Pupils tutored one-to-one under mastery conditions were reported to outperform conventionally taught pupils by around two standard deviations.
Bloom framed it as a challenge rather than a finding: tutoring works, and does not scale, so the task is to find group methods that approach it. The framing has outlasted the number, and most citations quote the number.
In practice
- A school buys a tutoring programme on a promise of two standard deviations and measures a third of one.
- A method that works with one graduate student per pupil tells you little about a class of thirty.
- Possible explanations
- One explanation is pacing and feedback: a tutor adjusts to the pupil continuously, which a class cannot. Another is mastery gating, since the pupil does not advance until the material is learned.
- In its favour
- Tutoring does work: meta-analyses find substantial average effects for structured one-to-one tutoring, and the challenge Bloom set, to find group methods that approach it, remains a useful one.
- Against it
- The two-sigma figure came from a small number of short studies run by his own graduate students, and later work generally finds effects well below it, though the number is still quoted as an established result.
- Standing
- The two-sigma figure comes from a small number of short studies with his own graduate students, and later work has generally found smaller effects. The problem it poses is real; the size is not settled.
- Subject
- Learning & Education
- Essay
- What works in learning, and what schools can and cannot change