Learning from Errors

No one likes to mess up, but going to great lengths to avoid errors - even when the consequences of making an error are benign - is unlikely to help you learn. In fact, in her review of the literature, Janet Metcalfe makes a compelling argument that making errors while learning - so long as you receive corrective feedback - results in better outcomes than making no errors at all.

Her findings are somewhat counterintuitive. If the goal is to perform flawlessly in high-stakes situations, shouldn't we pursue perfection in order to prepare for them? Early theorists feared that the commission of errors would make it harder to learn the correct response later on. One of the most famous psychologists of the 20th century, Albert Bandura, believed that only correct responses should be rewarded; errors, if they occurred, should be ignored. However, what Metcalfe's review of the literature suggests is that errors should be encouraged as part of an active exploratory learning process, so long as corrective feedback is provided when students make mistakes. Indeed, Metcalfe cites more than a dozen studies that show students achieve superior outcomes when they are encouraged to generate answers before being shown the correct response, rather than simply being shown the correct response. That may sound obvious as you read it now, but it's not. The vast majority of the American education system operates under an idea similar to Bandura's - namely, that only correct responses should be rewarded, and that errors should generally be ignored. The findings of this emerging body of research fly in the face of conventional wisdom.

The reasons for these findings are not yet well-understood. For example, in experiments where the experimenter suggests a wrong answer, memory for the correct answer later on is actually diminished. Thus mere exposure to errors does not explain these results. Rather, it seems that in order to benefit from making errors, you have to be actively involved in generating them. That is consistent with a general theme in all of my blogs, namely that actively engaging in the learning process results in materially superior outcomes than trying to passively absorb all that you need to know. The exciting thing about Metcalfe's and others' research is that it suggests that your active engagement need not be perfect to be effective: simply trying to engage with the concepts, attempting to come up with an answer before looking at the solution, and then seeking out the correct solution if you make a mistake is enough. If you find that you made a mistake, figure out how to reconcile what you did with the correct process. Were you on the right track and simply took a wrong turn? If so, take note of what threw you off. If your answer was completely different, did you get confused about something fundamental to the underlying concept? In that case, spend some time going over that concept before you move on to more questions.

Somewhat surprisingly, the students who benefit most from making errors are the ones who, prior to having their error pointed out, were highly confident their answer was correct. Consequently, it makes sense to avoid doing a bunch of practice questions on a topic you feel you know little about. A better approach would be to study the underlying concepts first, and then attempt a few easy questions to see if you're getting the hang of them before moving on to a more mixed practice set. For example, say you're working on problems about triangles but you don't know the properties of equilateral and isosceles triangles, you are unfamiliar with the idea of special right triangles, and you have never heard of Pythagorean Triplets. Before you torture yourself with a practice set that will inevitably require you to know about these topics, take a moment to learn about them. I personally like YouTube and often use the number of views a video has a proxy for quality, but even a simple Google search can suffice. After you develop a rudimentary foundation, test it out on a few easy problems, and then move on to a more mixed practice set. You don't need to know everything about these topics to advance to a mixed practice set - indeed, the point of this article is that making errors while you're learning is beneficial. You just need to learn enough to feel like you should be able to get some questions right. Whether or not you do is not what's immediately important. The point is that if you get questions wrong after you've learned enough to feel like you should get them right, you'll have an easier time correcting your mistakes. A tutor can be very helpful in this context, and their goal should be to help you understand what about your thinking led to the mistake, not just to show you how to answer the question correctly.

The truism that "everyone makes mistakes," which often carries a tinge of regret, need not lead us to feel disappointed. In the context of learning at least, mistakes appear to be necessary to achieve the best results. Does that mean you should rush to learn multivariable calculus without any training, since you know you will surely make mistakes? Obviously not. But it does mean that both teachers and students should embrace errors as part of the learning process. Even more importantly, it means that you do not have to be highly confident to become actively engaged in the learning process - the process of actively engaging by itself, even if you make errors, is what is likely to result in the best outcome for you. Making errors when you are actively engaging in the learning process is more likely to HELP you than hurt you. How cool is that?


There is a sci if story I read long ago with a character who never had a physical accident, never slipped, never tripped, never bumped into anything, etc. Her motion was completely lacking in grace and beauty because she never had to learn to move smoothly and gracefully to avoid mishaps. We learn from our mistakes. And without making mistakes, we never learn.


Peter A.

Harvard PhD Student | Expert SAT, ACT, & GRE Tutor

900+ hours
if (isMyPost) { }