Education is full of concepts everyone uses but few remember where they came from. Scaffolding, or in plain English: the right support at the right moment, is a perfect example. Since Bruner and Wood (yes, the same Bruner of social constructivism) introduced it in the 1970s, it has become one of the sacred pillars of effective teaching. You’ll find it in handbooks, policy documents and teacher training programmes: give pupils the right help at just the right time, and they’ll learn to think and act more independently.

But how strong is the actual evidence for that?

A team of researchers at amongst others Utrecht University, my colleagues Nienke Smit, Renske de Kleijn, Jelte Wicherts and Janneke van de Pol, decided to find out. Not by proposing a “new model”, but by doing something that still happens far too rarely in education research: a direct replication. They reconstructed the classic 1978 experiment by David Wood and colleagues, often cited as the first empirical proof that scaffolding works.

The Tower of Nottingham, revisited

The original experiment was both simple and brilliant: toddlers were asked to build a wooden pyramid, the now-famous Tower of Nottingham, while an adult guided them using different instructional strategies. The key idea was contingency: give more help when the child fails, less when the child succeeds. Wood found that children who received this kind of support performed much better than peers who only got demonstrations or verbal explanations.

The Utrecht researchers rebuilt the entire experiment, including wooden blocks, protocols, and instructional rules. But this time, they did it with 285 children instead of 32, several tutors instead of one, and under today’s open-science standards: preregistration, video analysis, independently coded data, and all materials freely available on OSF.

And… the effect disappeared

The outcome was as surprising as it was important: they couldn’t replicate the original effect.

Children who received “contingent” tutoring did not outperform those in the other conditions. They didn’t build faster, better, or more independently. The tutors followed the rules just as accurately as in the 1978 study, and all other conditions matched perfectly. Only the spectacular effect of the original study was gone.

That might sound like bad news, but it’s actually good news, and much needed. The findings show that what we’ve taken for granted for nearly fifty years may be too simplistic. “More help after failure, less after success” is a narrow definition of contingency. In reality, it’s something richer: continuously judging why a learner struggles, what they need, and how that moment fits within a broader context of motivation, attention and understanding.

Why this matters

Replications like this push education forward precisely because they bring nuance. The study reminds us that scaffolding can’t be reduced to an algorithm of “more or less help”. Context, individual differences, and patterns of interaction all matter.

That doesn’t mean scaffolding “doesn’t work”; it means it works differently than we thought. Effective tutors don’t simply react to success or failure. They constantly judge, anticipate, and adjust. It’s a form of professional sensitivity that’s hard to capture in an experimental rulebook.

And maybe that’s exactly what makes good teaching both difficult and beautiful.