Competency-Based Learning

Pattern · All patterns
Evidence · 65 claims (22 for, 36 mixed, 7 against) · 65 studies (20 causal, 10 quant-synthesis, 9 associational, 7 review, 7 design, 6 qualitative, 6 theoretical), q1–q4 · 13 of 65 report an effect size · 57 claims rest on one study

Description and scope

A reusable policy for organising a course around a map of stated competencies rather than a fixed calendar: each learner's starting response on each competency is elicited, instruction and support are directed at the competencies not yet shown, pace varies between learners, and a learner advances on a competency when a criterion-referenced performance shows it. It instantiates the Competency-Based Learning & Assessment principle; competency-based assessment concerns how the evidence of competence is gathered and judged.

This page owns the more general, course-level relationship: a map of several competencies, placement, flexible pace across them and progression on evidence. Mastery learning (and its principle) is a narrower configuration that sits inside it: a gate-correct-recheck loop on one cumulative unit at a time, usually with a group schedule. Use the mastery-learning pages for the loop at each competency's check; use this page for how the competencies are mapped, how learners are placed, how far pace may vary and what progression and completion rules apply. Programme governance (credit unbundled from seat time, assessment on demand, a competency transcript) is a further, programme-level object that this page does not describe.

The study configurations below are evidence; the wiki has no claim that tests a competency-based course as a whole against a time-based one. The closest is a quasi-experiment on personalized-learning schools in which competency-based progression was one bundled component among several. The response-dependent policy on this page is therefore an untested design proposal. Use it to collect better evidence about learners' responses, not to certify that a learner "is competent".

What the wiki holds at each level, read 2026-10-17. At the level of one competency's check, the evidence is strong and old: mastery-learning syntheses of controlled evaluations, mostly school and college classrooms from before 1990. At the level of the evidence itself, there is a separate body on simulation-based training with mastery standards in health professions, and on how far raters (human or AI) agree when they score performances. At the level of a whole competency-based programme in higher education, the wiki has one evaluation of online competency-based information-technology programmes in three community colleges (Mathematica, 2014–2016), which is descriptive except for a gateway-course comparison, and one implementation study of personalised-learning schools. For the parts a self-paced online programme adds (mentors, AI tutors, AI role-play partners), the evidence comes from studies of those components in other courses, not from a competency-based programme; each row below says so.

Inputs: establish the design brief

Field	Record or elicit
Learner and starting observation	A criterion-referenced response on each competency in the map (or a placement probe), prior experience the learner reports, earlier credentials, errors and their kind. Use expertise (novice, intermediate, advanced, mixed), age and population_band values where known; record unknowns as unknown, not as novice.
Context and activities	Setting (classroom, online-self-paced, online-instructor-led, workplace-clinical), duration (days-weeks, term-plus), who teaches and who judges evidence, what support is available between attempts (teacher, mentor or coach and their caseload and response time, peer, AI tutor, AI role-play partner, worked examples), whether any AI scores or comments on competency evidence and whether a person confirms it, aids permitted during assessment (and whether AI help is one of them), how far pace may vary within the term and what the administrative calendar requires.
Learner-valued goal	Ask what the learner wants from the course and by when: a credential, a job task, finishing within funded time, skipping what they already know. Record divergence from the competency map, and the learner's tolerance for extra time, as observations.
Designer objective	The competency map: each competency, its knowledge type (procedure, principle, complex-skill, metacognitive-strategy, ...), which competencies depend on which, the representative performance that would show each, the criterion and its justification, and which competencies must be shown together in an integrated task rather than one at a time.
Outcome	Competency-check performance, an independent course or programme examination (general-achievement), integrated or whole-task performance (skill-performance), delayed retention or transfer if targeted, completion (persistence) and time to criterion (learning-time), each with its instrument and horizon.

A request to "move to competency-based learning" or "let students go at their own pace" does not specify these. Ask for them before prescribing a placement test, a criterion, an attempt limit, a degree of self-pacing or an expected gain. Check that each competency's performance samples what the competency names, that a parallel form exists for reassessment, and that some task requires competencies to be integrated.

Sequence and conditional policy

1. Map the competencies. Name each competency, its representative performance, criterion, permitted aids and dependencies. Include at least one integrated task that requires several competencies together. Check that the number of competencies in each block fits the hours it has, and that the people who will teach and judge them have been prepared to plan for them (a default from one reform analysis, not a tested rule; see the situation table). Tell learners the map and the criteria, and show each learner where they stand on it.
2. Elicit a starting response. Place each learner by a criterion-referenced response on the competencies, not by self-report alone. Record what the learner already shows and what remains unknown.
3. Set the pacing and progression rules openly. Decide how far pace may vary, whether learners move as a group with individual correction or individually, what checkpoints or deadlines apply, and what happens when time runs out. Agree these with the learner where the course allows.
4. Teach and support the competencies not yet shown. Direct instruction, practice and support at those competencies. Record what each learner actually received, including how much was self-directed.
5. Gather performance evidence. Score responses against the criterion and record the reasoning or error, not only the score. Use the mastery-learning loop where a competency check falls short: interpret, correct responsively, recheck on a parallel form.
6. Advance on demonstrated competence, with the limit explicit. Advance on criterion, or by the agreed rule when the attempt or time limit is reached, recording any unresolved gap.
7. Check integration and retention. Re-observe competencies in the integrated task and at a stated delayed horizon; a sequence of single-competency passes is not yet evidence of integrated performance.

Observation	Discriminating probe	Proposed action, with its uncertainty
Learner shows a competency on placement	Give a parallel task with a changed relevant condition; compare with the learner's account of where they learned it.	If the parallel task is also met, credit it and move on; if only the familiar form is met, treat recognition or narrow practice as live explanations and give a shorter targeted route rather than full instruction. One placement response does not establish competence.
Falling behind under self-paced progress	Record time and attempts per competency; compare progress on a scheduled checkpoint with progress without one; ask about purpose, barriers and time available.	Add checkpoints, group sessions or a tutor contact, or negotiate the time limit; self-pacing that the learner cannot sustain may cost completion. Neither procrastination nor a missing prerequisite is established by the delay alone.
Passes each competency check, struggles on the integrated task	Compare which competency the failed step draws on with what its check sampled; give a smaller task combining two competencies.	Add integration practice or widen the earlier checks; a narrow check, forgetting and a genuinely new demand of integration each lead to a different action, and one observation does not separate them.
Repeated attempts on one competency without reaching criterion	Change the form of support and observe; probe an earlier competency in the map; check how other learners fare on the same item set.	Offer a different route (explicit strategy instruction, a worked example, one-to-one or small-group help) or revisit the criterion with evidence from other learners; do not keep cycling the same material.
Course-level results no better than the time-based version	Record which components were actually delivered: placement, targeted support, varied pace, reassessment; compare sections or terms that delivered more of them.	Read a null result as possibly a partial implementation before reading it as a failure of the pattern; equally, do not read larger gains where implementation was fuller as proof, since who implements fully is not random.

Choosing configurations from evidence

• The bundle as a whole. Personalized-learning effects and implementation [~M]: a quasi-experiment (controlled-nonrandom, matched virtual comparison groups built from national norm-referenced tests; pre-k-12; about 40 schools, 5,539 students in mathematics and 5,474 in reading, 2014–15) of schools implementing personalized-learning practices that included competency-based progression, flexible pacing, adaptive software and data-driven grouping. One-year effects were small: +0.09 SD in mathematics (significant), +0.07 SD in reading (not significant). Schools reporting fuller implementation and schools in their second year showed larger effects, which the authors call suggestive, not confirmed, since implementation was self-reported and the district subsample small. It cannot separate competency-based progression from the other components, and it does not predict an effect for one course. Not yet checked against its source. Treat the pattern as a bundle whose components need to be delivered and recorded, and expect small average effects rather than large ones.
• Gating each competency on a criterion. Mastery learning programmes, 108 controlled evaluations [+M] report positive effects on examination performance against conventionally paced instruction at college, high-school and upper-elementary levels, apparently stronger for weaker students and varying with the mastery procedure, design and content; possibly more time on instructional tasks; and, for self-paced college programmes, often reduced completion. Abstract only, judged to pass on it; no pooled size recorded. It supports criterion-based progression as a candidate and makes the degree of self-pacing a decision to take with completion in view.
• Group-paced versus self-paced progression. Group-based mastery learning and attrition [~M]: a 1985 conference synthesis reports, from five observational studies, more time-on-task in mastery classes (average effect size .68), and, from one community-college evaluation of over 7,000 students in 77 sections, lower attrition in mastery-taught classes in seven of eight disciplines (average .85), contrasted with findings for personalised self-paced systems. The attrition result rests on one evaluation and the time-on-task result on observation, not assignment; not yet checked against its source. Together with the synthesis above, it is a reason to consider group checkpoints with individual correction rather than unbounded self-pacing, not proof that either is better.
• Letting learners set the pace. Learner pacing and learner control [~M]: in one experiment with primary-school students learning the causes of day and night from an animation, learner-paced versions beat a system-paced one on difficult, high-element-interactivity questions only; a meta-analysis of 18 studies (29 effects) of learner control in educational technology found an overall effect near zero (g = 0.05). Both entries pass the judge on abstracts. Pace control within a lesson is not the same as pace across a course, so this does not test the pattern's flexible progression; it does mean "learners control their pace" should not be expected to help by itself.
• Support between attempts. Contingent versus fixed support [~M]: in one-to-one long-division tutoring with fourth and fifth graders (n = 8 per condition), fully contingent support beat fixed, moderate, partly contingent and no support at immediate and one-month tests; a dynamic-assessment synthesis ranks explicit strategy training above contingent scaffolding; most evidence is one-to-one, and classroom contingency is largely undemonstrated. Not settled against its sources. Make the support between attempts respond to what the response showed, and keep explicit strategy instruction as an alternative.
• Whether the evidence of competence means what the map says. Criterion-referenced tests need construct validation [~W] is a theoretical argument (checked by the judge on full text) that the tasks of a criterion-referenced test must be shown empirically to reflect the competence of interest; it reports no validation study. Progression is only as sound as that evidence, so the competency checks should be checked against an independent outcome before the map is trusted.
• Who gains, and whether it lasts. In the same 1990 meta-analysis, effects were larger for lower-aptitude students [~M] (0.61 against 0.40 SD in 13 evaluations, a difference the authors report as not significant; score spread was smaller in mastery classes in 37 of 52), and gains held at a follow-up about eight weeks later [+M] (0.71 SD in 11 evaluations). A 1985 synthesis of group-based programmes found larger effects in elementary than in high-school or college classes [~M] (.94, .72, .65), so adult and university programmes sit at the lower end of a still positive range. All three are syntheses read from their abstracts or a scanned conference paper. They support criterion-gated progression for weaker starters as well as stronger ones; they say nothing about horizons beyond a few months.
• A whole online competency-based programme in higher education. One evaluation of a three-college consortium running online, competency-based information-technology programmes adapted from Western Governors University's model reports that 35 percent of participants completed a programme of study, with high employment and wage growth [~M], and that completers took on average under two terms to a first industry-certification course and about four terms to a certificate or degree [~M]. Both are descriptive, with no comparison group; the one comparison in the evaluation, on gateway courses, ran the other way (see the situation table). Its implementation report found that the curriculum needed more collaboration and standardisation than community-college programmes usually have [~W]. In the personalised-learning schools, competency-based progression was the component least implemented, held back by grade-level content and testing [~M]. Read together: plan for fast progress by some, slower or no completion by many, and a curriculum and assessment team that works to one standard.
• Performance tasks in simulations and labs as the competency evidence. A meta-analysis of 14 comparisons found simulation-based medical education with deliberate practice, built on mastery standards and feedback on each attempt, beat traditional clinical education [+M] (d = 0.71, mostly procedural skills). A larger meta-analysis of 609 studies found technology-enhanced simulation beat no intervention [+M], which is a weaker comparison. Across 145 higher-education studies, simulation effects were largest for technical performance and smallest for communication [~M] (g = 1.06 against 0.44). A competency gated in a simulation or lab with repeated attempts against a stated standard is the configuration this evidence supports most directly; communication competencies need more attempts and a human-rated check.
• Whether a scored performance is dependable. When people score performances, raters differ in severity [~M] and no rubric item reached full agreement among ten trained raters [~M]. A policy report argues that performance assessments can be scored reliably when designed for it and used with other measures [~W], and states that build performance tasks locally and calibrate rubrics collaboratively [~W]. A gate can also be consistent and still not predict the work: a review reports that teacher licensure exams fail candidates of colour at higher rates without predicting their teaching effectiveness [-M]. Every gate in a competency map therefore needs rater calibration, a second scorer on a sample, and a check that pass rates and later performance do not diverge by group.
• Keeping self-paced learners moving. In one randomised evaluation, proactive one-to-one coaching of college students raised retention at 12 months [+M] (66.4 against 61.4 percent, effect size 0.13) but showed no discernible effect on degree completion at 24 months [~M]. Open-ended planning prompts at the start of online courses raised completion by 29 percent [+M]; self-regulated-learning interventions in online and blended courses had a moderate pooled effect [+M] (d = 0.65, 15 studies); a self-regulation management system was followed by more students meeting the course completion deadline [+W] (88 against 71 percent, one course); and a review finds deadlines and other commitment devices are in demand but do not reliably improve completion [~M]. None of these was run inside a competency-based programme. They support a mentor who reaches out, a plan the learner makes at the start, and milestones the learner commits to, as the mentor-support pattern sets out. Two syntheses, one a review and one theoretical and both read from their abstracts, place self-monitoring against a goal or criterion at the centre of self-regulated learning [+M], as the process through which learners notice a gap and change strategy; they report no effect size and no test inside a competency-based course. They are a reason to show each learner their standing on the map and their next criterion, so the learner as well as the mentor can see a stall, not evidence that doing so raises completion.
• AI tutoring as practice before a competency check. Intelligent tutoring systems raise achievement over non-individualised instruction [+S] (two meta-analyses, g = 0.42 and median 0.66 SD, larger on locally developed tests), but not over individual human tutoring. With a general chatbot, unrestricted GPT-4 access during practice lowered unaided exam performance [-M] (high-school mathematics, preregistered trial). In one large writing experiment practising with AI left learners' self-assessed skill no higher than practising without it [~M], and in one course instructor feedback raised lab scores more than AI feedback [~M] (35 students). Adding human tutors to AI tutoring for the students who needed them raised test-score growth [+M] (grades 5–8, quasi-experimental). So an AI tutor belongs in the practice loop between attempts, with the competency check taken without it and a person reachable; see AI tutoring.
• AI role-play and automated scoring. In one randomised study with medical students, multi-agent AI standardised-patient practice raised scores on an OSCE-aligned examination [+M] (71.8 against 55.6 percent). In one validation, an LLM assessor rated clinical communication with agreement comparable to individual human raters [~M] (ICC 0.75, one rubric). A position statement from a micro-credential report holds that AI is not yet best practice for assessing competency-based work [~W]. Use AI role-play for repeated practice and AI scoring for formative feedback; the wiki has no claim in which an AI's judgement alone certified a competency, so keep a person on that decision. See practice with simulated people.

The earlier page's requirements (defined competencies, valid evidence, flexible pacing or modular progression, feedback and reassessment) and constraints (design complexity, atomising learning into isolated skills, administrative systems built on seat time) are carried into the inputs and sequence above. None of them is tested as a requirement by a claim here; they are proposals from the earlier page. Do not combine these findings into one strength, rank unlike comparators by effect labels, or read "positive effects" as a size. A precise abstention names the missing comparison, instrument or horizon and how to obtain it.

Fitting the design to a situation

Ask for four facts when the brief omits them: how far learners' pace may vary and what deadlines remain, how starting points will be placed, what external test or credential the course answers to, and whether there is an integrated task that requires several competencies together. Where the brief adds mentors or AI, also ask who supports learners between attempts, with what caseload and response time, and who makes the advancement decision.

If the brief says…	Then change…	Basis
Each competency should be gated on demonstrated performance	Put a criterion check, correction and parallel recheck on each competency, as in mastery learning	Mastery learning improves outcomes [+M]
Learners will set their own pace (online-self-paced, adult or college)	Keep checkpoints or deadlines and monitor who stalls; self-paced mastery courses lost more students, and learner control in general showed almost no effect	Self-paced mastery programmes reduce course completion [-M]; Learner pacing helped on complex material only [~M]
A whole school or programme adopting personalised, competency-based progression	Expect small first-year effects that depend on how fully the practices are put in place; monitor implementation, not only outcomes	Personalized learning effects vary with fidelity [~M]
A community-college credential programme	Do not assume completion will rise; one evaluation found slightly lower gateway-course completion for CBE participants than for students in the same colleges' conventional online and campus programmes, and credential-completion differences that varied by college and may reflect who chose CBE (no effect size printed). Compare against a matched group and report by college	CBE gatekeeper completion lower than comparison [-W]
A national or system-wide curriculum recast as competencies or "expected learnings", taught by existing teachers on a fixed timetable	Count the competencies each block expects against its hours before rollout and trim or regroup where they do not fit; give teachers training in planning activities for the competencies, not only an induction session; monitor what is actually taught. Evidence is one analysis of one national secondary-science reform	The reform's science guide packed many expected learnings into few class hours [-W]; monitoring reports found teacher training for the reform insufficient [-W]; competency-based progression was the least implemented component where grade-level content and testing pressed [~M]
A placement test decides where each learner starts	Check that the test is precise around the cut that places learners, not only overall; one placement test lost precision above the 40th percentile	Placement test precision declines above the 40th percentile [~W]
Checks need to be short and frequent	An adaptive mastery test can cut length while reaching the same pass/fail decision for almost all learners	Adaptive mastery testing reduces length and preserves decisions [~W]
Learners will also sit a state test or a time-bound exam	Map which competencies the test covers and protect time for the later ones; slow mastery of early topics can leave later tested topics untaught	Tension between mastery and curricular coverage [-W]; Pacing calendars seen as impediments to mastery [~W]
Support between attempts is given by a tutor	Make the support depend on the learner's last response rather than a fixed sequence of hints	Contingent scaffolding improves learning [~M]
The goal is a complex-skill built from several competencies	Add an integrated whole task and recheck it; a set of single-competency passes is not evidence of integrated performance	Whole-task practice improves transfer, an untested argument [~W]
Group-paced class moving through competencies together	Keep the group pace with individual correction; attrition was lower in group-based mastery classes	Mastery improves engagement and lowers attrition [+W]
Online, asynchronous, self-paced, with mentors available (online-self-paced)	Have the mentor open with a planning contact in which the learner sets milestones per competency; trigger mentor outreach on a missed milestone or repeated failed attempts, not on calendar dates; plan coaching across terms and measure persistence each term. Follow mentor support in self-paced online learning	Coaching raised 12-month retention [+M] but not 24-month degree completion [~M]; planning prompts raised online completion [+M]; commitment devices are mixed [~M]; none tested inside a competency-based programme
Learners start with very different preparation (open admission, adults returning)	Keep the criterion the same for all and vary time and support; place by performance, and give the least-prepared learners the most mentor contact early. Do not assume self-pacing suits everyone: the one programme evaluation's authors proposed offering CBE as one option among several	Effects larger for lower-aptitude students [~M]; CBE gateway-course completion slightly lower [-W]
Adults bring work experience they want credited	Credit a competency only on a performance (a parallel-form task or a portfolio scored against the criterion), not on years of experience or self-report	untested: the wiki has no claim on credit for prior learning or prior-learning assessment
Health professions programme (clinical, procedural and communication competencies)	Gate procedural competencies in simulation with a mastery standard, feedback on each attempt and repetition; give communication competencies more attempts and a blinded rating with a human standardised patient; see practice with simulated people	Simulation with deliberate practice beat traditional clinical education [+M]; simulation effects smallest for communication [~M]
Teacher education (field placement, teaching performance assessment)	Treat the licensure performance assessment as one gate among several; add mentor feedback on evidence from the candidate's own students with revise-and-resubmit; check pass rates by route and group before relying on any single gate	Preservice candidates passed performance assessments more often than interns [~M]; licensure exams failed candidates of colour more without predicting effectiveness [-M]; probing feedback with revise-and-resubmit [~W]
Technology or IT programme aligned to industry certifications	Map competencies to certification objectives, expect a fast group and a long tail, and track gateway-course completion against a comparison; keep instructor or mentor scaffolds and probes in online labs	IT CBE completers averaged under two terms to a first certification course [~M]; 35 percent completed a programme [~M]; online labs judged more effective with instructor scaffolds [+W]
Business programme (analysis, decisions, client or team conversations)	Assess decisions and their justification on a case or simulation scored to a rubric, and practise conversations with an AI or human role-play partner before a rated performance with a different person	deliberate practice raised real-estate licensing pass rates [+W] (self-selected groups); otherwise untested in business programmes in the wiki
AI tutors give practice toward a competency assessment	Use the tutor between attempts, set it to withhold answers, and take the competency check without it; route repeated failure or answer-seeking to a mentor. Follow AI tutoring	Tutoring systems beat non-individualised instruction [+S]; unrestricted GPT-4 lowered unaided exam scores [-M]; human tutors added to AI tutoring raised growth [+M]
AI role-play partners rehearse an interpersonal competency	Allow repeated low-stakes attempts with rubric feedback and a debrief, then certify on a rated performance with a different person	AI standardised-patient practice raised OSCE-aligned scores [+M]; certification by AI role-play alone is untested in the wiki
An AI scores or comments on competency evidence	Use AI scores for formative feedback; for the advancement decision, have a person confirm, double-score a sample, and compare AI and human scores by group	An LLM assessor agreed with human raters about as well as one human [~M]; raters differ in severity [~M]; AI not yet best practice for competency assessment [~W]
A learner reaches the attempt limit without meeting the criterion	Change the support (a mentor session, a worked example, a smaller task) before another attempt, and record any promotion without the criterion as such	In one platform, students on a third attempt met the criterion only 14 percent of the time [~W]

The evidence by question

• Does a competency-based programme, taken as a whole, raise completion and achievement over a time-based one? Gatekeeper course completion rates were slightly lower for CBE participants than for comparison students, and credential completion differences varied by college [-W]
• Which variant: how high a criterion, group-paced or self-paced, and measured how? Mastery learning effects are larger with a higher mastery standard, group-based pacing, local tests, and less feedback for the comparison group [~M]
• Does a pass on the check mean what the criterion says: are the items, cut scores and raters dependable? Criterion-referenced tests require empirically-based construct validation studies [~W]; Raters differed significantly in severity/leniency [~M]; Teacher licensure exams disproportionately fail candidates of colour without predicting teaching effectiveness [-M]
• Do criterion-gated gains last, and for whom? Mastery learning gains persist at follow-up about eight weeks after instruction [+M]; Mastery learning effects are larger for lower-aptitude students [~M]. The wiki has no claim following a competency-based programme's graduates into later study or work against a comparison group.
• How many learners in an online competency-based programme complete, and how fast? 35 percent of participants completed a programme of study [~M]; completers averaged under two terms to a first certification course and about four to a credential [~M]
• Does mentor or coach outreach keep learners in online or self-paced study? Coaching raised 12-month retention [+M]; Coaching showed no discernible effect on degree completion [~M]
• Does practice with an AI tutor carry over to an unaided competency check? Unfettered GPT-4 access during practice harms subsequent unassisted exam performance [-M]; Intelligent tutoring systems improve achievement over non-individualised instruction [+S]
• Does simulation or AI role-play practice with a mastery standard raise a rated performance? Simulation-based education with deliberate practice improves clinical outcomes [+M]; Multi-agent AI standardised-patient training improved OSCE-aligned scores [+M]
• Can an AI score competency evidence as well as a person? An LLM assessor rated clinical communication with agreement comparable to individual human raters [~M]. One validation on one rubric; the wiki has no claim comparing advancement decisions made by AI with those made by people.
• Should prior experience earn credit? The wiki has no claim on credit for prior learning or prior-learning assessment; the placement probe in the sequence above is the untested proposal in its place.

Further evidence, not yet read against this model

Claims this page cited before it was rewritten as a conditional model. They are evidence about the relationship, but none has yet been re-read against the model above; each says how far its own sources have been checked.

Read 2026-10-19 at their evidence entries. Four of these five are now in the model above (noted on each); the one left here is relevant but does not change a decision.

• Self-regulated learning models treat self-monitoring as a core process through which learners notice gaps and adjust strategy [+M] — read 2026-10-19: now in the model, under keeping self-paced learners moving (show each learner their standing on the map); a review and a theoretical synthesis, no effect size
• Gatekeeper course completion rates were slightly lower for CBE participants than for comparison students, and credential completion differences varied by college [-W] — read 2026-10-19: already in the model (the community-college and varied-preparation rows, and the first question); the community-college row now also carries its credential-completion finding and the evaluators' caution about selection
• Experts predict classroom phenomena and student errors far more than novices, who rarely anticipate misalgorithms [+W] — read 2026-10-19: left here. One think-aloud study of teachers (an ERIC report) in which expert teachers anticipated students' likely errors and novices rarely did; it bears on who writes competency checks and interprets wrong answers (step 5), but it is about teacher expertise, not about competency-based progression, and changes no decision
• The SEP Ciencias II work guide packs many expected learnings into few class hours (e.g., 5 learnings in 6 hours for Block 1 and 8 in 5 hours for Block 2), which teachers find hard to cover [-W] — read 2026-10-19: now in the model, in the system-wide curriculum row and the fit-to-time check in step 1
• SEP national monitoring reports indicate insufficient teacher training for the reform: 35% of surveyed teachers requested training in planning and design of didactic activities, and about 30% of directors rated induction-workshop information as poorly adequate or inadequate [-W] — read 2026-10-19: now in the model, in the system-wide curriculum row and step 1

Added 2026-10-17 after a search for competency-based, mastery-based and self-paced evidence. Each was read at its evidence entries; none changes a decision above on its own.

• A three-college consortium implemented online competency-based programmes as planned, meeting most milestones and outcome targets [+W] — the evaluators' own summary of implementation, not an outcome comparison
• The Sinclair-led consortium's competency-based IT programmes, documented at or near baseline [~W] — implementation description, no outcome data
• Mastery learning sections showed higher earned-credit rates than comparable sections at the City Colleges of Chicago [+W] — retrospective institutional records, eight of nine comparisons
• Mastery learning did not reduce the material covered or the pace [~W] — teacher self-report, group-paced classes
• Self-assessing against a rubric improved the quality of student work in two non-randomised studies [+M] — school writing; bears on sharing the criteria with learners
• Performance-based assessments are used to give students more opportunities to demonstrate mastery [+W] — state officials' accounts
• LLMs support competency-based assessment design by generating rubrics, varied items and scenario-based tasks [+W] — a literature review's assertion, no outcome data
• Micro-credential programmes appealed through flexibility and financial savings [~W] — learners' and developers' reasons, rural postsecondary settings
• Student-driven, independent study is the dominant mode in online charter schools, a third offering only self-paced instruction [~W] — prevalence, not effect
• Online attrition in an orientation week was twice that of on-ground first weeks [~M] — continuing-education online classes; bears on the first contact in a self-paced programme
• Career coaches appeared to expand exposure to and completion of community-college health sciences programmes [+W] — associational

Illustrative design instance and observation record

Illustration, invented for this page; not a tested course. A term-long workplace course in electrical maintenance for adult apprentices (adult, adult-workplace, workplace-clinical, term-plus) is mapped to eight competencies, from reading schematics to fault-finding on a live panel, with a final integrated fault-finding task. One apprentice, who has worked as an electrician's helper, wants to qualify before a site placement in ten weeks; the designer's objective is criterion performance on all eight competencies and the integrated task. On placement the apprentice meets the criterion on three competencies; on a parallel task with a different panel layout, two of the three hold and one (isolating a circuit) does not, so that one is kept in the route. The group meets weekly for a checkpoint, with individual work between. In week 6 the apprentice passes every single-competency check but misses a step on the integrated task that draws on schematic reading under time pressure; a smaller two-competency task is added and the integrated task is reattempted in week 8, with the result recorded.

Record: competency, criterion and check form → placement response and parallel-form response → route agreed and pacing rule → instruction and support received, and how much was self-directed → response and reasoning → candidate interpretations and the probe chosen → recheck form, attempt number and response → advancement decision and rule applied → integrated task and delayed re-observation, alongside the learner's stated goal and any change in it. An agent may summarise this record but must not invent missing inputs or assign a probability that the learner "is competent" without a calibrated model.

Elements and limits

Assessment, formative assessment, feedback, reassessment, adaptive mastery learning, learning objectives, rubric, self-assessment, portfolio and practice.

This pattern is scoped to a course organised around a competency map with criterion-referenced evidence. It does not establish that a competency-based course outperforms a time-based one, an optimal degree of self-pacing, a criterion, a number of attempts or a unit size, or a forecast for an individual learner. The risk named by the earlier page, that a competency map atomises a complex performance into isolated skills, is why the sequence includes an integrated task; whether that guards against it remains to be tested. Programme-level competency certification and credit policy need their own configuration and evidence. In the wiki, nothing tests credit for prior learning, an AI's judgement as the basis for advancement, or a competency-based programme's graduates against a comparison group at work.

Claims

• The ELA placement test was most precise for students whose scores would place them into developmental courses, with precision declining above the 40th percentile [~W]
• Teachers viewed district pacing calendars and standardized assessments as impediments when they pressured rushing through material regardless of mastery [~W]
• Self-paced mastery programs (PSI) reduce college course completion, though not by enough to explain their examination advantage [-M]
• El uso de IA para evaluaciones basadas en competencias no es actualmente la mejor práctica [~W]
• Competency-focused internships improved the depth, clarity, and practicality of principal preparation internships compared with hour-accumulation models [+W]

Related Patterns

• Mastery Learning
• Mentor Support in Self-Paced Online Learning: who keeps self-paced learners moving between competency checks
• AI Tutoring: practice between attempts, and how to keep the check unaided
• Practice with Simulated People: rehearsing interpersonal competencies with simulated or AI role-play partners before a rated performance

Examples

• An online, asynchronous, self-paced degree programme in which learners progress through competencies by submitting performance tasks (cases, simulations, labs) scored against rubrics, with a mentor who sets milestones with each learner and reaches out when they stall, an AI tutor for practice between attempts, and AI role-play for rehearsing professional conversations; advancement decisions stay with trained human evaluators. A design scenario using this pattern with the three above, not a tested programme.

• Integrated learning pathways: four principles — district-wide Portrait of a Graduate progressions written as "I can" Practices

• A competency map that lets learners reassess specific standards until they show proficiency.

• Modular technical training where prior experience allows early demonstration and faster progression.
• Promote student agency by letting students decide the pace of their work and whether to complete it independently or collaboratively

Key Sources

• Le, C., Wolfe, R. E., & Steinberg, A. (2014). The past and the promise: Today's competency education movement. Jobs for the Future.
• Sturgis, C., & Casey, K. (2018). Designing for equity: Leveraging competency-based education to ensure all students succeed. CompetencyWorks.
• Berliner, David C. (1988). The Development of Expertise in Pedagogy. American Association of Colleges for Teacher Education. https://eric.ed.gov/?id=ED298122
• Alfonso Cuervo, César Mora y R. García-Salcedo. (2008). Analysis of the Educative Reform in the Secondary School Education in Mexico and its implications in Science II in the new curriculum. http://www.journal.lapen.org.mx
• Kulik, C.-L. C., Kulik, J. A., & Bangert-Drowns, R. L. (1990). Effectiveness of Mastery Learning Programs: A Meta-Analysis. Review of Educational Research, 60(2), 265–299. doi:10.3102/00346543060002265
• Guskey, T. R., & Gates, S. L. (1985). A Synthesis of Research on Group-Based Mastery Learning Programs. AERA Annual Meeting, Chicago. https://eric.ed.gov/?id=ED262088
• Person, A. E., Thomas, J., Bruch, J., Johann, A., & Maestas, N. (2016). Outcomes of Competency-Based Education in Community Colleges: Summative Findings from the Evaluation of a TAACCCT Grant. Mathematica Policy Research. https://www.mathematica.org/publications/outcomes-of-competencybased-education-in-community-colleges-summative-findings-from-the-evaluation
• Person, A. E., Goble, L., Bruch, J., & Mazeika, J. (2015). Implementation of Competency-Based Education in Community Colleges: Findings from the Evaluation of a TAACCCT Grant. Mathematica Policy Research. https://www.mathematica.org/publications/implementation-of-competencybased-education-in-community-colleges-findings-from-the-evaluation
• Pane, J. F., Steiner, E. D., Baird, M. D., & Hamilton, L. S. (2015). Continued Progress: Promising Evidence on Personalized Learning. RAND Corporation. doi:10.7249/rr1365
• McGaghie, W. C., Issenberg, S. B., Cohen, E. R., Barsuk, J. H., & Wayne, D. B. (2011). Does Simulation-Based Medical Education With Deliberate Practice Yield Better Results Than Traditional Clinical Education? Academic Medicine, 86(6), 706–711. doi:10.1097/acm.0b013e318217e119
• Cook, D. A., et al. (2011). Technology-Enhanced Simulation for Health Professions Education: A Systematic Review and Meta-analysis. JAMA, 306(9), 978–988. doi:10.1001/jama.2011.1234
• Chernikova, O., Heitzmann, N., Stadler, M., Holzberger, D., Seidel, T., & Fischer, F. (2020). Simulation-Based Learning in Higher Education: A Meta-Analysis. Review of Educational Research. doi:10.3102/0034654320933544
• What Works Clearinghouse. (2019). Supporting Postsecondary Success intervention report: InsideTrack Coaching. https://ies.ed.gov/ncee/wwc/InterventionReport/696
• Yeomans, M., & Reich, J. (2017). Planning Prompts Increase and Forecast Course Completion in Massive Open Online Courses. doi:10.1145/3027385.3027416
• Ma, W., Adesope, O. O., Nesbit, J. C., & Liu, Q. (2014). Intelligent tutoring systems and learning outcomes: A meta-analysis. Journal of Educational Psychology, 106(4), 901–918. doi:10.1037/a0037123
• Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. PNAS. doi:10.1073/pnas.2422633122
• Yang, L., et al. (2026). Evaluating Scaffolding-Oriented Multi-Agent Large Language Model System for Clinical Interview Training. doi:10.48550/arxiv.2609.10939
• Patrick, S. K. (2024). How preparation predicts teaching performance assessment results in California. Learning Policy Institute. doi:10.54300/544.849
• Carver-Thomas, D. (2018). Diversifying the teaching profession: How to recruit and retain teachers of color. Palo Alto, CA: Learning Policy Institute. doi:10.54300/559.310

Page metadata

Field	Value
type	pattern
id	competency-based-learning
title	Competency-Based Learning
description	A reusable course-level policy that maps a course to stated competencies, places each learner by a criterion-referenced response, lets pace vary and advances on demonstrated competence, with the validity of the evidence, the support offered, time and completion stated as conditions.
status	review
grain_size	course
author	competency-based education tradition
moved_from	principle/competence-is-realistic-goal-for-teacher-preparation, principle/expected-learnings-backbone-of-planning
canonical	True
generated	by claude/unspecified on 2026-10-19
trust	unverified — no one has confirmed this page's content
sources	2 — berliner-1988, alfonso-cuervo-2008

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