Comparing cases side by side improves learning and transfer by a moderate average amount, though in one algebra study it raised procedural knowledge and flexibility but not conceptual knowledge¶

Claim · All claims
Evidence · 3 studies · 2 causal r?, 1 quant-synthesis r? · q3–q4 · i2 medium

Learners who compare two or more cases that differ on key features — side by side, with attention directed to what varies and why — learn more than learners who study the same cases one at a time or study a single case (d = .50 across 57 experiments), and transfer the principle more often (48% vs 19% in one negotiation experiment); in one algebra experiment the gain was in procedural knowledge and flexibility, not conceptual knowledge.

Subclaims¶

q4 i2 Across 57 experiments and 336 tests, case-comparison activities produced greater learning than sequential or single-case study, non-analogous cases, traditional instruction and controls (d = .50), with larger benefits when learners are asked to find similarities, when the principle is given after the comparison, for perceptual content, and on immediate tests. → Alfieri et al. 2013

q3 i? Undergraduate novices who compared two negotiation cases on one page were more than twice as likely to transfer the principle to a new negotiation as those who studied the same two cases separately (48% vs 19%). → Gentner et al. 2003

q3 i? Seventh-graders who compared alternative algebra solution methods side by side gained more procedural knowledge and flexibility than peers who reflected on the same methods one at a time, but gained no more conceptual knowledge. → Rittle-Johnson & Star 2007

q4 i2 Across 57 experiments (336 tests), case-comparison activities produced greater learning than sequential, single-case or non-analogous case study, traditional instruction and controls (d = .50), with larger gains when learners look for similarities, receive the principle after comparing, and are tested immediately. → Alfieri et al. 2013

q3 i? Novice undergraduates who compared two negotiation cases were more than twice as likely to transfer the underlying principle to a new negotiation (48%) as those who studied the same two cases separately (19%); no standardised effect size is reported. → Gentner et al. 2003

Evidence¶

Alfieri et al. 2013¶

Alfieri, L., Nokes-Malach, T. J., & Schunn, C. D. (2013). Learning through case comparisons: A meta-analytic review. Educational Psychologist, 48(2), 87–113. doi:10.1080/00461520.2013.775712

q4 · meta-analysis · i2 · medium effect, d=0.50, 95% CI [.44, .56] · n=57 experiments (336 tests) · quant-synthesis · r?

A random-effects meta-analysis of 57 laboratory and classroom experiments (336 tests) that compared learning from case-comparison activities with other ways of studying cases (one case, cases in sequence, non-analogous cases) and with traditional instruction or control conditions. Comparison led to greater learning overall, d = .50. Of 15 moderators tested, four reliably changed the effect: asking learners to find similarities, presenting the principle after the comparison, perceptual content, and testing immediately were all associated with larger gains. The pooled estimate mixes several kinds of comparison condition, so it is not solely a "compare vs. study the same cases one at a time" effect.

Gentner et al. 2003¶

Gentner, D., Loewenstein, J., & Thompson, L. (2003). Learning and transfer: A general role for analogical encoding. Journal of Educational Psychology, 95(2), 393–408. doi:10.1037/0022-0663.95.2.393

q3 · peer-reviewed randomised experiment (3 experiments) · i? · no standardised effect size reported; 48% vs 19% transfer, χ²(1, N=128)=11.85 · n=128 (Experiment 2) · causal · r?

In Experiment 2, 128 undergraduates learning negotiation strategies were randomly assigned either to read two example cases on one page and describe their similarities, or to read and describe the same two cases on separate pages. On a later test negotiation, 48% of the comparison group used the target principle against 19% of the separate-cases group, and comparers were also more likely to state the principle in full. Experiment 1 found a benefit of comparison over no case study, and Experiment 3 found that more comparison support raised transfer in a face-to-face negotiation. The n code counts Experiment 2 only, the source of the 48% vs 19% comparison; the paper reports three experiments, and a page counting all three records a larger n.

Rittle-Johnson & Star 2007¶

Rittle-Johnson, B., & Star, J. R. (2007). Does comparing solution methods facilitate conceptual and procedural knowledge? An experimental study on learning to solve equations. Journal of Educational Psychology, 99(3), 561–574. doi:10.1037/0022-0663.99.3.561

q3 · peer-reviewed randomised experiment · i? · no effect size in the abstract read · n=70 · causal · r?

Seventy seventh-grade students learning to solve algebra equations were randomly assigned to compare and contrast alternative solution methods or to reflect on the same methods one at a time. At posttest the compare group had made greater gains in procedural knowledge and procedural flexibility, and comparable gains in conceptual knowledge. This qualifies the claim: here comparison helped learners use and choose methods, not their grasp of the underlying concepts. Only the abstract was read.

Studies bearing on this claim¶

Does comparing analogous cases improve transfer more than studying cases one at a time? This table gathers the studies behind this claim and the 5 narrower claims that bear on it: 7 distinct studies (3 causal, 2 quant-synthesis, 2 review), q2–q4, 1 of 7 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¶

Why comparison helps. Contrasting cases make the deep structure of a domain visible: features that remain constant across cases are likely central to the principle, while features that vary are incidental. Comparison also prompts learners to notice dimensions they would otherwise overlook, supporting analogical reasoning and transfer to novel situations. This aligns with the broader claim that analogical reasoning improves transfer.

Multiple cases and cognitive flexibility. Studying multiple cases with varied surface features is a core prescription of cognitive flexibility theory; the specific benefit claimed here depends on learners actively comparing the cases rather than merely encountering them sequentially — see Cognitive flexibility theory: multiple cases support transfer..

Boundary conditions. Comparison is demanding: simultaneously holding two cases in working memory can overload novices, so the benefit likely depends on case familiarity, the salience of the contrasting features, and instructional support such as prompts naming what to compare — consistent with cognitive load management concerns. Poorly chosen cases that differ on too many dimensions invite surface-level alignment rather than structural insight. Comparison should also be distinguished from simply reading multiple examples; without explicit comparison activity, learners often process cases independently and miss the contrast. Side-by-side spatial alignment of cases appears to matter: sequential presentation of the same cases tends to yield weaker structural alignment than simultaneous, aligned presentation.

Expertise considerations. As with worked examples, the value of guided comparison likely follows an expertise-reversal pattern: novices benefit from explicit prompts naming the contrasting dimensions, while advanced learners may compare spontaneously and find such scaffolding redundant. See expertise reversal effect.

Open questions. Three studies are recorded (Gentner et al. 2003, Rittle-Johnson & Star 2007, and the Alfieri et al. 2013 meta-analysis, coded medium). Still open: how learner expertise, domain and case format moderate the effect.

Merged from “Comparing contrasting cases improves learning” (comparing-contrasting-cases-improve-learning): Mechanism. Contrasting cases work by making discriminating features visible. When two cases differ on exactly the dimension the designer wants learned, learners' attention is drawn to that dimension in a way that studying a single case cannot achieve. This aligns with the broader claim that analogical reasoning improves transfer: comparison invites learners to map structure from one case onto another, and the mapping process surfaces the relational schema that supports transfer.

Relation to multiple-case instruction. The claim is closely related to Cognitive Flexibility Theory and its recommendation of multiple representations and cases for ill-structured domains — see Cognitive flexibility theory: multiple cases support transfer in ill-structured domains.. It also underpins case-based learning designs, where case-based learning improves exam performance under some conditions.

Moderators and boundary conditions. Comparison is not automatically beneficial. If the cases differ on too many dimensions at once, learners may attend to surface features rather than the intended deep structure, and the added processing can contribute to cognitive overload degrading learning. Sequencing matters: learners typically need enough prior knowledge to interpret the cases before comparison pays off, and for novices the comparison itself imposes working-memory demands that may need scaffolding — consistent with cognitive load theory. As expertise grows, the value of guided comparison may decline in line with the expertise reversal effect.

Design implications. Effective comparisons tend to (a) vary on one or few dimensions at a time, (b) present cases side by side or in immediate succession rather than spaced apart, (c) prompt learners to state what differs and why rather than leaving comparison implicit, and (d) align case surface features so that the intended deep structure is the salient difference. These follow directly from the mechanism above: anything that obscures the discriminating dimension — extra differences, temporal separation, or unprompted comparison — weakens the effect. Prompts to articulate differences can be treated as a form of self-explanation layered onto the comparison itself.

Open questions. How the number of cases, their similarity and the amount of guidance during comparison moderate the effect is not settled by the entries recorded here.

Merged from “Multiple Contrasting Cases Support Abstraction” (multiple-contrasting-cases-support-abstraction): Mechanism. Contrasting cases make the deep structure of a domain visible by highlighting what varies and what stays constant across examples. When learners compare cases that differ on surface features but share a principle, they are pushed to explain the differences, which supports abstraction and transfer. This is the core rationale of Cognitive Flexibility Theory, which argues for multiple representations and cases in ill-structured domains — see Cognitive flexibility theory: multiple cases support flexible transfer..

Comparison as the active ingredient. The benefit is generally attributed to comparison itself, not mere exposure to multiple cases. Learners need prompts to align the cases and articulate similarities and differences; simply presenting several examples side by side without comparison support is unlikely to produce the same abstraction. This connects closely to Analogical reasoning improves transfer., where structured comparison of analogs drives schema formation. Structured comparison prompts — asking learners to identify what is the same, what differs, and why — function as a form of self-explanation applied across cases rather than within one.

Case selection matters. Cases should be minimally contrasting — similar enough to align, but differing on the dimension the learner is meant to abstract. Overly dissimilar cases are hard to align; overly similar cases highlight nothing. This mirrors the use of contrasting examples and non-examples in Concept Attainment, where the boundary between instances and non-instances carries the definitional information.

Sequencing and load. Whether to present cases simultaneously (side by side) or successively is an open design question, and the optimal arrangement may depend on learner expertise and working-memory demands — see Cognitive Load Theory and Chunking reduces working memory load.. Simultaneous display requires holding two cases in mind at once, which can overload novices even when the comparison itself would be valuable; successive presentation reduces load but makes alignment harder because the first case must be retrieved from memory.

Single-case limitation and expertise reversal. A single well-chosen case can sometimes suffice for novices, and multiple cases impose additional processing demands that may not pay off when the target concept is simple or the learner is very inexperienced. As with worked examples, an expertise-reversal dynamic is plausible: see Expertise Reversal Effect. More advanced learners, who already possess a schema, may extract little from additional cases and may benefit more from varied practice than from further comparison.

Open questions. The three entries above bear on this folded claim: Gentner et al. (2003) found comparing two cases beat studying the same two cases separately, which supports comparison rather than exposure as the active ingredient, and Rittle-Johnson & Star (2007) is the one mathematics study recorded. No entry compares simultaneous with successive presentation or covers physics, and none tests the single-case limitation for novices. Designers should weight design decisions toward the comparison-support and case-selection conditions described above, which are the moderators most likely to determine whether multiple cases help or merely add load.

Merged from “Analogical Reasoning Improves Transfer” (analogical-reasoning-improves-transfer): Mechanism. Analogical reasoning supports transfer because it directs attention to the deep, relational structure shared by source and target situations rather than to surface features. Comparing two or more cases side by side — a common instructional implementation — makes that shared structure visible and supports schema abstraction, which is what learners carry into novel problems. This aligns with the broader argument that comparing and contrasting cases improve learning and with cognitive flexibility theory's recommendation of multiple, varied cases.

Boundary conditions. The benefit depends on learners actually engaging in the mapping. Surface similarity can mislead: learners may over-extend a source case to targets that look alike but differ structurally, producing negative transfer. Novices also often fail to retrieve relevant source analogs spontaneously, so instruction typically needs to supply the analog, prompt comparison, or activate relevant prior knowledge — see Analogies and prior knowledge activation and Activation. Without such scaffolding, analogy instruction can consume working memory without producing structural insight, a concern consistent with cognitive load theory.

Design implications. Effective implementations tend to (a) supply the source analog rather than hoping learners retrieve one, (b) prompt explicit mapping of relations ("how is this like the last problem, and how is it different?"), and (c) use at least two cases so the shared structure can be abstracted rather than bound to a single context. Self-generated analogies may produce deeper processing for learners who already have relevant knowledge, but risk failure for novices — a moderation pattern consistent with the expertise reversal effect.

Open questions. Two sources are recorded: the Alfieri et al. (2013) meta-analysis of case comparison (d = .50, larger with immediate testing) and Gentner et al. (2003)'s randomised experiments with novice undergraduates learning negotiation. Key moderators still to document: learner expertise, degree of surface vs. structural similarity between cases, whether the analog is provided or self-generated, and domain (well-structured domains such as mathematics vs. ill-structured ones).

Learner Variables¶

• Executive Function — outcome: instruction changes it
• Algebraic Thinking — outcome: instruction changes it
• Mathematical Flexibility — outcome: instruction changes it
• Operations and Place Value — outcome: instruction changes it

Related Claims¶

• Students who self-explain worked examples learn more effectively than those who do not, and the quality of self-explanations relates to explaining deep structure — the general claim this finding bears on, with the studies behind it

• Strategy Instruction Needs Conditional Knowledge — the general claim this finding bears on, with the studies behind it

• Erroneous examples improve conceptual understanding by forcing comparison with correct models. — the general claim this finding bears on, with the studies behind it

• Cognitive flexibility theory: multiple cases support transfer. — multiple varied cases are the raw material for flexible knowledge

• Case-based learning improves exam performance. — cases as instructional units, of which contrasting cases are a structured variant
• Concept attainment. — comparing positive and negative examples to induce a concept is the classic contrasting-cases pattern
• Chunking reduces working memory load. — relevant to the working-memory demands of holding multiple cases in mind simultaneously
• Expertise reversal effect. — comparison scaffolds that help novices may burden more advanced learners
• Cognitive overload degrades learning. — poorly designed comparisons can overload working memory
• Interleaving category examples improves inductive category learning for visual and mathematical materials, but not for expository texts or word categories — related
• PAIR-C scaffolding shows mixed evidence for deep understanding and reduced misconceptions in emergent-phenomena instruction — related
• Learning in multiple contexts and generalizing across them makes knowledge both specific and general, addressing why transfer is difficult — related
• Cognitive flexibility theory: multiple cases support flexible transfer. — the theoretical framework behind multiple-case instruction in ill-structured domains
• Prompting learners to self-explain improves understanding and problem solving on immediate tests by a moderate average amount, with little evidence yet on delayed or classroom gains. — the prompts that make case comparison productive operate through self-explanation
• Case-based learning improves exam performance. — cases as the unit of instruction in professional education
• Worked examples reduce unnecessary search for novices. — worked examples are a single-case alternative whose multi-case extensions raise the same design questions
• Encoding variability across varied example contexts produces decontextualization supporting transfer (review reports DiVesta and Peverly) — related
• Alternative-strategy practices are most effective for improving procedural flexibility, with weaker evidence for conceptual and procedural knowledge — related
• Teaching reflective questioning or using graphical representations improves procedural and conceptual knowledge across diverse populations — related
• Relational understandings of the equals sign are associated with middle grades mathematics success — related
• PS-I fosters learning only if specific design features (contrasting cases or building instruction on student solutions) are implemented — related
• Activation improves learning. — retrieval of relevant prior knowledge supplies the source analog
• Advance organizers improve learning. — organizers can frame new material as analogous to known structures
• Deep understanding is characterized by coherent, interconnected mental schemas that make knowledge transferable — related
• Building extensive, well-organized domain knowledge may enhance functional use of intelligence in novel problems — related
• Providing both a simple and a complex example facilitates transfer better than a single example or an example plus procedures — a narrower finding that bears on this claim
• Lesson variability benefits transfer only in combination with worked-example instruction, as shown by a variability-by-instruction interaction — related