Why the best learners talk to themselves | Tania Lombrozo

Learning can begin by working with information already in your mind, rather than immediately searching for more. Tania Lombrozo argues that memories and skills can be recombined to answer questions and make sense of a problem. A programmer who explains a problem to a rubber duck may learn from the exercise even though the duck supplies no information: the value lies in putting the problem into words and examining what one can explain. Lombrozo has studied learning for more than 25 years.

Many successful students explain material to themselves, ask questions, and try to answer them. Self-explanation is associated with better learning and with applying knowledge more broadly, not just repeating what was studied. It can also reveal a mismatch between confidence and understanding. People may believe they understand how something works but struggle to explain it; Lombrozo describes this as the illusion of explanatory depth. Trying to explain is therefore not only a way to learn, but also a way to test what one actually knows.

A study with children illustrates how explanation can change the way a task is understood. Adults typically paired a card showing two ducks with a card showing two frogs, grouping examples by the same number of matching objects. Five- and six-year-olds typically paired the two ducks with a duck and a fox, focusing on the shared duck. Half the children were prompted to explain why example cards belonged together. They received no feedback or new information, yet later matched new cards by same-with-same, like adults. Lombrozo’s interpretation is that explaining can alter how learners represent information and approach a task.

Noticing a gap between what seems familiar and what can actually be explained can help learners identify what to ask, what information to seek, and how to connect an answer to what they already know. Thinking alone, however, cannot provide information a person has never encountered. Learning by thinking is not a substitute for acquiring new information; it can help direct that search and shape what someone does with the information once it is found.

Lombrozo expects information overwhelm to worsen as tools for finding and synthesizing information become more pervasive. AI-generated search responses and chatbot explanations can be useful, but she cautions against letting them replace the work of learning by thinking. Starting with what one knows can help distinguish understanding that can be explained from information that is merely accessible. The practical point is not to avoid outside information, but to use reflection to identify what is missing and make new information more meaningful.