Annotated bibliography

A growing, honestly incomplete bibliography of the research this project draws on. Each entry notes what it actually found, its methodology at a level relevant to judging confidence, and where it sits in the “academic versus non-academic outcomes” and “supported versus contested” distinctions used throughout this site. Entries are added only once independently checked. This page grows slowly on purpose.

Systematic reviews and meta-analyses

Randolph, J.J., Bryson, A., Menon, L., Henderson, D.K., Kureethara Manuel, A., Michaels, S., Rosenstein, D.L.W., McPherson, W., O’Grady, R. and Lillard, A.S. (2023). “Montessori education’s impact on academic and nonacademic outcomes: A systematic review.” Campbell Systematic Reviews, 19(3), e1330. DOI: 10.1002/cl2.1330.

• Design: Systematic review and meta-analysis, with searches ending in February 2020. Of 2,012 articles found, 32 studies met its inclusion criteria, which required evidence of baseline equivalence between the Montessori and comparison groups, through random assignment or statistical adjustment. 204 effect sizes across 132,249 data points.
• Findings: Composite effect size g = 0.24 for academic outcomes (from 0.06 for social studies to 0.26 for general academic ability; mathematics 0.22, language 0.17) and g = 0.33 for non-academic outcomes (from 0.23 for social skills to 0.41 for children’s inner experience of school; executive function 0.36, creativity 0.26). Effects were larger in randomised than in non-randomised studies, at preschool and elementary than at middle- and high-school level, and in private than in public programmes. The authors found a shortage of small studies favouring traditional education, a possible sign of publication bias, but judged their findings robust in a sensitivity analysis.
• Methodology-quality flag: As of 2026, a later review reaches a sharply different conclusion about the Montessori evidence base (see León et al. 2026, below), and researchers currently disagree about how strong that evidence is. Full author list confirmed via Crossref publisher metadata; figures above from the published abstract. This project notes that the tenth author, Angeline S. Lillard, also led the 2025 lottery-based trial listed below, so the review and one of the researchers whose primary work bears on its question are not fully independent of each other. The 2025 trial itself falls outside the review, whose searches ended before the trial was published.
• Used on: Understanding the Stages, Evidence Status.

Demangeon, A., Claudel-Valentin, S., Aubry, A. and 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. DOI: 10.1016/j.cedpsych.2023.102182.

• Design: Meta-analysis of 33 experimental or quasi-experimental studies comparing Montessori education with other approaches (268 effect sizes; n = 21,670), conducted in North America, Asia and Europe and published between 1991 and 2021, using a three-level model to allow for several effect sizes from the same study.
• Findings: Positive pooled effects in all five fields: cognitive abilities g = 0.17, social skills 0.22, creativity 0.25, motor skills 0.27 and academic achievement 1.10. Moderator analyses found no differences by school level, type of publication or continent.
• Methodology-quality flag: Its academic-achievement effect (g = 1.10) is far larger than its own cognitive effect (0.17) and than the academic composite in Randolph et al. (2023, above; 0.24): the two 2023 syntheses agree on direction but not on size. This project has read the abstract, not the full text, so it cannot say why the estimates differ, and nothing is said here about this review’s checks for bias. Full citation (authors, title, journal, volume, article number, DOI) confirmed via Crossref publisher metadata; figures from the authors’ abstract, read through their deposit in the HAL open archive.
• Used on: Evidence Status.

León, S.P., Lipnevich, A.A. and García-Garrido, M. (2026). “The Montessori Method Under Scrutiny: A Meta-Analytic and Metascientific Review for Greater Transparency in Education Research.” Review of Educational Research, published online 7 July 2026. DOI: 10.3102/00346543261458396.

• Design: Preregistered systematic review and meta-analysis. Identified 198 studies, of which 114 provided enough quantitative data for meta-analysis.
• Findings: Pooled effects on learning outcomes were “nominally positive and conventionally large” in pretest–posttest and posttest-only designs, but the authors judge these estimates not robust: they showed “extreme between-study heterogeneity, substantial small-study effects, and clear indications of publication bias”, and inconsistent reporting of implementation fidelity, low transparency and limited experimental control further constrain how far they generalise. The authors conclude that the large average effects reported in the literature are “largely attributable to structural weaknesses of the evidence base rather than to robust educational impacts”. The review does not conclude that Montessori education has no effect.
• Methodology-quality flag: Its conclusion differs sharply from those of the 2023 syntheses (Randolph et al. 2023 and Demangeon et al. 2023, above), and researchers currently disagree about how strong the Montessori evidence base is. Treat that disagreement as live and unresolved: this review is neither an accepted correction nor a finding to be set aside. Full citation (authors, title, journal, online publication date, DOI) confirmed via Crossref publisher metadata; no volume or issue has been assigned yet. This project has read the published abstract, not the full text, so nothing is said here about its methods beyond the abstract. Two questions remain open: how far its study pool overlaps with that of Randolph et al. (2023), and whether either review’s estimates survive the other’s bias diagnostics. This project has not independently re-run any of these meta-analyses.
• Used on: Understanding the Stages, Evidence Status.

Randomised trials

Lillard, A.S., Loeb, D., Berg, J., Escueta, M., Manship, K., Hauser, A. and Daggett, E.D. (2025). “A national randomized controlled trial of the impact of public Montessori preschool at the end of kindergarten.” Proceedings of the National Academy of Sciences, 122(43), e2506130122. Published online 20 October 2025. DOI: 10.1073/pnas.2506130122.

• Design: Randomised by admission lottery: 588 children who applied at age three to 24 public Montessori schools in the United States, 242 offered a place and 346 not. Intention-to-treat effect sizes (Hedges’ g, controlling for baseline scores and demographics) at the end of each of two preschool years and of kindergarten (ages five to six).
• Findings: At the end of kindergarten, statistically significant effects on reading (0.297), executive function (0.250), short-term memory (0.293) and theory of mind (0.222); not significant: maths (0.172), vocabulary (0.107), working memory (0.160), social problem-solving (−0.055) and persistence (−0.167). The authors report “no notable impacts” at the end of either preschool year (one significant difference at the end of the first year, on an unvalidated persistence task, reversed in the second). A cost analysis estimated that three years of public Montessori from age three to six cost school districts USD 13,127 less per child than traditional programmes, mainly because of higher child-to-teacher ratios in the two preschool years.
• Methodology-quality flag: Consent was 20.9% overall (31% of those offered a place, 17% of the others); kindergarten assessments are missing for 32% of the children offered a place and 42% of the others; only about half of the children offered a place were still in Montessori in kindergarten, so the estimates are of the offer of a place, not of attendance. Public programmes only. The lead author, Angeline S. Lillard, is a co-author of Randolph et al. (2023, above) and the author of a 2026 book on the method, The Montessori Difference (Oxford University Press), not read for this page. Full citation confirmed via Crossref publisher metadata; figures checked against the full text (Europe PMC, PMC12582262); confidence intervals not reproduced here. A 2026 American Educational Research Association conference paper by Daggett, Lillard, Loeb and colleagues (doi:10.3102/2277608) analyses achievement gaps in the same study and has not been read for this page. No published comment on, or correction to, the trial had been found when this project checked in October 2026.
• Used on: Evidence Status.

Descriptive and census data

Debs, M., de Brouwer, J., Murray, A.K., Lawrence, L., Tyne, M. and von der Wehl, C. (2022). “Global Diffusion of Montessori Schools: A Report From the 2022 Global Montessori Census.” Journal of Montessori Research, 8(2), 1–15. DOI: 10.17161/jomr.v8i2.18675.

• Design: Descriptive count built from a survey of national and regional Montessori organisations (81 responses from 59 countries), triangulated with other published estimates and direct inquiries in key countries, and resting on those organisations’ broadest definitions of a Montessori programme. The authors state that several countries may be undercounted, and that the English-language survey, first sent out by AMI, may have missed schools working in local languages.
• Findings: 15,763 Montessori schools worldwide, roughly 9% of them fully government funded; the United States, Thailand, the Netherlands and India have the largest numbers of government-funded or public programmes. The authors note that respondents may have read their funding question differently, and that the line between government schools and privately run schools with some public funding needs further study.
• Note: Mira Debs is also the author of the India adoption study below; the same researcher accounts for two of this bibliography’s strongest entries, which is useful to know when judging how much independent corroboration this site’s international data actually has. Citation confirmed via Crossref publisher metadata; findings checked against the published full text.
• Used on: International Comparative Data.

Named bodies of work not yet individually cited here

Angeline Lillard’s earlier lottery-based studies, which predate both the Randolph et al. (2023) review above (she is its tenth co-author) and the 2025 trial above (she is its lead author), have not yet been individually checked and cited on this page. They are flagged here as the next addition needed, rather than relied upon without attribution.

Biography and history

de Stefano, C. Biography of Maria Montessori, as reviewed by Lucy Hughes-Hallett, “A Complicated Reformer,” The New York Review of Books (9 March 2023).

• Design: Secondary source: this project has read the NYRB review, not the original biography text, so citations drawn from it should eventually be checked against the primary biography.
• Findings: Source for Mario Montessori’s concealed birth, the 1913 reunion, and the family circumstances surrounding both.
• Used on: Maria Montessori: a Fuller Biography.

Debs, M. (2022). “Montessori in India: Adapted, Competing, and Contested Framings, 1915–2021.” History of Education Quarterly, 62(4), 387–417. DOI: 10.1017/heq.2022.25.

• Design: Peer-reviewed historical study.
• Findings: Confirms an orthodox/adapted split in Indian Montessori provision by wealth, including Tarabai Modak’s low-cost material adaptations, and confirms Montessori’s adoption as the preschool curriculum across Tibetan refugee schools in India from 1968.
• Methodology-quality flag: Full citation (author, volume, issue, pages, DOI) confirmed via Crossref publisher metadata. The article’s full text remains behind Cambridge Core’s paywall and has not been read in full by this project; findings here rest on Crossref metadata and the publisher’s listing, not a complete read-through.
• Used on: International Comparative Data.

Nobel Prize nomination archive (nobelprize.org, person id 6383).

• Design: Primary archival record (nomination database).
• Findings: Confirms Montessori was nominated for the Nobel Peace Prize in 1949 (four nominators), 1950 and 1951.
• Methodology-quality flag: This project accessed the archive’s content via indexed search snippets, not a direct page fetch (the site blocks automated requests). Recommend direct manual verification before academic citation.
• Used on: Maria Montessori: a Fuller Biography.

What’s still missing

This bibliography does not yet cover: the US Department of Education / What Works Clearinghouse’s historical assessments of Montessori evidence; primary sources on fidelity variation between AMI- and AMS-trained settings; Lillard’s earlier lottery-based studies, named so far only as a body of work (see above); the full texts of León et al. (2026) and Demangeon et al. (2023), of which this project has read only the abstracts; or dedicated adolescent-programme (Erdkinder) outcome studies, which are genuinely scarce in the literature. These gaps are listed deliberately, so the absence is visible rather than silently implied to be “nothing to report.”

Correction, 11 October 2026: an earlier version of this page described the 132,249 data points in Randolph et al. (2023) as participants and its 2,012 search results as studies screened; called its inclusion criteria a strict bar and a methodological strength, an assessment this project had not checked; described Angeline Lillard as the field’s single most-cited source of lottery-based evidence, which this project had not checked; and described the 2022 Global Montessori Census as a count of verified schools made by direct outreach, with public funding concentrated in three countries, which its published text does not support. See the corrections log.

Correction, 10 October 2026: an earlier version of this page gave the wrong initials for two authors of León et al. (2026), omitted its DOI, and described it as re-examining the same studies as Randolph et al. (2023) and as finding the earlier effects inflated and the earlier reviews’ search strategies insufficiently transparent, which the published abstract does not support. See the corrections log.