Self-Regulated Learning¶

Pattern · All patterns
Evidence · 73 claims (25 for, 44 mixed, 4 against) · 64 studies (23 causal, 12 review, 10 associational, 9 quant-synthesis, 8 theoretical, 1 qualitative, 1 design), q1–q4 · 16 of 64 report an effect size · 66 claims rest on one study

Description and scope¶

A reusable policy for making planning, monitoring against a criterion and choosing a next action explicit within subject-matter work, then reading each response before reducing support. It instantiates the conditional principle. The study configurations below (prompts in computer-based environments, school training programmes, web-based training with a diary and peer feedback, modelled self-assessment) are evidence; the response-dependent policy is an untested design proposal. Use it to gather better evidence about a learner's regulation, not to certify a trait called "self-regulated".

Inputs: establish the design brief¶

Field	Record or elicit
Learner and starting observation	Task-specific prior knowledge (a short content probe); a prediction or self-rating set beside a scored result; whether the learner already plans or reviews unprompted; age band and setting. Preserve unknowns.
Context and activities	Subject task the regulation is attached to; `classroom`, `online-self-paced` or other setting; time available across the unit; who delivers any training; prompts, diary, rubric, peer groups and feedback actually available; what the learner can choose (chapters, order, pace, help).
Learner-valued goal	Ask what the learner wants from this unit and why. Record disagreement with the designer objective; negotiate rather than assume agreement.
Designer objective	Which capability matters: subject performance, accurate self-assessment, a planning habit, or independent regulation without prompts. Say which supports are permitted at the target.
Outcome	Instrument for each objective (subject test aligned with what was regulated; calibration of self-ratings against scores; behaviour traces such as logged time or revisions; questionnaire, labelled as self-report), the local criterion, and the horizon (`immediate`, end of unit, `delayed`).

A request to "make learners more independent" does not specify these inputs. Ask for them before prescribing a prompt schedule, a diary or a fading point. Check that the criterion learners monitor against is itself correct and understandable; monitoring against an unclear rubric produces confident error.

Sequence and conditional policy¶

Elicit briefly. Before support, ask for a short plan for the coming task and a prediction of performance; then score the task. Record the gap between prediction and score, and whether a plan was written at all. These probes are also learning events; record them.
Model the regulation on this task. Show a worked example of someone setting a goal, checking work against the criterion, rating it and choosing a next action, in the subject the learner is studying rather than as a separate study-skills session. Name the strategy and say when and why it helps; classroom observation finds this said in only a quarter of strategy promotions (see the teacher row below), and whether adding it improves uptake is untested in the wiki. Model all three phases (plan, monitor, review) rather than one: in a motor-skill experiment, more phases trained gave better scores.
Prompt at natural pauses. Use short, task-specific planning, monitoring and evaluation prompts before, between and after task segments, paired with feedback on the task where possible. Offer a named set of next actions (review, try a different method, ask for help, move on) so that monitoring has something to act on.
Choose the next activity provisionally. Use the table below. Repeat matched observations where feasible; prior knowledge, task difficulty and the prompts themselves confound a single reading.
Reobserve, then reduce support. Score a new representative task with fewer prompts at the agreed horizon, and compare self-ratings with the score again. Reduce prompts only after the learner's choices and calibration hold without them; keep unresolved explanations on the record.

Observation	Discriminating probe	Proposed action, with its uncertainty
Large overestimate of own score	Ask for a prediction after a retrieval attempt rather than after rereading; have the learner score a worked exemplar with the rubric; run a short content probe.	If calibration improves after retrieval or exemplar scoring, keep that step before every self-rating. If the content probe is weak, treat it as a knowledge gap first and do not rely on self-assessment yet. Neither branch is a validated diagnosis.
Accurate rating, next choice does not change	Present a fixed menu of next actions and ask why the learner chose one; ask what they are aiming for.	If no option seems usable, teach one control strategy explicitly. If the goal is not valued, negotiate the task before adding prompts. More monitoring prompts alone are unlikely to help: in one university trial a monitoring exercise alone did not raise exam scores, and monitoring with a regulation strategy did [+W].
Prompts and diary completed, subject scores flat	Compare entries with traces (time, revisions, chapters chosen); score the chapters the learner chose to focus on separately from the overall test.	If focus-area scores rise and the overall score does not, the outcome may be misaligned, not the regulation. If traces do not change either, replace generic prompts with task-specific ones or add feedback on the plan.
Uses supports only when present	Remove one prompt type on a matched task and observe whether planning or checking still occurs.	Fade one prompt at a time and reobserve. Keep the prompt if removal costs performance at the target and permanent prompting is acceptable to the objective.
Little participation or abandons plans	Ask about purpose, time and barriers; supply the plan and see whether execution follows.	If execution follows a supplied plan, provide plans and teach planning later. If not, the barrier may be content, load or motivation rather than planning skill.

Choosing configurations from evidence¶

For learning in a computer-based environment, task-specific metacognitive prompts [+M] are a candidate configuration: a meta-analysis against no-prompt controls reported g = 0.40 on learning outcomes, varying with feedback pairing, task specificity and adaptivity. It does not set a schedule, and it does not show effects outlast the prompts; one reviewed study found lower-prior-knowledge learners needed training before prompts helped. A separate meta-analysis of SRL scaffolds in computer-based environments [+W] (29 controlled studies, 2,648 learners) pooled 0.438 on academic performance, with large heterogeneity. Prompts may not work through the route assumed: in one experiment prompts raised self-efficacy within the prompted group but left no between-group difference [~W], and did not change strategy use or note quality [-W]. Check the product a prompt asks for rather than assuming it was acted on.
For school learners, SRL training programmes [+M] have a pooled average of 0.69 across 84 studies, but over performance, strategy use and motivation together, and larger when researchers delivered them. Do not forecast a teacher-delivered unit's achievement gain from it; in online settings support helped only when it was used. In higher education, a meta-analysis found SRL interventions improved both SRL activity and achievement [+W], with SRL activity partly mediating the achievement effect and most study and intervention moderators explaining little; its authors' suggestion that task motivation and time on task moderate effectiveness [~W] was not tested there. In programming courses, regulation interventions averaged 0.59 [+W] (844 students, CI 0.28–0.89, effect-size metric not labelled).
Pair monitoring with a way to act on it. In a 2×2 university trial of an online tool, a monitoring-and-regulation strategy raised exam scores while a monitoring exercise alone did not [+W] (means 6.78 against 6.35, controlling for GPA; no standardised effect printed). This is the basis for step 3's menu of next actions. In an expert-novice line of work on Zimmerman's cyclical model, training more phases of the cycle gave better free-throw scores [+W] (50 college students, linear trend, no effect size), which favours modelling planning, monitoring and review together.
Before relying on self-ratings, modelling self-assessment [~M] is a candidate step: in a randomized experiment with 80 pre-university students on genetics problems, modelling examples improved self-assessment accuracy (ηp² = .10). It shows calibration can be trained; it does not show the trained learners then learned more.
A diary alone [~M] showed no detected pre–post gain on any measured outcome in a randomized four-group online mathematics preparation trial; equivalence was not tested, so this is no detected gain, not no effect. Not yet checked against its sources.
In the same trial, training plus a diary [~M] raised SRL knowledge, self-reported planning and metacognition, and self-efficacy, but not mathematics scores (checked by the judge: all 2 entries pass, full text); across groups the overall mathematics score missed significance while a score on self-chosen chapters did not [~M] (checked by the judge: all 2 entries pass, full text). Choose the outcome instrument before choosing the configuration: a broad test may not register what learners chose to regulate.

Do not rank these configurations by their effect labels: the comparators, learners, outcomes and horizons differ, and the 0.69 and 0.40 figures are not on one outcome. A meta-analysis reporting d = 0.859 for SRL strategies across 21 Turkish studies [+W] pools nine intervention studies with twelve correlational ones, so it is not an intervention effect; its moderator tests found no significant difference by school level [~W] or by subject [~W] on 3–8 studies per subgroup, which is no detected difference, not evidence that age or subject does not matter.

Fitting the design to a situation¶

Ask for three facts when the brief omits them: the setting (taught classroom, or online-self-paced with no instructor), the learners' prior knowledge of this content, and the outcome and horizon that count. Each row changes one step of the sequence above.

If the brief says…	Then change…	Basis
`online-self-paced`, no instructor to model the steps	Deliver step 2 as a short web-based training on the regulation steps, and pair it with a daily diary; do not run the diary alone	Training plus a diary raised SRL knowledge and self-efficacy [~M]; A diary alone gave no significant gains [~M]
A cohort that can work in small groups	Add peer-feedback groups to step 3, where learners comment on each other's plans and diaries	Peer feedback groups added self-motivation, volition and reflection gains [+M]
A computer-based environment that can check what learners do with a prompt	In step 3, prefer prompts that ask for a product (revise your summary, review your notes, set a subgoal) and check the product; prompts that only move learners through pages showed no link to learning, and after an add-subgoal prompt, learners who kept attending to the content gained more than those who complied. In a personalised-scaffold study, carrying out suggested review actions (review annotations, review the rubric) went with higher essay scores and note-taking after the last scaffold with lower. All are correlations of compliance with gains in small samples (7–28 learners), not experiments	Review-notes prompt compliance predicts gains [+W]; Revise-summary compliance predicts gains [+W]; Suggest-subgoal compliance predicts gains [+W]; Four explicit SRL prompts showed no learning effect [~W]; No learning effect for page-navigation monitoring prompts [~M]; Add-subgoal compliance may not help [~M]; Compliance with suggested actions and essay scores, mixed [~W]
Learners have low prior knowledge of the content	In step 1, run the content probe first and keep the plan short or supplied; in step 2, model self-assessment on a worked example before asking learners to rate themselves; in step 3, prefer prompts after a segment (what went well, what to change) to prompts before it. In one experiment with apprentices, before-task prompts lowered self-efficacy growth for those with less prior knowledge and after-task prompts raised it; self-efficacy, not learning, was the outcome	Self-assessment accuracy is low without training [~M]; Prior knowledge is needed for accurate self-assessment [~M]; Prior knowledge moderates prompt direction [~W]; Low-prior-knowledge students struggle to self-regulate [~W] (second-hand)
Learners judge how well they understood a text or solved a problem, children included	Before the judgement, have them score their own answer against a key; generating keywords or a summary first did not sharpen sixth-graders' judgements. Expect less gain for the lowest and highest performers and on harder tasks	Self-scoring improves regulation accuracy, not on division for low and high performers [~W]; Sixth-graders' monitoring accuracy low, no cue-prompt effect [~W]
The materials give little guidance (minimal worked examples, discovery or inquiry)	In step 1, record whether the learner plans and checks unprompted, and add steps 2–3 for those who do not; outcomes under minimal guidance tracked learners' own regulation skill, and the instructional approach alone did not change how much learners monitored or reflected	Self-regulation predicted outcomes only with minimal worked examples [~W]; No approach differences in monitoring or reflecting [~W]
Classroom teachers deliver the regulation teaching	In step 2, have the teacher name the strategy, model it and say when and why it helps; observe some lessons to see whether that happened. Unprompted, teachers rarely make strategies explicit or mention their benefit, and see SRL as desirable but hard to fit in; professional development raised observed modelling in one programme, raised explicitness in a simulation-based one, and left explicit instruction rare in another. Expect smaller effects than researcher-delivered trials	Teachers provide little explicit SRL support [+W]; Benefit of strategy use rarely mentioned [~W]; Teachers see SRL as desirable but not feasible [~W]; SRI teachers modelled and evaluated strategies more [+W]; SRL-AIDE raised explicitness, not duration [~W]; Explicit SRL instruction rare in both conditions [~W]; Low time and fidelity for predicting and purpose setting [~W]
Learners regulate as a group on a long project, online or face to face	Expect explicit group regulation mainly when the group meets a challenge; put planning and monitoring prompts at those points (a stuck problem, a disagreement), and prefer a collaboration script the group can adapt	Explicit group regulation infrequent and challenge-driven [~W]; Challenging situations drive plan–perform switching [~W]; Only the adaptable script improved monitoring [~W]
Learners are acquiring a new procedure	In step 1, have learners plan around process goals (the strategy steps), and move to outcome goals once the steps are secure	Process goals outperform outcome goals for novices [+M]
Learners rate their work after rereading or after a fluent explanation	Ask for the self-rating after a retrieval attempt or after scoring an exemplar, not straight after study; fluent presentation raises confidence without raising learning	Fluent illusions mislead self-assessment [~M]
The outcome is a broad end-of-course test over a few weeks	In step 5, score the learner's chosen focus areas and the broad test separately, and keep a behaviour trace beside any questionnaire	Overall mathematics score marginal, focus score significant [~M]
Learners use a generative AI assistant during the task	Add a step-3 prompt before learners accept AI output (confidence, how would you check it), and feedback on their own AI use	Reflection reduces acceptance of incorrect AI advice [+M]; Metacognitive feedback reduces answer offloading [+M]; Advice feedback reduced reminder bias, ranking feedback changed beliefs only [~W] (review); As reliance on AI rises, error detection falls and confidence does not [~W] (second-hand)
The AI tool shows its reasoning traces to learners	Do not treat reading the traces as self-evaluation; keep a scored check against which learners compare their own rating, and do not hand that check back to the AI: the learner still has to accept or reject what it returns	No trace format supported calibrated self-evaluation [~W]; Delegating the terminal evaluation is self-defeating [+W] (argument, no data)
The unit is long and prompts would run throughout	In step 5, remove one prompt type at a time and reobserve; repeated prompts may become routine rather than build a habit	Repeated reflective prompts may habituate [~W]; untested proposal for the order of removal
A single session, not a unit	Keep steps 1 and 3 only (a plan, a prediction, feedback on the gap); skip fading, which needs repeated observation, and do not read a change in regulation within the session as learning	untested proposal; Within-session SRL change not associated with post-test [~W]

The evidence by question¶

Does teaching and prompting self-regulation improve achievement over ordinary instruction? Self-regulated learning improves achievement [+M]
Do planning, monitoring and evaluation prompts help, and which ones? Metacognitive prompts improve learning [+M]
Can learners' own monitoring be trusted, and can it be trained? Self Assessment Accuracy Is Low Without Training [~M]; Self-scoring improves regulation accuracy [~W]
Why make monitoring explicit at all? The theoretical account: self-monitoring is the process through which learners notice gaps and adjust strategy [+M] (a review and a theoretical synthesis, no effect size); and, second-hand, learners often fail to regulate in computer-based environments [~W].
Does monitoring help without a way to act on it? Monitoring with a regulation strategy raised exam scores; monitoring alone did not [+W]
Does it survive delivery by ordinary classroom teachers? Not tested directly in the wiki; teachers provide little explicit SRL support [+W], so fidelity has to be observed.

Further evidence, not yet read against this model¶

Re-read against the model on 2026-10-19 (wave 2): the nine claims listed here and the 41 the overnight linker had added under a ## Claims heading were read from their Evidence entries (50 distinct claims; four were already in the model). Thirty-eight moved into the model (the prompt and programme configurations, step 2, the observation table, the computer-based, low-prior-knowledge, AI and single-session rows, four new rows on self-judgement, low-guidance materials, teacher delivery and group regulation, and the evidence by question). The eight below were left here because they do not change a step, a row or a condition.

A randomized controlled trial with 86 K-12 students found a significant gain in self-regulated-learning strategy use from agentic coaching systems [+W]. Read 2026-10-19: a review's one-line report (p < .05, no effect size, no description of the coaching); strategy use, not achievement, was the outcome. Too thin to shape the AI rows.
Students showed significant learning gains on domain knowledge (large effect) and concept maps (medium effect) after a 45-minute SRL-rich learning task [+W]. Read 2026-10-19: a pre–post gain in 44 university students with no comparison group, so it shows learning happened, not that regulation caused it.
Absolute monitoring accuracy showed a mixed, inconsistent group-by-time interaction rather than a stable competitiveness difference [~W]. Read 2026-10-19: monitoring accuracy across eight classroom quizzes by competitive preference; no stable difference and no regulation support tested.
No intervention effect on directly observed on-task self-regulation in the musical play study [-W]. Read 2026-10-19 (not about this page): a musical-play programme for young children, outside this pattern's scope of regulation attached to subject tasks.
No effects of reported planner or reviser scores were found on evaluating own text [-W]. Read 2026-10-19: 20 writers' think-aloud protocols; a self-reported writing style did not predict how often they evaluated their own text. A null on a questionnaire's predictive value, not on any support.
Students use strategies to regulate their own motivation and affect [+W]. Read 2026-10-19: a handbook chapter's description of college students' motivational self-talk and goal-evoking; motivational regulation is a distinct configuration this page scopes out.
Student-reported SRL promotion is not significantly associated with students' perceived cost of using strategies at class level, and only negligibly (r = −.04) at student level [~W]. Read 2026-10-19: a correlational null (167 teachers, 2,785 students), equivalence not tested; it does not change the teacher row.
Teachers' SRL&T module scores improved modestly over the year, with Adam improving more (three to six points) than Ben (five to six) [~W]. Read 2026-10-19: a two-teacher case study of teaching modules; consistent with the teacher row but too local to add a condition.

Illustrative design instance and observation record¶

Illustration, invented, not a tested design. An adult learner in a four-week self-paced statistics refresher says she wants to stop "freezing" on hypothesis-test questions; the designer's objective is a pass on the end-of-course test and a weekly planning habit. Agree that both count and record them separately. Week 1: a short plan and a prediction before a mixed problem set; she predicts 80% and scores 45%, with errors concentrated in test selection. A content probe shows she cannot yet say which test fits which design, so self-rating is not yet relied on; a modelled self-check of a worked solution against a three-point rubric is added before each later self-rating, with planning prompts at the start of each session and a menu of next actions. Week 3: prediction and score are within ten points on test-selection items; one prompt type is removed on a matched set. The end-of-course test and a score on the chapters she chose are reported separately.

Record: task and criterion → supports present (prompts, model, rubric, feedback) → plan and prediction → scored result → candidate interpretations and their basis → discriminating probe → chosen next activity and why → next observation at the stated horizon. Preserve the learner's stated purpose and any change in it. An agent may summarize this record, but must not fabricate missing inputs or assign numerical state probabilities without a calibrated model.

Elements and limits¶

Goal setting, self-monitoring, prompts, metacognitive strategies, self-assessment, rubric, feedback, peer feedback, reflection, daily before-and-after SRL learning diary and fading scaffolding.

This pattern is scoped to regulation attached to subject-matter tasks over a unit. Standalone study-skills courses, motivational interventions, behavioural self-monitoring for on-task conduct and adaptive systems that take regulation decisions for the learner are distinct configurations with their own evidence. The present policy supports observation and design reasoning; its branches, fading rule and the role of the learner's valued goal remain to be tested with actual learners.

Claims¶

Key Sources¶

Zimmerman, B. J. (2002). Becoming a self-regulated learner: An overview. Theory Into Practice, 41(2), 64-70. https://doi.org/10.1207/s15430421tip4102_2
Lallé, S., Mudrick, N., Conati, C., Taub, M., Azevedo, R. (2017). On the Influence on Learning of Student Compliance with Prompts Fostering Self-Regulated Learning. Proceedings of the 10th International Conference on Educational Data Mining. https://educationaldatamining.org/EDM2017/
de Bruin, A. B. H., Kok, E. M., Lobbestael, J., & de Grip, A. (2017). The impact of an online tool for monitoring and regulating learning at university: overconfidence, learning strategy, and personality. Metacognition and Learning, 12, 21–43. https://doi.org/10.1007/s11409-016-9159-5
Jansen, R. S., van Leeuwen, A., Janssen, J., Jak, S., & Kester, L. (2019). Self-regulated learning partially mediates the effect of self-regulated learning interventions on achievement in higher education: A meta-analysis. Educational Research Review, 28, 100292. https://doi.org/10.1016/j.edurev.2019.100292
Zheng, L. (2016). The effectiveness of self-regulated learning scaffolds on academic performance in computer-based learning environments: a meta-analysis. Asia Pacific Education Review. https://doi.org/10.1007/s12564-016-9426-9
Vosniadou, S., Bodner, E., Stephenson, H., Jeffries, D., Lawson, M. J., Darmawan, I. G. N., Kang, S., Graham, L., & Dignath, C. (2024). The promotion of self-regulated learning in the classroom: a theoretical framework and an observation study. Metacognition and Learning. https://doi.org/10.1007/s11409-024-09374-1

Page metadata

Field	Value
`type`	`pattern`
`id`	`self-regulated-learning`
`aliases`	`explicit-versus-inferred-compliance-prompt-taxonomy`
`title`	`Self-Regulated Learning`
`description`	`A reusable plan–monitor–act policy that makes regulation explicit with task-specific prompts, modelled self-assessment and a usable next action, then reads the response before handing regulation to the learner.`
`status`	`review`
`grain_size`	`unit`
`author`	`self-regulated learning tradition`
`canonical`	`True`
`generated`	by `claude/unspecified` on `2026-10-19`
`trust`	unverified — no one has confirmed this page's content
`sources`	2 — `zimmerman-2002`, `lallé-2017`

Frontmatter is the page's machine-readable record — see the Schema & Guide.