Training in text structure improves college students' summaries, but summarizing has not been shown here to improve recall or writing more than rereading or other active reading¶

Claim · All claims
Evidence · 2 studies · 2 causal r? · q2–q3
Currency · newest study 2024 · not yet checked against its sources · not reviewed by a person

Writing summaries of to-be-learned material means identifying main ideas and condensing them into the learner's own words. Two studies are recorded: explicit training in a text's hierarchical structure improved the quality of developmental college students' summaries, with only a marginal effect on delayed recall and no transfer to a reading test (Selinger et al. 1993); and in a multi-document writing task, adults who summarized each source wrote worse integrated essays than adults who reread (McNamara et al. 2024).

Subclaims¶

q2 i? Group-based training in identifying a text's hierarchical structure (thesis, major concepts, supporting details) substantially improves developmental college students' summary quality relative to untrained active-reading practice, with the training effect holding equally across cognitive-ability levels, but the gain does not reliably transfer to delayed recall of the passage or to a standardized reading-comprehension test. → Selinger et al. 1993

q3 i? In a multiple-document integrated-writing task, prompting readers to summarize each source text did not improve integrated-essay quality relative to simply rereading the texts twice — rereaders scored higher on holistic essay quality and on source use/inferencing — though within the group that summarized, higher-quality summaries predicted better essays. → McNamara et al. 2024

Evidence¶

Selinger et al. 1993¶

Selinger, B., Hutson, B., & Fortune, J. (1993). Starting at the top: Using hierarchical structure to train college students to summarize [Conference paper]. Paper presented at the Annual Meeting of the Eastern Educational Research Association, Hilton Head, SC. ERIC ED364296. https://files.eric.ed.gov/fulltext/ED364296.pdf

q2 · randomized experiment, unpublished conference paper (not peer-reviewed) · i? · no standardized effect size reported (R²/percent-variance figures only) · n=58 developmental community-college students · causal · r?

58 community college students in developmental English classes were randomly assigned to a summarization-training group (explicit instruction in identifying central thesis, major concepts and supporting details, ~1.25 hrs/week for 5 weeks) or an active-reading-strategies control group receiving the same amount of vocabulary/comprehension instruction. Both groups were tested on summarization and reading immediately after training, and on delayed recall of the last passage one week later. Training group membership accounted for 40% of the variance in final summarization scores (p < .001), independent of cognitive ability, and 73% of trained students captured at least 70% of the required information in their summaries versus 14% of controls. The training effect on one-week delayed recall was only marginal (p < .10, 12% of variance), and there was no significant transfer to a standardized reading-comprehension test.

McNamara et al. 2024¶

McNamara, D. S., Watanabe, M., Huynh, L., McCarthy, K. S., Allen, L. K., & Magliano, J. P. (2024). Summarizing versus rereading multiple documents. Contemporary Educational Psychology, 76, 102238. doi:10.1016/j.cedpsych.2023.102238

q3 · peer-reviewed randomized experiment · i? · no d/g/r/OR reported (η²/Pillai's trace on a multivariate omnibus test only) · n=295 adults · causal · r?

295 adult participants were randomly assigned to either summarize or reread five short texts on sun exposure and radiation, then wrote an integrated essay drawing on all five texts, which expert raters scored on holistic quality, argumentation, source use/inferencing, language sophistication and organization. A MANCOVA (controlling for topic/science prior knowledge and vocabulary) found a significant overall effect of condition, but in the direction opposite the claim: participants who reread scored higher than those who summarized on holistic essay score and on source use/inferencing, with no group difference on argumentation, language sophistication or organization. Within the summarizing group only, higher average summary quality (rated on inclusion of main points, details, cohesion, etc.) was a significant positive predictor of essay scores over and above prior-knowledge covariates. The authors explicitly frame this as consistent with prior work that summarization aids single-document comprehension but can hinder multi-document integration by biasing readers toward processing each source independently.

Studies bearing on this claim¶

Does writing summaries improve learning more than rereading or other study activities, and does training learners to summarize matter? This table gathers the studies behind this claim and the 5 narrower claims that bear on it: 6 distinct studies (3 causal, 1 quant-synthesis, 1 review, 1 associational), q2–q3, 1 of 6 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¶

The recorded evidence qualifies the claim. Training in text structure improved the quality of summaries (Selinger et al. 1993, an unpublished conference paper), but in a multi-document writing task, summarizing each source led to worse integrated essays than simply rereading (McNamara et al. 2024). The benefit depends on what learners are asked to do with the material afterwards.

Summarization is a generative learning activity: producing a summary asks learners to select the most important information, organize it coherently, and connect it to prior knowledge, which is more demanding than re-reading; whether it is more effective depends on the task, and in McNamara et al. (2024) it was less effective than rereading. It is closely related to other generative strategies such as annotating, and it imposes working-memory demands that must be managed — see Chunking reduces working memory load and Cognitive overload degrades learning.

Moderators the two recorded studies do not yet test (only training, in Selinger et al. 1993, and the multi-document task, in McNamara et al. 2024, are covered):

• Training and scaffolding. Untrained summarizers, especially younger learners, often copy verbatim or fixate on surface features rather than main ideas. Modeling, explicit criteria, and gradual fading of support are typically required before summarization pays off — a constraint consistent with Cognitive load theory.
• Prior knowledge. Learners without sufficient domain knowledge struggle to distinguish main ideas from detail, which limits the strategy's value; summarization tends to benefit learners with at least moderate familiarity with the material.
• Summary quality matters. Benefits accrue when summaries genuinely capture the structure of the material; poorly constructed summaries can reinforce misconceptions or omit critical content.
• Timing and format. Summarizing from memory (a closed-book summary) adds retrieval demands that may strengthen learning further, whereas open-book summarizing emphasizes selection and organization.

Open questions include how summarization compares with other generative strategies (e.g., self-explanation, concept mapping) for the same study time, and how its effectiveness varies across text genres and age groups.

Merged from “Summarization Effective With Training” (summarization-effective-with-training): that page carried the same two studies; its opening paragraph repeated the one at the top of this Discussion.

Why training matters. Effective summarization requires selecting main ideas, deleting redundancy, and paraphrasing — skills that novices do not apply spontaneously. Without instruction, learners tend to copy verbatim text or summarize surface features rather than the underlying structure, which consumes effort without producing the generative processing that makes summarization effective. Training typically involves modeling, explicit criteria for good summaries, and practice with feedback. In the one recorded training study (Selinger et al. 1993), five weeks of group instruction in identifying thesis, major concepts and supporting details raised summary quality, but it did not reliably improve delayed recall or reading-test scores, so better summaries were not shown to mean more learning.

Boundary conditions. The claim's scope is conditional: summarization is a strategy whose effectiveness is mediated by summary quality, and summary quality is mediated by training. Untrained summarization is a condition under which the strategy fails or backfires — it can add rather than reduce load for novices [-M], because the production demands of writing compete with comprehension for limited working-memory resources. Learners with low domain knowledge may struggle to identify what is important even with strategy training, so training may need to be paired with adequate prior knowledge activation [~M]. Summarization also imposes production demands that may overload working memory for novices — see Chunking reduces working memory load and Cognitive overload degrades learning.

Moderators to expect. Plausible moderators of the training effect include learner age and expertise (younger and less knowledgeable learners need more scaffolding), text structure (well-structured texts make main ideas easier to identify), and the fidelity of training (whether learners receive feedback on summary quality rather than merely being told to summarize). None of these is tested by the two recorded studies.

Open questions. The optimal amount, timing, and format of summarization training (e.g., written vs. oral, individual vs. collaborative, summary length constraints) remain under-specified. The page records two studies, one an unpublished conference paper and one a peer-reviewed experiment whose result runs against summarizing; a synthesis of summarization studies, and studies with recall or comprehension outcomes, are still missing.

Learner Variables¶

• Writing Skills — outcome: instruction changes it

Related Claims¶

• Self-questioning improves comprehension — the general claim this finding bears on, with the studies behind it

• Learner highlighting of text gives a small memory benefit but no reliable comprehension benefit, and evidence on other forms of annotation is not yet recorded here — a sibling generative strategy that also requires selecting and processing key information during study.

• Chunking reduces working memory load — summarization works by organizing material into larger, meaningful units.
• Cognitive overload degrades learning — summarization can overload novices if the source material exceeds working-memory capacity.
• Cognitive load theory — the theoretical framework for managing the demands summarization imposes.
• Information at higher levels of an acquired hierarchical organization is recalled better than information at lower levels — related
• Peer Feedback Improves Writing — related
• Peer assessment improves academic performance by a small average amount relative to no assessment or to teacher assessment — related
• Activation improves learning — prior knowledge supports identifying main ideas in a text
• Group-level scaffolding training increases teacher process support and student participation — related
• Summarization, highlighting, the keyword mnemonic, imagery for text, and rereading receive low utility assessments — related
• Well-organized texts following rhetorical schemata are better remembered than disorganized texts, and high-structure information is more likely to be recalled — related
• Group A scored higher than the other groups on Knowledge Quality of collaborative summary notes, with effect sizes of at least 0.7 — related
• Expert revisers' improvements are inconsistent across texts, with only signaling-structure strategies reliably corresponding to improved retention — related
• Strong acquisition tasks explicating the organization produce better internalization than weak tasks asking only for a structured summary — related
• Higher compliance with MetaTutor's revise-summary prompt is associated with larger proportional learning gains — related
• Compliance with MetaTutor's summarize, stay-on-subgoal, move-to-next-subgoal, and open-diagram prompts shows no significant effect on learning gains — related
• Across ten learning techniques, distributed practice and practice testing have the largest effects and underlining and summarization the smallest — a broader claim this one bears on