Have you ever been sure your answer was right, but it was wrong? That feels bad. But there is good news.

Scientists found that when you are very sure and still make a mistake, you remember the correct answer better after you see it. This is called the hypercorrection effect. Surprise makes your brain pay attention. But you must check the right answer soon, and try the question again later.

Making a mistake on a question you were sure about is embarrassing, but it may be one of the best chances to learn. In 2001, Butterfield and Metcalfe at Columbia University gave students general knowledge questions and asked how confident they were in each answer. After wrong answers, the students were shown the correct ones.

On a later test, the errors made with high confidence were the most likely to be corrected. This is called the hypercorrection effect. One explanation is surprise: the shock of being wrong focuses attention. Another is that confident people often already know something about the topic. However, a 2011 study found that some confident errors came back a week later.

Intuition suggests that an error held with strong conviction should be the hardest to dislodge. Research on the hypercorrection effect indicates the opposite. In a 2001 study at Columbia University, Butterfield and Metcalfe asked students general-knowledge questions, collected confidence ratings for each answer, and showed the correct answer after every mistake. On a retest, high-confidence errors were corrected more often than low-confidence ones. Two explanations have been proposed.

The first is that the surprise of discovering a confident belief is wrong increases attention at the moment the correct answer appears. The second, supported by a 2011 study by Metcalfe and Finn, is that people who make confident errors frequently have relevant background knowledge and may even have partly known the correct answer. Despite its robustness in the laboratory, the effect has limitations. It has mostly been demonstrated with short factual questions, so whether it applies equally to deep scientific misconceptions is uncertain.

In addition, Butler and colleagues found in 2011 that, although the effect persisted after a week, some high-confidence errors reappeared. The practical message is twofold: always seek feedback immediately after a mistake, and retest corrected items after a delay so that old errors do not return.