Montessori remains one of the world’s most widely discussed educational approaches. For some, it is almost synonymous with high-quality education. For others, it is primarily a collection of appealing ideas supported by limited evidence. That makes every new research synthesis worth paying attention to.

A new preregistered meta-analysis by Samuel León and colleagues sets out to do exactly that. Rather than asking only whether Montessori education works, the authors also examine how trustworthy the existing body of research actually is. To do so, they reviewed no fewer than 198 studies, of which 114 contained sufficient data for a meta-analysis. This is likely the largest synthesis on Montessori education to date. And, as is often the case when such a large body of evidence is brought together, the conclusions are considerably more nuanced than the headlines might suggest.

Impressive effects at first glance

If you only look at the average effect sizes, the results appear encouraging. Across many studies, Montessori education seems to produce substantial benefits for learning.

But, as you have probably guessed, that is not the whole story.

The researchers found that these average effects are accompanied by:

• very large differences between studies;
• clear indications of publication bias, suggesting that studies with positive findings are more likely to be published;
• a literature dominated by relatively small studies reporting comparatively large effects;
• and a noticeable lack of methodological quality.

Once these issues are taken into account, the originally impressive average effects become much less convincing. According to the authors, the large pooled effects are more likely to reflect weaknesses in the research literature than robust evidence of exceptional educational benefits associated with Montessori education.

Very much in line with what Nature already argued in 2017

A comment on social media about this study reminded me that these conclusions are not entirely new.

Back in 2017, Chloë Marshall published an extensive review in npj Science of Learning (Nature). She also concluded that the evidence for Montessori education certainly exists, but that several recurring methodological issues make it difficult to draw firm conclusions. Those issues sound remarkably familiar:

• few high-quality randomized studies;
• small sample sizes;
• considerable variation between Montessori schools;
• uncertainty about how faithfully schools actually implement Montessori principles;
• difficulty identifying which specific elements are responsible for any observed effects.

Marshall argued that the real question is not simply “Does Montessori work?” but rather “Which elements work, for whom, and under what circumstances?” The new meta-analysis largely confirms that conclusion, but now does so by synthesizing nearly 200 studies and applying modern techniques to assess publication bias and methodological quality.

“But didn’t you write something different before…?”

This new paper also means I should partly qualify an earlier blog post of mine. At the time, I discussed a large-scale study of Montessori preschool education that reported positive effects. Don’t get me wrong, that study remains an important and methodologically strong piece of research.

What this new meta-analysis shows, however, is why we should be cautious about treating even a single high-quality study as definitive evidence. It places that study within the context of almost a century of Montessori research and demonstrates that the broader literature is much less robust than the popularity of the approach might suggest.

To be clear, this is not a contradiction.

A well-conducted individual study can identify genuine positive effects, while the research field as a whole is still characterised by publication bias, small samples, and substantial variation in study quality. That is precisely why meta-analyses are so valuable, even if I have noticed growing scepticism towards them again in recent years.

This is about much more than Montessori

Perhaps the most interesting aspect of this review is that it is really about much more than Montessori education.

It illustrates how an entire research field can accumulate positive findings over decades without necessarily producing a stronger overall evidence base.

Among other things, the authors highlight:

• the near absence of preregistration;
• very limited use of open data;
• insufficient reporting of implementation fidelity;
• little information about the reliability of measurement instruments;
• and the large proportion of studies in which researchers themselves were directly involved in delivering the intervention.

These are precisely the reasons why educational research has increasingly embraced open science, preregistration, and greater transparency over the past decade. Still, do know researching such a thing remains difficult!

So… does Montessori work?

As is so often the case in education, the honest answer is less spectacular than many people might hope.

There is evidence that Montessori education can benefit learning in certain contexts. There are also well-conducted individual studies that report positive effects.

However, when all the available evidence is considered together, the overall case is considerably less robust than the popularity of the approach sometimes suggests.

That may ultimately be the most important lesson from this new review. It is not simply about Montessori. It is about the importance of evaluating the quality of the evidence, not just the size of the reported effects.