Chapter 1. Our story begins with the following question: Suppose you have to run 100 miles. Would you run at the same speed from start to finish, or vary your pace? (If you don't know the answer, please go out for a run immediately.) A century ago, almost all American runners chose to vary their pace. They did it by setting up a metronome-like series of repeated, small sprints throughout the course. This technique was known as "interval training."
Why, though, would you sprint a little bit more each time? Is it because, as you approach the finish line, you start to slow down? In fact, in the long run (pun intended), there's no obvious explanation for why people found interval training to be so effective. It was almost as if an intuitive, biological "sport science" had emerged in those days—in which scientists, coaches, and runners developed trial-and-error solutions to human performance challenges, often as a practical matter without having an especially clear understanding of the rationale or long-term health implications.
That's not the sort of thing we want from our running science. With the recent explosion in human genome sequencing, more sophisticated scientific techniques, and the invention of the smartphone, we now have a new golden age of sport science. And over the past century, most of us have already adopted a scientific approach to training and competition. You'd think that we'd already made all the basic discoveries about running—that the last major advance would be a definitive study demonstrating that a certain training protocol is highly effective for running performance.
Except it's hard to find any such definitive study. After all, when it comes to training, what's most important is not what we might _know_ , but what we _don't_ know. So while I am writing this book, I invite you to join me in imagining a hypothetical world in which interval training might actually be _the most effective way to run long distance_. What would be the reasons, if any, to think so?
•
There is only one example from which we can learn about what to do when we want to learn about the world: children. You can get a better understanding of learning by observing learning in children. This approach is not a new idea: the philosopher Francis Bacon, writing about 3,000 years ago, described the study of how animals learn as a process for investigating the workings of nature. When we think about how animals learn, there are typically a small number of options to consider. Perhaps the most important thing about how animals learn is that they almost all do it _intuitively_ , on the basis of a few concrete instances and rules of thumb.
When a baby learns language, for example, the first two things he does are babble and experiment. Babbling, as any parent who has to stop her baby talking to get her to eat knows, is a trial-and-error learning process: babbling is the process of experimenting with the language sounds and vocabulary with which an infant is born. Experimentation is what children do for months as they play with language. They gradually add words and phrases to their repertoire. For example, when an infant hears the first word in the word _bottle_ —the /b/—he may babble by inserting a /t/ sound between the first two letters. Over weeks, and with many repetitions, he learns to repeat _ba-ba-ba_ in anticipation of being given a bottle.
These two processes—trial-and-error learning and experimentation—are the basis for the classic model of learning that has been taught in schools since the nineteenth century. It is one that we all learned when we were young. We can sum up the main ideas in just a couple of phrases:
• You start with your instinctive reflexes.
• Then, as you interact with the world, you gradually refine and expand them.
Learning is a refinement of your initial reflexes. In this way, you do not progress simply from a less developed state to a more developed state, as in the linear progression of the "birth-to-death" graph in figure 1. You might start from a point B, where your ability to learn has just started, and develop along the way from that point to point A, where your ability to learn is complete. In the "birth-to-death" metaphor, the birth-to-death curve means that you start with your instincts, and your learning is a process by which you evolve them into a more rational order of reasoning.
But this model of learning is more than a metaphor. Your experience with the first steps of language learning as an infant can give us a concrete example of how human learning progresses. Babbling, in other words, is trial-and-error learning. For you to refine your instincts, you must experiment—and this experimental learning is "trying stuff out" in the real world.
Experimentation, for animals, means something that is _physically_ possible: what you experiment with must be physically possible. For example, it would not be right to try to experiment with a learning task that required you to take flight—for example, to experiment with the ability to fly by trying to jump off a table (as most children do at one time or another when learning to walk). Similarly, a squirrel who tries to leap from the branch of a tree by trying to jump through the air will just crash down to the ground. A pigeon that decides that it would be fun to leap from a balcony rail at the top of the Empire State Building to the ground will find that it hurts too much when he leaps, or that it is too hard to jump high enough to clear the rail. Similarly, a rabbit will find out that, in trying to jump as high as possible, it is painful to jump so hard. However, if the rabbit decides to hold back its jump and instead leap from a balcony that is a little lower than the Empire State Building's balcony, then, like a cat, it can land safely on the sidewalk and continue running for its life, never to be seen again by the same individual.
Figure 1. The classic model of learning.
Our point is that learning is about doing stuff, and in order to do stuff, the learning tasks we can try must be physically possible, or have a reasonable likelihood of success. This is the principle of "experimental learning": any task you try to learn must be one that you have a reasonable probability of succeeding at. Experiments are not always completely safe: we know that doing some experiments with fire is not just potentially dangerous—it can kill you. With this principle in mind, we can see how animals "adapt" from one situation to another. In order to learn to walk, for example, the first thing a baby bird does is bounce. But soon, as the baby tries the new trick of falling, falling, falling—getting its limbs wet with the water spray, trying to get enough air in its beak for breathing, getting the skin around its beak to get stretched—it learns the skill of walking. Anticipating that he will experience the same difficulty again, the baby refines and improves his strategy: when he falls, he tries to hold still and breathe, instead of frantically trying to get up. Eventually, he is able to walk with some control.
This learning process sounds complicated and counterintuitive, but we use it implicitly every time we want to learn something. When we try to learn how to fly, for example, we first try to learn how to flap our arms like the wings of a bird, which is what hummingbirds do. We then move on to more abstract ideas, such as, "What happens if I try to flap my arms like this?" and, "Maybe I'll get enough air if I flap really fast." Our instinctive reflexes can be changed in different ways through experimental learning: you can train yourself to experiment in different ways, to learn different things, through practice. For example, imagine that you've never seen this: You look at a tree with one long horizontal branch sticking out of it. You experiment with jumping from the branch, but land on the ground and can't get up. Then, you try jumping from the branch again and this time— _wow_ , you can get up. You have learned how to get up from the branch, even though you had never seen anyone else do it before.
By the way, this idea of how animals experiment with their environment as they learn is supported by numerous scientific studies of animal behavior. A key example is that of Albert Bandura's famous "Bobo doll experiment." In this experiment, which was first conducted by Bandura in 1961, two puppets are placed in front of one of the classic test subjects for this sort of research: a rhesus macaque monkey. One of the puppets is always a very aggressive monkey named Bobo, and the other is a less aggressive monkey named Curio. In one trial, Curio grabs a banana, munching away at it. Bobo, meanwhile, does not react at all. The monkey, instead of attacking the puppet, looks away from it and goes to other activities. But on the next trial, Curio reaches out and, in violation of his previous behavior, throws the banana at the Bobo puppet, thus eliciting a highly aggressive reaction. Curio gets a harsh scolding from Bobo.
On the next trial, Curio again gets the banana. This time, he puts the banana inside his mouth before reaching out. Bobo does nothing, as the monkey has observed that Curio ignores Bob
