Mastery learning holds that almost any student can reach a high standard on a topic if given enough time and the right help, so the variable that should change between learners is the time and support they need, not the standard they reach. Benjamin Bloom set out the model in 1968 as "Learning for Mastery", inverting the usual classroom logic: instead of fixing the pace and letting achievement spread out, fix the achievement target and let the pace vary.
The idea rests on John Carroll's 1963 model of school learning, which expressed the degree of learning as a ratio: time actually spent on a task over the time the learner needs to master it. Time spent depends on how much opportunity the learner is given and how long they persevere; time needed depends on aptitude, how readily the learner understands instruction, and the quality of that instruction. Carroll's reframing was quietly radical. Aptitude stops being a ceiling on how much a student can learn and becomes an index of how long they will take. A "slow" learner is not one who cannot reach the standard, only one who needs more time at it. Mastery learning is the instructional system built to deliver that extra time where the ratio falls short.
In practice a unit runs as a loop. The teacher teaches, then gives a short formative test (a check used to guide learning, not to grade it) that locates exactly what each student has and has not grasped. Students who clear a mastery threshold, commonly set around 80 percent, move on. Those who fall short receive corrective work targeted at their specific gaps, then take a parallel re-test. The loop repeats until the learner reaches the threshold. Failure on the first test is treated as a normal, recoverable step in learning rather than a verdict on ability, which is why the diagnostic quality of the check and the precision of the correction matter more than any single grade.
Fred Keller's Personalized System of Instruction, introduced in his 1968 paper "Good-Bye, Teacher...", pushed the same principle toward full self-pacing. Course content is broken into small written modules; a learner studies a unit, takes a unit test, and must score around 90 percent to advance, retaking as needed. Student proctors mark tests on the spot and tutor one-to-one, and lectures are demoted to occasional motivational events rather than the main delivery channel. PSI spread widely through university science teaching in the 1970s; a meta-analysis of PSI courses found students scored roughly 8 percentage points higher on examinations than peers in conventional courses.
Bloom's most cited claim came in his 1984 paper in Educational Researcher, "The 2 Sigma Problem". Comparing conventional teaching, classroom mastery learning, and one-to-one tutoring that used mastery methods, he reported that the average tutored student scored about two standard deviations (two "sigma") above the average of the conventional class, ending up above roughly 98 percent of conventional students. Around 90 percent of tutored students reached a level only the top 20 percent of conventional students reached. The "problem" Bloom posed was practical: tutoring every child is unaffordable, so the research goal is to find group methods that close that two-sigma gap. Mastery learning, in his data, recovered about one of the two sigma.
The two-sigma figure is best read as a provocation, not a settled effect size. It rests on two small University of Chicago dissertations and has not been reproduced at anything like that magnitude in large-scale studies. A central confound undercuts the comparison: the tutoring condition used a 90 percent mastery threshold while the classroom mastery condition used 80 percent, so the tutored students were simply held to a higher bar, which inflates the apparent advantage of tutoring as a method. Beyond the statistics, the model carries heavy logistical costs. Generating diagnostic tests, parallel re-tests, and tailored corrective material for every unit is demanding, and self-paced classes drift out of synchrony as fast and slow learners separate, raising real management and equity problems about who gets the extra time. Effect-size estimates for mastery learning in meta-analyses vary widely and are sensitive to how outcomes are measured and how long mastery standards are held.
The durable, low-cost takeaway is the loop, not the two-sigma headline. Build short diagnostic checks into a unit, define a clear "good enough" standard in advance, and reserve time for targeted re-teaching before moving on rather than after a summative test, where it is too late. Keep corrective work specific to the gaps the check reveals instead of re-teaching the whole unit. Where full self-pacing is impractical, group-based mastery, the whole class loops together on common sticking points, captures much of the benefit without the scheduling chaos of individual pacing. Treating a failed check as information rather than a final grade is the cultural shift that makes the rest work.