The studies
Reviewed · 1 min read
Every grade on this site rests on the studies in this table. Each row says what kind of study it was, who was studied, what it was compared with, what it found, and where it falls short. Filter by grade band, subject, design or principle.
How to read the design column, and why a correlation never earns a grade, is explained on How we grade the evidence.
This is the seed of a larger database. The plan is to add every randomized trial of a classroom practice in K–12, with the same columns, so that a reader never has to go to a second site to check a claim. New studies are logged monthly and reviewed quarterly; see Keeping it current.
46 of 46 studies
- Design
- Review or practice guide
- Grade bands
- K–2, 3–5, 6–8, 9–12, adult
- Subjects
- general
- Sample
- 50 classroom experiments, 5,374 students
- Compared with
- Business as usual
- Found
- 57% of effects were medium or large.
- Limits
- Benefits smaller when final test questions were reworded.
- Design
- Randomized trial
- Grade bands
- K–2
- Subjects
- Math
- Sample
- 110 elementary schools, 12 districts, grades 1–2
- Compared with
- Four curricula against each other
- Found
- Math Expressions, Saxon, and SFAW/enVision all outperformed Investigations by end of grade 2.
- Limits
- Early grades; older editions.
- Design
- Meta-analysis
- Grade bands
- K–2, 3–5, 6–8, 9–12, adult
- Subjects
- general
- Sample
- 164 studies
- Compared with
- Unassisted discovery vs explicit teaching; enhanced discovery vs other methods
- Found
- Explicit teaching beat unassisted discovery; guided or enhanced discovery beat other methods.
- Limits
- Includes adult and university studies; "enhanced discovery" is broad.
- Principles
- Design
- Meta-analysis
- Grade bands
- 3–5, 6–8, 9–12, adult
- Subjects
- Math
- Sample
- Random-assignment studies of worked examples in math, most in classrooms
- Compared with
- Problem solving without examples
- Found
- Worked examples improved math performance (g = 0.48), for first learning and for practice.
- Limits
- Effects vary with how examples are paired with practice.
- Principles
- Design
- Comparison study
- Grade bands
- 3–5
- Subjects
- Math
- Sample
- Nearly 6,000 schools, grades 4–5
- Compared with
- Schools using other textbooks
- Found
- Little difference in average math gains by textbook; more variation where texts were used more.
- Limits
- Not randomized; low PD and partial use.
- Design
- Meta-analysis
- Grade bands
- K–2, 3–5
- Subjects
- Reading and ELA
- Sample
- Same studies as Ehri 2001
- Compared with
- Same
- Found
- Smaller but still positive effect, about d = 0.24.
- Limits
- Different coding choices.
- Design
- Comparison study
- Grade bands
- 6–8
- Subjects
- Social studies
- Sample
- 8th grade U.S. history
- Compared with
- Review without testing, or no review
- Found
- After nine months, students who had been tested on the facts remembered more than students who restudied them.
- Limits
- One school; facts only.
- Design
- Meta-analysis
- Grade bands
- adult
- Subjects
- general
- Sample
- 317 experiments
- Compared with
- Massed practice
- Found
- Spacing helps; the longer you need to remember, the longer the best gap between sessions.
- Limits
- Mostly lab, verbal materials.
- Principles
- Design
- Review or practice guide
- Grade bands
- adult
- Subjects
- general
- Sample
- Review of memory experiments, mostly adults
- Compared with
- —
- Found
- Working memory holds about 3 to 5 chunks, not 7 plus or minus 2.
- Limits
- Lab tasks with adults; exact limits for children less precise.
- Principles
- Design
- Review or practice guide
- Grade bands
- 6–8, 9–12, adult
- Subjects
- general
- Sample
- Review of ten study techniques
- Compared with
- —
- Found
- Practice testing and spaced practice rated high; rereading and highlighting rated low.
- Limits
- Ratings, not a single pooled effect.
- Design
- Randomized trial
- Grade bands
- 9–12
- Subjects
- general
- Sample
- 140 English secondary schools, about 25,000 students
- Compared with
- Business as usual
- Found
- About two extra months of progress on overall GCSE scores; very high security rating.
- Limits
- No significant effect on English or math alone.
- Design
- Meta-analysis
- Grade bands
- K–2, 3–5
- Subjects
- Reading and ELA
- Sample
- 38 experiments, 66 comparisons
- Compared with
- Unsystematic or no phonics
- Found
- d = 0.41; 0.55 when started early, 0.27 after first grade; effects lasted after instruction ended.
- Limits
- Mix of experimental and quasi-experimental studies.
- Design
- Trend data
- Grade bands
- K–2
- Subjects
- Reading and ELA
- Sample
- All Year 1 pupils in England
- Compared with
- Earlier years
- Found
- 58% met the standard in 2012, 82% in 2018.
- Limits
- Not causal; schools learn the test.
- Design
- Randomized trial
- Grade bands
- 3–5
- Subjects
- Math
- Sample
- About 290 at-risk 4th graders
- Compared with
- School's usual fraction teaching
- Found
- Better on fraction comparison, number line, calculations, and released NAEP items.
- Limits
- At-risk students only.
- Design
- Randomized trial
- Grade bands
- 3–5
- Subjects
- Math
- Sample
- 277 at-risk 4th graders, 49 classrooms, 14 schools
- Compared with
- Usual teaching
- Found
- Gains on NAEP fraction items and number line.
- Limits
- Researcher-developed program.
- Design
- Randomized trial
- Grade bands
- 3–5
- Subjects
- Math
- Sample
- 213 at-risk 4th graders
- Compared with
- Usual teaching
- Found
- Better understanding and calculation; gains explained by better fraction-size understanding.
- Limits
- Small groups, not whole class.
- Design
- Review or practice guide
- Grade bands
- 3–5
- Subjects
- Math
- Sample
- Five trials with at-risk 4th graders
- Compared with
- —
- Found
- Consistent effects across five years of trials.
- Limits
- Same research group.
- Design
- Lottery study
- Grade bands
- 3–5, 6–8
- Subjects
- Reading and ELA, Social studies, Science
- Sample
- 2,310 applicants to 9 Colorado Core Knowledge charters
- Compared with
- Lottery losers who attended other schools
- Found
- About 16 percentile points higher in reading in grades 3–6; income gap closed at the one low-income school.
- Limits
- Charter schools, so curriculum and school model came together.
- Design
- Randomized trial
- Grade bands
- 9–12
- Subjects
- Math
- Sample
- About 2,600 and 2,700 Chicago 9th and 10th graders, two trials
- Compared with
- Business as usual
- Found
- Math scores up 0.18 to 0.40 SD; effects lasted into later years.
- Limits
- High school math only; one city.
- Design
- Review or practice guide
- Grade bands
- K–2, 3–5
- Subjects
- Reading and ELA
- Sample
- Expert panel review of rigorous studies
- Compared with
- —
- Found
- Strong evidence for sound-letter links and decoding; moderate for daily connected text; minimal for academic language.
- Limits
- Panel judgments.
- Design
- Randomized trial
- Grade bands
- 6–8
- Subjects
- Math
- Sample
- 363 teachers, about 27,600 Virginia middle school students
- Compared with
- No lessons or support
- Found
- Lessons plus support: +0.086 SD; lessons alone smaller and not consistently significant.
- Limits
- Inquiry-oriented lessons; not a test of explicit vs inquiry.
- Design
- Randomized trial
- Grade bands
- K–2
- Subjects
- Reading and ELA, Science, Social studies
- Sample
- 30 schools, nearly 3,000 students
- Compared with
- Business as usual
- Found
- Science reading comprehension up (0.18); smaller summer losses.
- Limits
- Overall district reading score not significant in the WWC review.
- Design
- Randomized trial
- Grade bands
- 3–5
- Subjects
- Science
- Sample
- 112 third and fourth graders
- Compared with
- Direct instruction vs discovery learning of experiment design
- Found
- 77% mastered fair-test design after direct instruction vs 23% by discovery; transfer was equal for those who mastered it.
- Limits
- Single lab-style study; short task.
- Principles
- Design
- Meta-analysis
- Grade bands
- adult
- Subjects
- general
- Sample
- 607 effect sizes
- Compared with
- No feedback
- Found
- Feedback raised performance on average, but over a third of interventions lowered it; feedback about the self hurt most.
- Limits
- Mostly adults and lab tasks.
- Principles
- Design
- Comparison study
- Grade bands
- K–2, 3–5, 6–8
- Subjects
- Reading and ELA, Math
- Sample
- Over 6,800 students across three years
- Compared with
- Students not tutored, often getting other small-group help
- Found
- Reading up 0.04 to 0.09 SD; no average effect in math or grades.
- Limits
- Many sessions before or after school; program changing as it scaled.
- Design
- Randomized trial
- Grade bands
- 3–5
- Subjects
- Math
- Sample
- 225 at-risk 4th graders
- Compared with
- Usual teaching
- Found
- Gains transferred to decimals without decimal instruction.
- Limits
- Same research group.
- Design
- Comparison study
- Grade bands
- 6–8
- Subjects
- Science
- Sample
- 8th grade science, one school
- Compared with
- Content not quizzed
- Found
- 13 to 25% gains on unit exams; benefits lasted to semester and end-of-year exams.
- Limits
- One school.
- Design
- Trend data
- Grade bands
- 3–5
- Subjects
- Reading and ELA, Math
- Sample
- 4th graders, state results
- Compared with
- Other states, demographics-adjusted
- Found
- First in the nation in 4th grade reading and math after adjustment; 49th in reading in 2013.
- Limits
- Not causal; many reforms at once.
- Design
- Meta-analysis
- Grade bands
- K–2, 3–5, 6–8, 9–12
- Subjects
- Reading and ELA, Math
- Sample
- 89 randomized trials
- Compared with
- No tutoring
- Found
- 0.29 SD average; largest with teachers or aides, early grades, 3+ days a week, during school.
- Limits
- Mostly pre-pandemic, smaller programs.
- Pashler et al. 2008 (opens in a new tab)Retired idea
- Design
- Review or practice guide
- Grade bands
- adult
- Subjects
- general
- Sample
- Review of properly designed tests of matching teaching to a style
- Compared with
- —
- Found
- No support for matching teaching to a learning style; several studies contradicted it.
- Limits
- Few studies used the right design.
- Principles
- Design
- Review or practice guide
- Grade bands
- K–2, 3–5, 6–8, 9–12
- Subjects
- general
- Sample
- Review of classroom applications of cognitive science
- Compared with
- —
- Found
- Classroom evidence is thinner than lab evidence; multi-strategy programs tested at scale often showed small effects; little early-years evidence.
- Limits
- Review, not pooled estimates.
- Design
- Lab study
- Grade bands
- 6–8
- Subjects
- Reading and ELA
- Sample
- 64 seventh and eighth graders
- Compared with
- Students with less baseball knowledge
- Found
- Baseball knowledge, more than reading ability, predicted comprehension of a baseball text.
- Limits
- Small; disputed in later critiques.
- Design
- Lab study
- Grade bands
- adult
- Subjects
- general
- Sample
- College students
- Compared with
- Restudying
- Found
- Restudying won at 5 minutes; testing won at 2 days and 1 week. Students felt more confident after rereading.
- Limits
- Lab, adults, short passages.
- Design
- Comparison study
- Grade bands
- 6–8
- Subjects
- Social studies
- Sample
- 6th grade social studies, one school
- Compared with
- Not quizzed, or content shown twice
- Found
- Better chapter and semester exams after quizzing.
- Limits
- One school.
- Design
- Randomized trial
- Grade bands
- 6–8
- Subjects
- Math
- Sample
- 787 seventh graders, 54 classes, 5 Florida schools
- Compared with
- Mostly blocked practice, same problems
- Found
- 61% vs 38% on an unannounced test a month later (d = 0.83). Meets WWC standards without reservations.
- Limits
- One district, one research group; state-test outcomes untested until the WestEd replication reports (through 2027).
- Design
- Review or practice guide
- Grade bands
- 6–8
- Subjects
- Math
- Sample
- 13,505 problems in six textbooks
- Compared with
- —
- Found
- Under 10% of practice problems were mixed.
- Limits
- Six texts.
- Design
- Comparison study
- Grade bands
- 6–8
- Subjects
- Math
- Sample
- 7th graders, 9 weeks, one school
- Compared with
- Blocked practice
- Found
- 72% vs 38%, even for dissimilar problem types.
- Limits
- One school.
- Design
- Comparison study
- Grade bands
- 6–8
- Subjects
- Math
- Sample
- 126 seventh graders, 3 months
- Compared with
- Blocked practice, same problems
- Found
- Higher on immediate and delayed tests (d = 0.42 and 0.79).
- Limits
- Within-student design.
- Design
- Correlation
- Grade bands
- 3–5, 9–12
- Subjects
- Math
- Sample
- U.S. and U.K. longitudinal samples
- Compared with
- —
- Found
- 5th grade fraction and division knowledge predicted high school algebra and overall math, after accounting for IQ, reading and family income.
- Limits
- Not causal.
- Design
- Meta-analysis
- Grade bands
- 6–8, 9–12, adult
- Subjects
- Math, Science
- Sample
- 53 studies
- Compared with
- Instruction first
- Found
- Problem solving before instruction helped conceptual understanding, mostly for secondary and university students, less for primary.
- Limits
- Contested; applies to students who already have the basics.
- Principles
- Design
- Review or practice guide
- Grade bands
- K–2, 3–5, 6–8
- Subjects
- Reading and ELA, Math
- Sample
- Review of progress-monitoring studies
- Compared with
- —
- Found
- Achievement improved when teachers used progress data to change instruction; collecting it alone did little.
- Limits
- Mostly special education settings.
- Design
- Meta-analysis
- Grade bands
- K–2, 3–5, 6–8, 9–12
- Subjects
- Reading and ELA, Math, general
- Sample
- 328 studies, almost 4,000 effects
- Compared with
- Other programs
- Found
- Positive effects in reading, math, language and spelling; held after instruction ended; larger with more exposure.
- Limits
- Many studies not randomized.
- Principles
- Design
- Review or practice guide
- Grade bands
- K–2, 3–5
- Subjects
- Reading and ELA
- Sample
- Response to the reanalyses
- Compared with
- —
- Found
- Effects of the reanalyzed size still matter in practice.
- Limits
- Commentary.
- Design
- Lab study
- Grade bands
- 3–5
- Subjects
- Math
- Sample
- 4th graders, prism problems
- Compared with
- Blocked practice
- Found
- 77% vs 38% on a test one day later.
- Limits
- Short delay; lab-like setting.
- Design
- Correlation
- Grade bands
- K–2, 3–5
- Subjects
- Social studies, Reading and ELA
- Sample
- ECLS-K 2011 national data
- Compared with
- Less social studies time
- Found
- More social studies time linked to better reading; more ELA time was not.
- Limits
- Not causal.
- Design
- Meta-analysis
- Grade bands
- K–2, 3–5, 6–8, 9–12, adult
- Subjects
- general
- Sample
- 222 classroom studies, about 48,000 students
- Compared with
- Other learning activities or none
- Found
- Quizzing raised achievement by about half a standard deviation across school levels.
- Limits
- Mixed designs and comparison groups.