Earlier this week, I wrote about the possibility that educational research might be heading for a crisis. Not necessarily a disaster, but perhaps a necessary shock. I received many reactions across the many channels social media now offers. For which, I thank you. One of those reactions pointed me towards a fascinating study that adds an important layer to the debate, particularly relevant for the role of replication in educational research.

In Replication after Collins: An ethnography of current replication studies in psychology, Jonna Brenninkmeijer and colleagues explored what replication research in psychology really looks like today, not from an ivory tower or some lofty vantage point, but by observing what actually happens in labs, meetings and everyday research practice. What they found was strikingly familiar.

Many researchers began with the idea that replication is simple: you take the protocol of an earlier study and repeat it. In practice, that turns out to be impossible. The context, participants, technology, and researcher are different. Even if you try to follow everything “exactly”, it will always be a new performance. As the authors, following Harry Collins, put it, science is less about following recipes and more about learning to cook.

That brings me back to educational research today. We, too, often work with an algorithmic model of science: a carefully written methods section, a pre-registration, and an analysis plan. All of that is valuable, but it sometimes suggests a precision that doesn’t actually exist. Every intervention, every lesson, every school context is different — as I’ve argued many times before. And that’s precisely where both the richness and the complexity of our field lie.

What I found so compelling in Brenninkmeijer and colleagues’ work is how the replication researchers themselves changed over time. They began with the thought: “We’re just repeating what’s already been done.” But they ended up realising: “We’ve mainly learned a lot about how we ourselves do research.” Some even said they could no longer read an article without thinking, “What’s missing here?” Replication turned out not to be a technical exercise, but a form of self-reflection.

Perhaps that’s precisely what educational research needs right now. A period of reflection, not panic. A moment to hold up the mirror and ask: Are we still doing research that fits the complexity of education today? Have we become too reliant on protocols, questionnaires and averages, while the real insights often lie in how context, interaction and culture shape what happens?

But that nuance also carries a risk, as I noticed in some reactions and at several international conferences. It’s tempting for some to conclude that if replication is so tricky, then science itself is relative. That “everything depends on context”, and you might as well trust your gut. That would be a mistake. The fact that perfect replication is challenging, or that context matters, doesn’t mean anything goes. It means we must better understand when and why something works—and under what conditions. The toolkits from the EEF and Leerpunt follow that same reasoning. Science that questions itself is not weak; it is practising its strongest form of self-correction.

So the crisis I mentioned earlier need not be a collapse, but a turning point. Not the end of educational research, but a chance to make it stronger — through:

• Greater attention to the context in which data are produced.
• More openness about the uncertainty in our findings.
• A deeper awareness of the human side of research; of the teacher, the learner, and the researcher alike.

In that sense, we can learn a great deal from the psychologists in that ethnographic study. They discovered that replication is not only a test of reliability but also a mirror for their discipline. Perhaps that’s exactly what we need now, too: a mirror. Not a crisis to fear, but a chance to reinvent ourselves.