I have lost count of how many times I have heard people say that we learn from mistakes. Sometimes it even turns into a slogan: “If you never make mistakes, you never learn.” It sounds appealing, and there is some truth to it. After all, I do not know anyone who has never made a mistake.

But as is often the case in education, there is an important nuance.

A recent study in Learning and Instruction by Xing Li and colleagues examines this idea more closely. Their question is simple: when do mistakes actually help learning, and when do they not? Their conclusion is both interesting and reassuring for those who have followed this line of research for some time. Mistakes can support learning, but not every mistake does.

This is not the first time this topic has appeared here. Last year, I wrote about how students can learn from mistakes in worked examples, so-called erroneous examples. The common thread across these studies is clear: mistakes only become useful when learners actively engage with them, analyse them, and try to understand them.

The new study by Li and colleagues approaches the issue from a slightly different angle. The researchers compare two ways of integrating mistakes into the learning process.

In the first approach, called induced errors, students are asked to answer a question or define a concept before they are shown the correct answer. Because they often lack the necessary knowledge, they make mistakes. Only afterwards do they receive the correct explanation.

The second approach is called deliberate errors. Here, students first see the correct answer and are then asked to deliberately generate a plausible mistake and correct it.

Both approaches fit within a broader body of research showing that errorful learning, learning that involves making mistakes, can be more effective than errorless learning. This involves several mechanisms, including generating answers, activating prior knowledge, and timing feedback.

In the first experiment, Li and colleagues compared these two approaches with a simple baseline condition in which students just copied definitions. The result was not particularly surprising but still important: both error-based approaches led to better retention than simple copying. That makes sense. Copying involves little thinking, and thinking is what supports memory.

Things became more interesting when the researchers looked at the nature of the mistakes students made. Not all mistakes proved equally useful. The crucial factor was the semantic relationship between the mistake and the correct answer. Mistakes that were conceptually close to the correct answer were much more beneficial than those that were far off. This is also in line with this earlier study.

Again, this is not entirely surprising. A nonsensical answer does not stimulate much thinking, whereas an answer that could have been correct forces learners to process the material more deeply.

To explore this further, the researchers conducted a second experiment in which they manipulated students’ prior knowledge. Some participants received examples beforehand that helped them build a conceptual framework for the material they would later learn. This gave them a bit more to work with.

The effect was clear. Students with more prior knowledge produced mistakes that were more closely related to the correct answers. Under these conditions, the approach in which students first tried to generate an answer themselves proved even more effective than the deliberate-error approach.

In other words, mistakes can support learning, but only when they are part of a meaningful attempt to find the correct answer.

This may sound like a small detail, but it is important. When learners have no prior knowledge, their mistakes are often random or irrelevant. In those cases, mistakes contribute little to learning. When learners already have some conceptual framework, their mistakes can activate exactly what is needed to make new information stick.

The study by Li and colleagues aligns well with a broader conclusion from cognitive psychology: learning is rarely about simply making or avoiding mistakes. What matters is what happens cognitively around those mistakes.

A mistake can trigger curiosity, direct attention to feedback, and make knowledge gaps visible. But this only works when learners have enough background knowledge to make sense of the mistake.

For educational purposes, this leads to a fairly simple yet important implication. The idea that students should just try and make mistakes is too simplistic. Productive mistakes usually emerge within a framework of instruction, prior knowledge, and targeted feedback.

Or put differently, mistakes can be a powerful driver of learning, but only when they do not occur in a vacuum.