This article was originally published on 7 February 2024

Books and nature programmes love to tell stories about parent birds teaching their fledglings to fly and mother cats teaching their kittens to chase mice. But these stories are blatant examples of anthropomorphising. Birds fly and cats rush after small objects, all without being taught. When a kitten chases a ball of paper, this is because the kitten’s brain has built-in motor responses for hunting, triggered by certain stimuli.

While most young animals, including humans, learn from observing their elders, this isn’t teaching. Their nervous system is primed to copy what other “agents” are doing, while the agents display little, if any, interest in them. We would not call these agents “teachers”; they are simply behaving as usual.

Teaching, at a minimum, requires agents to directly address their pupils and actively facilitate the transfer of information. Such a basic form of teaching, although rare in animals, has been observed in some species.

Meerkats, for example, give pups the opportunity to interact with dangerous prey, taking account of their level of experience. They give dead scorpions to the youngest pups and offer live scorpions with their stings removed to the older ones. At a later stage, the pups are confronted with intact live scorpions, as the parent remains nearby to monitor the pups’ handling of the prey.

As far as we know, the meerkats do not have any explicit recognition that they are teaching their young. They are simply responding to the cries of the pups, which change as the pups get older, and adapting their behaviour accordingly.

Humans, however, know exactly what they are doing when they teach - and it is this explicit recognition that can elevate our teaching to a different level.

We do not wish to claim that teaching is always superior to other forms of learning. Learning with and without a teacher happens side by side. Different types of learning have their place, with unconscious and conscious processes often intermingled.

For instance, we can learn spoken language efficiently by mere exposure, “zombie” fashion. However, we learn written language only by being taught explicitly.

In general, explicit instructions are helpful for learning rule-based tasks. But there are many complex tasks where it is difficult to state explicit rules, and so it is difficult to give instructions. In these cases, people are better off learning by copying or mere exposure, especially when it is necessary to integrate information from many sources.

Whatever other kinds of teaching are practised, though, it is explicit teaching that enables humans to share stories that aim to explain the world and our place in it. Such stories are continuously and flexibly constructed through our ability to create and share meaning. Indeed, sharing them is crucial for creating a cohesive society.

So how does explicit teaching work?

In discussions of education, it is easy to remain fixated on the learner. We suspect that the extent to which teaching is actually based on mechanisms of social cognition has not always been acknowledged. This is a topic we explore in our new book, What Makes Us Social? (see box).

“Social cognition” is our term for capturing everything that can, in principle, explain how animals, including humans, learn from each other, cooperate and compete with each other. It also covers how humans create culture.

Drawing on our growing understanding of social cognition, we would like to overturn the view that teaching is a one-sided affair, with the teacher providing information and the pupil receiving it.

Instead, we want to make the case that, for teaching to be more successful, information providers and receivers need to be working together, through mutual adaptation.

The role of implicit teaching

Before discussing advanced forms of formal teaching, we need to consider a form of implicit teaching identified by Gergely Csibra and György Gergely in 2009: “natural pedagogy”.

In natural pedagogy, parents or carers of infants use ostensive (directly demonstrative) gestures to show when they are teaching, and infants respond to these gestures. For example, infants will follow the direction of the gaze of an adult specifically, if this is preceded by gestures such as eye contact or infant-directed speech.

These interactions are effective means of learning. Infants spontaneously assume that messages conveyed in this way are worth remembering and that they contain durable meanings that can be generalised. In this way, children effortlessly learn that apples are fruit and good to eat. In contrast, when infants observe an adult either eating or rejecting an apple, without there being an ostensive gesture, they learn that this adult likes or dislikes apples.

All this makes natural pedagogy a new type of social learning that may well represent an evolutionary adaptation in humans.

In addition to using ostensive gestures to indicate that useful information is forthcoming, adults modify their teaching behaviour to match the needs of the learner. Language directed towards children, known as “motherese”, is fine-tuned to the development of the individual child. For example, in a “guess the animal” game, parents give more informative clues about animals that they think a particular child does not know.

All of this happens spontaneously. But beyond natural pedagogy, adults can use more sophisticated forms of teaching once children acquire language, allowing them to turn the learning process into one of joint action.

Mutual adaptation in teaching and learning

Evidence suggests that a key feature of working together is mutual adaptation in both partners. But for teaching, this presents a problem: teachers and pupils inevitably differ in competence.

So, how do we overcome this? The answer is that the teacher must adapt downward, the pupil upward. The teacher always has to be a step ahead, but not too far ahead.

We can throw some light on the mutual adaptation that occurs between teacher and learner with the help of an experiment conducted in 2010, by Ivana Konvalinka and colleagues. In it, two players had to tap on a keyboard in sync while also keeping to a particular tempo, indicated by a metronome. They were wearing earphones that allowed them to hear each other’s taps, but not their own.

The taps did not always occur at exactly the same time, but the players continually attempted to adjust their speed to achieve synchrony. When A’s tap was later than B’s, A tried to speed up, while B slowed down. This adjustment was aimed at what we have called “closing the loop”.

In many examples of social interaction, one agent simply responds to another, but this is different with mutual adaptation. There are always tiny gaps between responses, and these elicit a reciprocal response, so the loop can be closed.

Usually, a leader-follower relationship develops. In the case of the tapping task, this happens when A concentrates on keeping in tempo and B is mainly concerned with maintaining synchrony. The rapid and continuous nature of the adaptation is possible only because it happens through automatic information processing that takes place outside conscious awareness.

The leader-follower relationship in the tapping task seems to us highly relevant to teacher-pupil interactions. If the teacher aims to transfer new knowledge to the pupil, he or she must adapt to the pupil’s existing knowledge. The pupil, on the other hand, must be able to integrate the new information into his or her repertoire before moving to the next stage.

But how does the transfer of information between teacher and pupil actually work?

The basis of deliberate teaching

The kind of communication required for mutual alignment in deliberate teaching is neither simple nor direct, as it involves recognising the hidden states of knowledge that exist in both the teacher and the pupil.

The aim is for the hidden state of knowledge in the mind/brain of the teacher to be recreated in the mind/brain of the pupil. However, neither teacher nor student can directly perceive the hidden state of the other.

Given that we believe the brain is essentially a prediction engine, one solution is to let this engine estimate what the hidden state might be.

How this works is illustrated by a 2007 study, which found that it is possible to read a person’s intentions from their movements in a particular context. For example, if a person is standing in the kitchen and picks up a glass left near the sink, it is probably more likely that their intention is to wash the glass up than to throw the contents into an irritating person’s face.

The engine (the brain) makes an informed guess about what is going on in the brain of the other person (the hidden state), and then predicts what that person will do. Their prediction will then be updated by any further information they receive. This process allows a continuous adjustment of guesses until the predictions are sufficiently accurate. At this point, the interpretation is likely to be correct.

Read more:

• Can you see learning in the brain?
• Is knowledge transfer education’s holy grail?
• Why explicit instruction matters, according to neuroscience

We can apply this same theory to a teaching situation. Consider a scenario in which the teacher (let’s call him Karl) is helping the student (let’s call him Chris) to understand a new concept (“free energy”). Chris will assign great importance to prediction errors to update his naive prior beliefs about the nature of this new concept, while the knowledgeable teacher will try to change the belief of his listener.

For the teacher, prediction errors indicate that his listener still has not understood. What Chris has just said does not fit Karl’s understanding of the concept. So Karl will try to change Chris’ understanding of “free energy” and may even modify his own understanding to improve communication.

However, this is not enough. What Karl has just said does not fit Chris’ current understanding of the concept, so Chris must update his representation of it.

The ideal result of the interaction will be a generalised synchronisation between the speakers in which prediction errors have been minimised. The emergence of such synchronisation indicates that the concept has been successfully communicated - and each party can accurately predict what the other will say about it.

As a result, Chris’ concept of “free energy” will have changed a lot, but even Karl’s will have changed a little as a result of interacting with Chris.

While natural pedagogy happens in “zombie” fashion, deliberate and explicit teaching like this comes at a cost. It takes conscious effort and is far slower, and sometimes painful, to achieve.

We believe that improvements could be made if there was a more advanced science of teaching. As a start, it would be helpful to know what makes an effective teacher.

What makes a good teacher?

It seems that, very early on, children recognise that some people are better teachers than others. At 14 months, children already copy the actions of a competent model more often than an incompetent model, where competence has previously been demonstrated in perfecting a skill, such as putting on a pair of shoes.

And by four years, children can predict whether an informant will be accurate in the future. They seek and endorse information from accurate rather than inaccurate informants.

Competence and accuracy are obvious criteria for being a good teacher, but there must be many others. From our work, we would predict that teachers also need to be sensitive to the knowledge state of the pupil and able to adapt their teaching to this state.

By studying the effect of verbal instructions in the lab and using brain-scanning equipment, we can make some progress in further understanding the processes involved in explicit teaching. And we would do well to understand them, since explicit teaching greatly facilitates the acquisition of knowledge about all that is important in our culture.

An enlightening 2019 study from Mahzarin Banaji’s lab investigated how 7- to 11-year-old children learn stereotypic attitudes toward outgroups (a social group that the subject does not identify themselves as being a member of). They compared the effects of explicit verbal instruction and implicit association learning.

The experimenters artificially created two groups, which were called “longfaces” and “squarefaces”. The children were shown repeated pairings of pictures of longface avatars with nasty things, such as snakes, and squareface avatars with nice things, such as puppies. This had no effect on their attitudes to the two groups. But when these images were combined with verbal instructions, children rapidly learned that longfaces were “bad” and that squarefaces were “good”.

This shows how powerful explicit verbal instructions can be. A moment’s thought can convince us that the human ability to learn by following instructions can have undesirable effects. For instance, instructions enable efficient coercion and oppression. They can make people do things that they would not do of their own accord.

That is why telling people what to do produces intense worry for us, as experimental psychologists. For every experiment, we have to carefully consider how to tell our participants what we expect them to do. The words must be the same for every participant and they need to be utterly clear. But even obsessive care about the exact wording of the instructions is not sufficient. We also need to monitor participants to make sure the task they are doing is the same as the one that we intended them to do.

We are only at the beginning of explaining how instructions can reconfigure the brain, but what we do know is that our human ability to teach new concepts provides a powerful means for creating shared models of the world, which are fundamental to cumulative culture from prehistoric times to the present day.

The evolutionary anthropologist Robert Boyd made a strong case for cultural learning as the basis of the success of human adaptation in many environments, from jungle to desert and from arctic to tropic. Cultural learning can evolve over generations and can be transmitted over long distances of time and space.

It is this kind of learning that will determine the highest goals that we can pursue - and it’s only through explicit teaching that this kind of learning is possible.

Chris Frith is emeritus professor of neuropsychology at the Wellcome Centre for Human Neuroimaging at University College London and honorary research fellow at the Institute of Philosophy, University of London. Uta Frith is emeritus professor of cognitive development at the Institute of Cognitive Neuroscience at University College London. This article is an edited extract of What Makes Us Social? by Chris Frith and Uta Frith

For the latest research, pedagogy and practical classroom advice delivered directly to your inbox every week, sign up to our Teaching Essentials newsletter