Foundations for Success: The Final Report of the National Mathematics Advisory Panel
National Mathematics Advisory Panel · 2008
grade Creviewindependentreplicatednumbers spot-checked
Sample
Consensus panel report. Over 16,000 research publications and policy reports screened by 5 task groups and 3 subcommittees; a strict experimental/rigorous-quasi-experimental filter left small qualifying sets per question (8 studies on teacher-directed vs student-centered, 26 on LD/low-achieving instruction, 11 on calculators, 10 on real-world contexts, 8 on gifted students after the criteria were relaxed). Also draws on public testimony from 110 individuals, written commentary from 160 organizations, and a survey of 743 algebra teachers.
Population
K-12 US; separate analysis for low-achieving/LD students.
Design
Strict evidence filter on the instructional-practice questions. Politically contested (Educational Researcher Dec 2008 special issue attacked the RCT-only filter and panel composition; Benbow & Faulkner rejoined) — but critics produced no counter-RCTs. THIS DOCUMENT IS THE 120-PAGE FINAL REPORT, and it is deliberately unreferenced: "the sections below are not extensively referenced, because the goal of this report is to communicate the Panel's main conclusions without distractions from detail" (p.11). The study-level evidence lives in eight separate task-group and subcommittee reports not held in this archive, so nothing here can be traced to a primary trial from this file alone. Grade lowered B -> C on full-text read: this is a consensus panel's narrative synthesis, not a pooled quantitative synthesis, no effect sizes are reported in the Final Report, the "experimental/rigorous QED only" description does not hold across the whole document (the gifted-students section explicitly relaxed the criteria, and the automaticity conclusions rest on cognitive-science theory plus panel judgment rather than on the filtered trial set). The 26-study explicit-instruction finding is the best-evidenced conclusion in the report; the fact-fluency endorsement is the least.
Key findings
The federal review's two firmest conclusions cut different ways: the global teacher-directed-vs-student-centered question is empirically UNRESOLVED (8 rigorous studies, mixed — "all had limitations and no generalizations can be made"), while explicit instruction for STRUGGLERS is one-sided and repeatedly certified (26 studies, mostly RCTs). Also endorsed fact automaticity and rejected concepts-first sequencing dogma. On full-text read, the three conclusions are NOT equally evidenced and the archive should not treat them as one source of one grade: the 8-study and 26-study findings come from the Panel's strict design filter; the fact-automaticity endorsement is a consensus recommendation derived from cognitive-science theory with no qualifying-study count or effect size attached. The Panel's own summary of its evidence base is bleak — of the thousands of studies screened, "only a small proportion met standards for rigor for the causal questions the Panel was attempting to answer", and it concludes debates of national importance "have devolved into matters of personal opinion rather than scientific evidence".
Genetic confound
Review of experiments; conclusions concern domain skills only.
Replication notes
Struggler finding re-confirmed by Gersten 2009 and WWC 2021 (three federal cycles, same overlapping research community — not three independent looks).
Effects
| Outcome | Metric | Value | Measure | Timing | Vs | Horizon | Class |
|---|---|---|---|---|---|---|---|
| Teacher-directed vs student-centered (8 qualifying studies) | verdict | MIXED/inconclusive — 'all-encompassing recommendations... are not supported by research... should be rescinded' | mixed | end-of-treatment | active-alternative | end-of-treatment | domain-skill |
| Explicit systematic instruction for LD/lowest-third (26 studies) | verdict | 26 high-quality studies, "mostly using randomized control designs": consistently positive effects on computation, word problems, and problems requiring application of mathematics to novel situations; significant positive effects also for scripted Direct Instruction. Panel recommends strugglers receive explicit instruction regularly, while explicitly stating "this kind of instruction should not comprise all the mathematics instruction these students receive". Comparison conditions are not stated in the Final Report, so `baseline` stays unclear. | mixed | end-of-treatment | unclear | end-of-treatment | domain-skill |
| Visual representations for LD/low-achieving students | verdict | "Most of the small number of studies that investigated the use of visual representations yielded nonsignificant effects" — they only produced significant positive effects when bundled with the other components of explicit instruction (p.49). A null inside the 26-study set | mixed | end-of-treatment | unclear | end-of-treatment | domain-skill |
| Practice to automaticity with number facts | panel recommendation (no effect size) | endorsed; conceptual understanding, fluency, and recall are "mutually reinforcing" (rejects concepts-first orthodoxy). Recommendation 11: computational proficiency "is dependent on sufficient and appropriate practice to develop automatic recall of addition and related subtraction facts, and of multiplication and related division facts". IMPORTANT PROVENANCE CAVEAT: this is a Panel judgment grounded in the Learning Processes task group's cognitive-science synthesis (practice -> automaticity -> freed working memory) and in the algebra-readiness analysis. It is NOT an output of the Panel's rigorous-trials filter, and no qualifying-study count or effect size is attached to it anywhere in the Final Report | mixed | n/a | none | not-applicable | domain-skill |
| "Real-world" contexts as an instructional approach (10 studies, 4 meta-analyzed) | verdict | improves performance on assessments involving similar "real-world" problems, but performance on computation, simple word problems and equation solving "is not improved" (p.49-50). A near-transfer-only result — teaching through contexts buys you the contexts | mixed | end-of-treatment | unclear | end-of-treatment | near-transfer |
| Calculators (11 qualifying studies, only one less than 20 years old) | verdict | "limited or no impact of calculators on calculation skills, problem solving, or conceptual development over periods of up to 1 year". Panel cautions that to the degree calculators impede automaticity, computational fluency will suffer, but notes multiyear use from the early grades has never been adequately investigated | mixed | up to 1 year | business-as-usual | under-1yr | domain-skill |
| Computer-assisted drill/practice and tutorial software | verdict | generally positive for performance in specific areas and recommended as a tool for developing automaticity — BUT "one recent large, multisite national study found no significant effects of instructional tutorial (or tutorial and practice) software when implemented under typical conditions of use", and the Panel concludes the available research is insufficient to identify what makes such software work under conventional circumstances. Efficacy-to-effectiveness decay, recorded by the Panel itself | mixed | end-of-treatment | business-as-usual | end-of-treatment | domain-skill |
| Cooperative learning — Team Assisted Individualization (4 studies) | verdict | improves computation skills, but "effects of TAI on conceptual understanding and problem solving were not significant". Suggestive evidence only for peer tutoring on elementary computation | mixed | end-of-treatment | business-as-usual | end-of-treatment | domain-skill |
Cited by
- Explicit vs reform/constructivist math instructionmoderate supportconf: highgc: low
- Math-fact fluency and timed practice (and the anti-timed-test claims)strong supportconf: highgc: low