Assessment for learning improves achievement¶

Claim · All claims
Evidence · 3 studies · 2 quant-synthesis r2, 1 review r? · q2–q3 · i1 small

Formative assessment — eliciting evidence of student learning and using it to adapt teaching and learning activities — raises achievement relative to instruction without such feedback loops.

Subclaims¶

q3 i1 A meta-analysis of US K-12 formative assessment interventions with control conditions finds a small positive overall effect on learning (d = .29), larger where students are supported in self-assessment (d = .61). → Lee et al. 2020

q3 i1 A K-12 meta-analysis estimates an overall effect of about .20, below the .40–.70 often claimed. A 2015 correction changes the science subgroup to .19 (95% CI .06–.31) after correcting subject coding and excluding an inappropriate comparator; the overall conclusion remains about .20. → Kingston and Nash 2011

q2 i? The founding narrative review concludes that innovations strengthening the frequent feedback students receive about their learning yield substantial learning gains. → Black and Wiliam 1998

Evidence¶

Lee et al. 2020¶

Lee, H., Chung, H. Q., Zhang, Y., Abedi, J., & Warschauer, M. (2020). The effectiveness and features of formative assessment in US K-12 education: A systematic review. Applied Measurement in Education, 33(2), 124–140. doi:10.1080/08957347.2020.1732383

q3 · systematic review with meta-analysis · i1 · small effect, d=.29 · n=33 studies (126 effect sizes) · quant-synthesis · r?

A systematic review and meta-analysis of formative assessment interventions in US K-12 schools, restricted to studies with a control condition: 126 effect sizes from 33 studies representing 25 research projects. The overall effect on student learning was small and positive (d = .29), with reported benefits in mathematics (d = .34), literacy (d = .33) and arts (d = .29). Meta-regression suggested larger effects when interventions supported student-initiated self-assessment (d = .61), used formal formative evidence such as written feedback on quizzes (d = .40), and ran on a medium cycle within or between units (d = .52). These feature estimates compare study configurations in a synthesis; they are not direct randomized comparisons of the three features, are not additive, and do not predict an effect for a new course. The source abstract does not specify a common post-intervention measurement horizon.

Kingston and Nash 2011¶

Kingston, N., & Nash, B. (2011). Formative assessment: A meta-analysis and a call for research. Educational Measurement: Issues and Practice, 30(4), 28–37. doi:10.1111/j.1745-3992.2011.00220.x

q3 · meta-analysis (random effects) · i1 · small effect, weighted mean ES=.20 · n=13 studies (42 effect sizes) · quant-synthesis · r2

The authors screened more than 300 K-12 studies of formative assessment and found that only 13 reported enough to compute effect sizes, many of the rest having severely flawed designs. A random-effects model over 42 independent effect sizes gave a weighted mean of .20 (median .25), against the .40–.70 often cited. Implementations built on professional development (.30) or computer-based formative systems (.28) had larger subgroup estimates than other approaches; this does not isolate the approach as a cause. The source abstract does not establish one common assessment horizon or uniform comparator for all included studies.

Published correction to the content-area analysis: Kingston and Nash (2015), Erratum reports that eight science effects were classified as reading effects and that an effect comparing two formative-assessment approaches should have been excluded. The corrected content-area estimates are English language arts .32 (4 effects, 95% CI .30–.34), mathematics .17 (19 effects, CI .14–.20), and science .19 (17 effects, CI .06–.31). The originally reported science .09 and its zero-crossing interval are superseded. The correction says the overall estimate remains about .20 and calls for better research. Unequal subgroup bases and study differences prevent treating the contrast as a randomized choice of subject or a local effect forecast.

Black and Wiliam 1998¶

Black, P., & Wiliam, D. (1998). Assessment and classroom learning. Assessment in Education: Principles, Policy & Practice, 5(1), 7–74. doi:10.1080/0969595980050102

q2 · narrative review · i? · no pooled effect size reported in the abstract · n=unreported · review · r?

A narrative review of the classroom formative assessment literature, the source this claim traces to. It concludes that several studies give firm evidence that strengthening frequent feedback to students yields substantial learning gains, and it discusses students' role in self-assessment, teachers' formative strategies, and mastery learning, before developing a theoretical account of feedback. No pooled estimate was computed, which is part of why later meta-analyses contest the size of the effect.

Studies bearing on this claim¶

Does adjusting instruction to evidence of learners' thinking gathered during teaching raise achievement over teaching the plan unchanged? This table gathers the studies behind this claim and the 20 narrower claims that bear on it: 18 distinct studies (4 design, 4 qualitative, 3 quant-synthesis, 3 review, 2 causal, 1 theoretical, 1 associational), q1–q3, 3 of 18 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¶

Comparability of the pooled effects¶

The i1 label is a legacy Cohen-like bin for the reported statistic, not a judgment that d = .20 or .29 is educationally trivial. Kraft's education benchmarks concern causal K–12 interventions on standardized achievement outcomes. The two pooled estimates cannot be placed in that reference class without checking the included measures and designs. Neither mean predicts a transition for an individual learner.

Unknown here does not imply that every primary study omitted the detail. A usable design comparison would extract those study-level fields where reported, keep unlike measures and contrasts separate, and state which dimensions remain unresolved. A local target needs a learner's initial state, specific task and intended capability, observable criterion, instrument and time point. IES guidance documents differing effect distributions for broad standardized versus specialized researcher-developed tests.

The claim traces to Black & Wiliam's influential review, which argued that formative assessment practices produce some of the largest learning gains in the educational literature, particularly for lower-achieving students. The mechanism is a feedback loop: assessment tasks surface gaps between current and desired performance, feedback closes those gaps, and instruction adjusts accordingly. This aligns with feedback research showing that feedback is most effective when it addresses the task and process rather than the self, and with constructive alignment, which ties assessment tasks to intended outcomes.

Boundary conditions matter. Gains depend on the use of assessment information — assessment that is collected but not acted on, or that primarily serves grading and accountability, shows much weaker effects. Overly frequent or evaluative assessment can shift students toward performance goals and undermine the self-regulated learning that formative assessment is meant to build. Peer- and self-assessment are key vehicles: they require students to internalize quality criteria, which is where much of the achievement benefit appears to originate. Three studies are recorded above: two K-12 meta-analyses (Lee et al. 2020, d = .29; Kingston and Nash 2011, about .20) and Black and Wiliam's narrative review. Of the boundary conditions in this paragraph, only the self-assessment one has support here, as a meta-regression feature estimate (d = .61 in Lee et al. 2020) rather than a direct comparison; none of the entries tests the effect of grading stakes or of assessment that is collected but not acted on.

Open questions. The size of the effect claimed in early reviews has been contested; later syntheses suggest gains are real but more variable than "large effect everywhere" readings imply, depending on subject domain, grade level, and fidelity of implementation. Whether the benefit accrues mainly through improved instruction (teachers adapting) or improved learner behavior (students acting on feedback) remains unresolved, and likely differs across age groups — younger students depend more on teacher adaptation, older students more on self-regulation.

Implementation fidelity. Because the claim concerns a practice rather than a single instructional technique, effect sizes in field studies are highly sensitive to how faithfully the feedback loop is enacted. Professional development that targets teacher questioning, check-ins, and interpretation of student evidence is often the operative variable in successful implementations, not the presence of assessment instruments themselves.

Grading as a contaminant. When formative tasks carry evaluative stakes, students attend to the mark rather than the feedback, and the adaptive function of the loop collapses. Designers should keep elicitation low-stakes and separate it from summative judgment — see Feedback improves learning outcomes.

Related Claims¶

• Feedback improves learning outcomes. — feedback is the operative ingredient in the formative assessment loop
• Self-assessment improves self-regulated learning. — students internalizing criteria drives much of the gain
• Retrieval practice improves long-term retention. — low-stakes quizzing is a common formative assessment vehicle
• Peer assessment improves academic performance by a small average amount relative to no assessment or to teacher assessment — peer assessment operationalizes formative assessment at scale
• Universal school-based SEL programs improve students' behaviour, with a small and, in recent randomised trials, negligible effect on academic achievement — related
• The review reports that formative assessment produces significant and often substantial learning gains, citing Black and Wiliam — related
• Project-based learning improves achievement — related
• Reading and writing abilities share substantial overlap, supporting teaching writing based on reading — related
• High-stakes testing accountability is pervasive in American K-12 schooling while its efficacy remains contested — related
• Incorporating the ACT into high-stakes accountability has not, by itself, produced instructional practices teaching more high-level skills — related
• Schools where students and teachers reported frequent use of standards-aligned practices showed stronger assessment gains than schools with few reports of frequent use — related
• Diagnostic Review revisions contained larger shares of meaning-change, organization, and content/argument work, while Control revisions remained more dependent on surface-level editing — a narrower finding that bears on this claim
• Regular classroom formative assessment can raise student achievement by 0.4–0.7 standard deviations, per studies the chapter cites — related
• Attendees correctly identified AI-generated images in 82% of polling responses, and polling served as a scaffold for discussion rather than a measurement tool — a narrower finding that bears on this claim
• Teachers reported that CRS-and-TEFA use gave them more information about student thinking and that they changed lessons in response to that formative information — a narrower finding that bears on this claim
• Teacher feedback as part of formative assessment has been shown by multiple studies to move student learning forward — a narrower finding that bears on this claim
• Hattie's synthesis identifies lesson study, formative evaluation of teaching, discussion of learning, and feedback as instructional factors with high effect sizes on student learning — related
• Teachers use Microsoft Forms for real-time formative assessment to make in-the-moment instructional adjustments — a narrower finding that bears on this claim
• Strategy Instruction Improves Writing Quality — related
• In one nonequivalent-groups study of pre-school prospective teachers, the group taught with formative assessment embedded in explicit-reflective instruction scored higher on post-test NOS and SK than the group taught with explicit-reflective instruction alone — a narrower finding that bears on this claim
• The Diagnostic Review class gained more writing points across three cycles than the Control class (Hedges' g = 1.12) — a narrower finding that bears on this claim