Increasing instruction time in school does increase learning

90 schools and 1,931 fourth-grade students

Danish fourth graders, autumn 2013

PNAS 113(27), 7481-7484. A three-arm cluster randomised trial - two treatments and a control - which makes it the only true randomised test of adding instructional time in this topic. Both treatment arms added THREE HOURS A WEEK of reading and writing instruction for sixteen weeks; one arm came with a detailed expert-developed teaching programme and the other left the extra time entirely to teacher discretion. Outcome is the Danish National Tests, a national adaptive assessment, measured at the end of the sixteen weeks with no follow-up. Read at abstract depth: the effect sizes below were relayed from a full-text extraction rather than verified against verbatim source text, and pnas.org blocks direct fetching, so the record carries that debt.

Three extra hours a week for sixteen weeks raised reading by 0.15 SD (p = 0.02) IN THE ARM WITH NO TEACHING PROGRAMME, and by a non-significant 0.04 SD (p = 0.36) in the arm with the expert-developed programme, though that arm did move a language-comprehension subscale by 0.14 SD (p = 0.03) and reduced behavioural difficulties by 0.55 SD. THE RANKING IS THE INTERESTING PART AND IT RUNS AGAINST EXPECTATION: unstructured extra time beat expertly structured extra time, and the authors conclude that 'a general increase in instruction time is at least as efficient as an expert-developed, detailed teaching program'. They are appropriately cautious about the comparison - 'limited power' prevents distinguishing the two treatment arms statistically - and about generalisation, noting that Denmark already has high baseline instruction time so marginal effects may diminish elsewhere.

Low. Schools were randomly assigned, so child and family characteristics are balanced in expectation, and the outcome is the national adaptive assessment rather than a teacher grade.

Unreplicated. Its magnitude sits at the top of what the quasi-experimental literature implies: three extra weekly hours for 0.15 SD is far more per hour than Lavy's within-pupil estimate, Bietenbeck and Collins's replication range of 0.014-0.037 SD per weekly hour, or Thompson's four-day-week IV of 0.018 SD per weekly hour. Since it is measured at the end of treatment with no follow-up, it is not comparable to the extended-time evaluations that measure a year or more out and find nothing.