Article
Montessori is 100 years old and beloved. The science is thinner than you'd think — and stranger.
The Montessori lottery: what a century of evidence actually shows
The most famous educational method in the world began in a housing-project slum in Rome in 1907, run by a woman who had to fight to become one of Italy's very first female medical doctors. A century later there are thousands of Montessori schools on every continent — and for most of that century there was almost no rigorous evidence that the method worked at all.
Then, in 2006, the most powerful education experiment ever run on Montessori happened by accident. And in the last two years, the evidence base has finally caught up — with a twist that complicates the story its fans tell.
The method that looks like nothing else
Maria Montessori (1870–1952) initially resisted teaching — one of the few careers open to women in late-19th-century Italy — and trained as a doctor instead, working with children with intellectual disabilities. There she concluded that what those children needed was not medical treatment but the right pedagogy. In 1907 she opened her first "Casa dei Bambini" (Children's House) for 3–7-year-olds in a slum district of Rome and kept developing her method through systematic observation and experimentation [1].
Her pedagogy rests on a "dynamic triad of child, teacher and environment": a prepared classroom where children work hands-on with specially designed manipulable materials during a three-hour "work cycle," choosing what to do, where, with whom, and for how long — within class rules, without competition and without extrinsic rewards or punishments [1]. It looks strikingly different from a conventional classroom, which is precisely why people either love it or suspect it's a fad.
The lottery
Here's the part that makes Montessori's evidence unusual: the best study ever done on it used random assignment by lottery. In Milwaukee in the mid-2000s, oversubscribed public Montessori schools admitted children by a computer-drawn random order of rank — a natural experiment. Psychologist Angeline Lillard and statistician Nicole Else-Quest exploited it: children who won the lottery attended Montessori; children who didn't attended other schools. Both groups were then measured [2][3].
Published in Science in 2006, the study found that at age 5, Montessori children outperformed the control group on standardized measures including letter-word reading, word attack (decoding), applied math problems, a card-sorting test of executive function, and a "false belief" test of social cognition [2][3]. At age 12, differences persisted — Montessori children scored higher on a measure of referring to justice, a socially-oriented reasoning measure [2][3].
Random assignment matters: most "school A beats school B" comparisons are confounded — the families who choose a school differ from those who don't. A lottery removes that. Lillard and Else-Quest's design is why Montessori's flagship evidence is cited by even skeptical researchers [4, 5].
But the evidence base is thin
The other half of the story is how little rigorous research exists. In a 2017 review in npj Science of Learning, Chloë Marshall of UCL found that, despite over 100 years of the method, "peer-reviewed evaluations of Montessori education are few and they suffer from a number of methodological limitations" [1]. Most studies were small, non-randomized, or single-school. The 2006 lottery study remained, for years, the exception rather than the rule [1].
Marshall also flags a subtler problem: Montessori's stated goal is the child's optimal development — intellectual, physical, emotional, social — whereas most school systems measure academic attainment. "The outcome measures used to capture effectiveness do not necessarily measure the things that Montessori deemed most important," she writes [1]. If you evaluate Montessori only by test scores, you may miss what the method is aiming at — or, from the other side, forgive it for outcomes it never prioritized.
The meta-analysis era
The last two years changed the picture. In 2023, Demangeon, Claudel-Valentin, Aubry and Tazouti published a meta-analysis of Montessori education's effects on five fields of development and learning in preschool and school-age children [4, 5]. A 2025 research paper in Frontiers in Developmental Psychology sums up the current state: "meta-analyses (Demangeon et al., 2023; Randolph et al., 2023), tightly controlled longitudinal studies (Fleming and Culclasure, 2023), and random lottery studies suggest Montessori" effects across domains [5].
So the honest 2026 summary is: the evidence is no longer "almost nothing." It's a small but converging set of meta-analyses, longitudinal studies, and a genuine lottery experiment pointing the same direction — modestly positive effects, concentrated in areas Montessori was designed to train.
The twist: the delayed benefit
Then came the complication. In December 2025, a five-year longitudinal randomized controlled trial of Montessori education — by Le Diagon, Van der Henst and Prado, published in Scientific Reports — came to the opposite-looking conclusion, titled "Early Montessori education shows delayed benefits for mathematical problem-solving" [6]. Delayed. Not immediate.
The finding — benefits for math problem-solving appearing later, not in the early years — undercuts the "Montessori instantly unlocks kids" narrative as much as it supports a longer-horizon one. It's a genuinely interesting result for anyone deciding whether the method "works": the answer depends on when you look.
Don't over-read the label
One more honesty checkpoint: not every school called "Montessori" actually practices it faithfully. Studies vary wildly in implementation fidelity, and Marshall's review notes that schools may adopt the name with only loose adherence to the method's core elements [1]. A meta-analysis averages over "Montessori-ish" classrooms as much as faithful ones — which means true-fidelity effects could be larger, and label-only effects could be near zero.
Bottom line
The science on Montessori, a century in, is: one real lottery experiment with positive headline findings; a thin but now-converging set of meta-analyses and longitudinal trials; genuine effects in the domains the method targets; and at least one 2025 RCT showing the benefits can arrive late. That's a far better evidence base than almost any other educational fad — and still far less settled than Montessori's most passionate defenders claim. Both sides of that sentence are true, and the honest answer sits between them.
Sources:
- Marshall, C. (2017). Montessori education: a review of the evidence base. npj Science of Learning, 2, 11. https://www.nature.com/articles/s41539-017-0012-7 (fetched 2026-10-13)
- Lillard, A. & Else-Quest, N. (2006). Evaluating Montessori Education. Science, 313(5795), 1893–1894. https://www.science.org/doi/10.1126/science.1132362
- Free full text of the 2006 Science paper (ERIC mirror). https://files.eric.ed.gov/fulltext/ED622858.pdf
- Demangeon, A., Claudel-Valentin, S., Aubry, A., & Tazouti, Y. (2023). A meta-analysis of the effects of Montessori education on five fields of development and learning in preschool and school-age children. Contemporary Educational Psychology, 73, 102182. https://www.sciencedirect.com/science/article/abs/pii/S0361476X2300036X
- Frontiers in Developmental Psychology (2025), original research reviewing the Montessori evidence: meta-analyses (Demangeon et al. 2023; Randolph et al. 2023), longitudinal studies (Fleming & Culclasure 2023), random lottery studies. https://www.frontiersin.org/journals/developmental-psychology/articles/10.3389/fdpys.2025.1546451
- Le Diagon, S., Van der Henst, J.-B., & Prado, J. (2025). Early Montessori education shows delayed benefits for mathematical problem-solving in a 5-year longitudinal randomized controlled trial. Scientific Reports, 15, 43961. https://www.nature.com/articles/s41598-025-27687-2