contested. For entirely novel material, Kornell et al. (2009) found failed retrieval followed by the answer beat equal-time reading (fictional trivia, d = 0.58 in Experiment 1), though in the laboratory and over short intervals.

Related Claims¶

• Retrieval practice improves retention — the closest cousin; pretesting extends testing effects to pre-instruction attempts
• Activation improves learning — pretesting is a form of prior-knowledge activation that surfaces gaps
• Conflict-based instruction improves science conceptual learning, though staged contradictions helped only learners who reported being confused, and no study isolates disequilibrium as the mechanism — failed pretest attempts create the disequilibrium that instruction then resolves
• Assessment for learning improves achievement — pretests as low-stakes, formative rather than evaluative assessment
• Productive failure improves learning — the broader pattern that failed attempts before instruction can outperform instruction alone
• Retrieval practice — the strategy page covering test–study–test cycles
• Pre-tests reduced adult learners' persistence in a MOOC, though they improved post-test scores among those who completed it — related
• Feedback Enhances Retrieval Practice — related
• Retrieval Failure Reduces Benefit — related
• Question prompts improve learning — related
• Chunking reduces working memory load. — overloaded working memory during instruction prevents encoding in the first place
• Cognitive overload degrades learning. — a primary cause of encoding failure
• Worked examples reduce unnecessary search for novices. — an encoding-first alternative to premature problem-solving or retrieval
• Advance organizers improve learning. — a structure-building device that supports the encoding retrieval depends on
• Retrieval practice effects on mediator-cued final tests have been positive, but Coppens et al. (2016) concluded the true effect may be only about 0.10 to 0.20 — related
• Two-part collaborative assessment, individual then group answering, transforms summative testing into a learning experience and may reduce test anxiety — related
• Retrieval practice strengthens memory and produces stronger, more enduring learning — related
• Classroom pretesting improves high-stakes exam performance on pretested content by 8–9% over a multi-week course — a narrower finding that bears on this claim
• A review reports that pretesting helps paired-associate learning only when correct-answer feedback is immediate, but helps trivia-fact learning even when feedback is delayed up to 24 hours — a narrower finding that bears on this claim
• Learners fail to recognize the benefits of pre-instruction testing, ranking it as least effective despite benefiting from it — a narrower finding that bears on this claim
• Pre-instruction testing enhances memory for directly tested video content, with benefits shown for videos from 5 to 30 minutes — a narrower finding that bears on this claim
• Effects of pre-instruction testing on untested but studied materials are highly inconsistent across experiments and materials — related
• Pretesting benefits for paired associates depend on semantic association between pair members, with cued-recall benefits absent for unrelated pairs — a narrower finding that bears on this claim
• Practicing retrieval of some portions of an educational text can enhance retention of related nontested portions (retrieval-induced facilitation), but not when relational encoding is disrupted — related
• Questions interpolated between video lecture segments improve subsequent test performance and reduce mind wandering — related
• Multiple-choice testing improves later cued recall of prose facts, a benefit that persists though reduced after a one-week delay — related
• Active control over the learning experience enhances recognition memory from age 7 into adulthood — related
• Reading passages without intent to learn was classified as rest because participants were not informed of a later test — related
• Pretesting before instruction enhances learning of subsequently presented material even when pretest answers are wrong, with multiple-choice pretesting more effective than cued-recall pretesting — a narrower finding that bears on this claim
• The pretesting effect on text memory grows with retention interval, being largest at one week — related
• Generation Effect Improves Retention — related
• Unsuccessful retrieval attempts confer a smaller benefit to future performance than successful retrievals, whose advantage grows with item difficulty — related
• Students already show a pretest congruency effect: existing knowledge yields better baseline performance for schema-related questions — related
• The study addresses whether performance on the low-stakes PISA reading and mathematics assessments can be fostered by coaching and pretesting — a narrower finding that bears on this claim