Richard Feyman, a Nobel Prize-winning physicist, told the story of an explanation his father gave him when he was a boy.

One day, I was playing with what we call an express wagon, which is a little wagon with a rail ing around. I had a ball in it, and I noticed some thing about the way the ball moved. So, I went to my father, and I said, “Pop, I noticed that when I pull the wagon, the ball rolls to the back, and when I’m pulling along and I suddenly stop, the ball rolls to the front of the wagon. Why is that?”

And he said,“Nobody knows.”

He said the general principle is that things that are moving try to keep on moving and things that are standing still tend to stand still unless you push on them. He said this tendency is called inertia, but nobody knows why it’s true.

That’s a deep understanding. He doesn’t give me a name. He knew the difference between know ing the name of something and actually knowing something.

“If you look close,” he said, “you’ll find the ball does not rush to the back of the wagon, but it’s the back of the wagon that you’re pulling against the ball. The ball stands still.”

So, I ran back to the little wagon and set the ball up again and pulled the wagon from under it and looked sideways and saw he was right. The ball never moved backwards in the wagon when I pulled the wagon forward. It moved backward relative to the wagon, but relative to the sidewalk, it stood still until the wagon caught up with it.

What more surprising and curiosity-inducing expression can a small boy hear from an expert than “nobody knows”? Having hooked young Feynman, his father gave him a brief “general principle” and an idea for an experiment to confirm it.

Explanations are the mainstay of school instruction, accounting for the vast majority of what teachers do in class each day. But research has shown repeatedly that they don’t reliably lead to learning. (They are, in fact, very effective for the giver but seldom for the receiver.)

Why not? Nobody knows. But researchers have uncovered some general principles.

Two expert cardiologists based in Chicago, Joel Michael and Allen Rovick, analyzed transcripts of tutoring sessions in which they were the tutors. Their approach was to show tutees scenarios in which a person’s blood pressure was perturbed and to ask them to predict cardiovascular responses. Michael or Rovick then began a dialog with the tutee about the accuracy of their predictions. The researchers were well aware of the limited value of explanations. But when they examined transcripts of their own tutoring, they were shocked to discover that, at the first sign of an error in tutee thinking, they flipped into full-on explanation mode and never flipped back. They had unwittingly demonstrated that, like a trout and a dragonfly lure, not even an experienced tutor can resist the draw of explaining.

Most pernicious are unsolicited explanations in which tutors launch into a lecture, often at the start of a session. As well as being ineffective, they waste valuable tutoring time.

But there is “a time for telling,” as the title of a famous paper by education researchers Dan Schwartz and John Bransford has it. They showed that there are moments when an explanation is not only okay but that failing to provide one would be a missed opportunity.

For example, a student might have the insight that when comparing similar right triangles—triangles that are the same except for their size—the two sides next to the right angle are always in the same ratio to each other. But they will not figure out for themselves that the ratio is called the tangent and that someone back in antiquity made a handy lookup table of that ratio for every triangle. That is a time for telling.

Schwartz and Bransford call this the “detective story” approach. The explanation is the solution to the story. It comes at the end. Instead of starting with an explanation of trigonometry and then giving examples to illustrate it, the student divines something in the examples first and the denouement comes with the tutor’s explanation.

In other words, an explanation is the right move when the tutee is ready for it, not when the tutor is. This is the first of many examples we are going to encounter showing that effective tutors are following the student’s lead, even though it is almost universally assumed by tutors that it should be the other way around.

The research literature is unequivocal on one way through which most tutors can instantly improve their effectiveness: talk less. “Most tutors just won’t shut up,” says Micki Chi, a professor in the Institute for the Science of Teaching and Learning at Arizona State University. She once ran an experiment in which she attempted to train tutors to stop making long explanations. (“How long is long?” I asked. “You know it when you hear it,” she replied.) Chi trained the tutors to replace explanations with “content-free” prompts such as these:

But the training didn’t take. Tutors found it difficult to keep to purely content-free prompts. Instead, they often substituted content-specific prompts such as “Which of Newton’s laws would help here?” or “How could you get the x on its own?” The problem with these prompts is that they reflect the pathway the tutor has in mind, which may not match the pathway the student has in mind. This approach to tutoring isn’t totally ineffective, though it can turn sessions into a game of “guess what the tutor wants me to do next.” Students can get quite good at this game, but it leaves them unprepared when they face similar problems in the future and the tutor isn’t around.

Although Chi’s tutors didn’t keep to content-free prompts, they were successful in reducing long explanations. What effect did that have on student learning? Surprisingly, learning without explanations was as good as learning with explanations, even though students who got the long explanations heard a lot more information than those who heard only the prompts. Without explanations, the tutor–tutee conversation became much more interactive. Students initiated more of the dialog and did proportionately more of the work and that led them to learn more.

Research has uncovered some principles for how, as well as when, to give an explanation. The first is that the explanation must be at the learner’s knowledge level. We’ll call that a considerate explanation. When technical support agents help customers fix technology problems, they very often bamboozle the caller. That’s if they give an explanation at all. They may simply tell you a sequence of mysterious steps. As a result, when the problem reoccurs, you are no better able to solve it. But when agents were given some information about a caller’s it expertise—callers answered questions such as “Do you know what ftp is used for?”—the agents were able to adapt their explanations in a way that made them far more effective for the callers. The way they did this was to give more definitions to less-skilled customers while talking more about processes and events (that rely on an understanding of definitions) to more skilled customers. They recalibrated their mental model of whom they were talking to: instead of picturing a generic layperson, they pictured a specific individual with competencies and confusions.

Tutors, likewise, must encourage tutees to make their thinking visible, like video gamers who narrate their gameplay on YouTube. That allows the tutor to pitch their explanations at the level of the tutee, or a bit above. This is akin to adjusting your language when trying to be comprehensible to a non-native speaker: slowing down and using higher-frequency vocabulary (rather than just talking more loudly). Hence, the term “considerate explanation”.

The second general principle, which was noted over 30 years ago by Paul Vedder, a Dutch teacher, is a wonderfully simple technique for supercharging explanations. We’ll call it active explanation. Here is an example from vocabulary instruction:

S: What does “heathen” mean? T: Sort of uncivilized. It’s not usually meant kindly. (Pause) So, would you call the King of England a heathen, do you think? S: No. T: Neither would I. Who would you call a heathen? S: (Thinks) You? T: What! Why? S: You eat with your fork and knife in the wrong hands.

Vedder suggested explanations that include an opportunity for the learner to immediately put new knowledge (such as the meaning of “heathen”) to work are more effective. An easy way to do this is to come up with an example or counterexample and ask the learner to say which it is—for example, is the King of England a heathen? This technique is effective for two reasons. It gives the learner an opportunity to use their new knowledge, which helps memory. And it uncovers situations where the explanation didn’t click.

I once ran a workshop with a group of business executives, and after 20 minutes of explaining, I finally gave them an exercise to check how much they had absorbed. The answer was next to nothing. I was shocked, though every teacher I’ve described this to has a similar story. The solution, of course, is not to wait 20 minutes.

The revelation that they didn’t understand can also come as a surprise to the learner, who frequently has the illusion of understanding to match the teacher’s illusion of having explained. Active explanations rapidly disabuse them both of the notion.

Richard Feyman’s father intuitively followed these prescriptions: he pitched his description of the physics at the level of a boy’s understanding (considerate explanation), gave young Feyman a follow-up to pursue (active explanation), and provided the explanation in response to a specific question (time for telling). Perhaps it should not be surprising that Feyman himself went on to become one of the world’s great explainers.