Make It Stick: The Science of Successful Learning

by Peter C. Brown

Elizabeth and Robert Bjork, who coined the phrase “desirable difficulties,” write that difficulties are desirable because “they trigger encoding and retrieval processes that support learning, comprehension, and remembering. If, however, the learner does not have the background knowledge or skills to respond to them successfully, they become undesirable difficulties.”19

Having somebody whisper in your ear while you read the news may be essential training for a TV anchor. Being heckled by role-playing protestors while honing your campaign speech may help train up a politician. But neither of these difficulties is likely to be helpful for Rotary Club presidents or aspiring YouTube bloggers who want to improve their stage presence.

Learning is at least a three-step process: initial encoding of information is held in short-term working memory before being consolidated into a cohesive representation of knowledge in long-term memory. Consolidation reorganizes and stabilizes memory traces, gives them meaning, and makes connections to past experiences and to other knowledge already stored in long-term memory. Retrieval updates learning and enables you to apply it when you need it.

When practice conditions are varied or retrieval is interleaved with the practice of other material, we increase our abilities of discrimination and induction and the versatility with which we can apply the learning in new settings at a later date. Interleaving and variation build new connections, expanding and more firmly entrenching knowledge in memory and increasing the number of cues for retrieval.

Monitoring your own thinking is what psychologists call metacognition (meta is Greek for “about”). Learning to be accurate self-observers helps us to stay out of blind alleys, make good decisions, and reflect on how we might do better next time. An important part of this skill is being sensitive to the ways we can delude ourselves. One problem with poor judgment is that we usually don’t know when we’ve got it. Another problem is the sheer scope of the ways our judgment can be led astray.1

Our understanding of the world is shaped by a hunger for narrative that rises out of our discomfort with ambiguity and arbitrary events.

The results showed that overhearing one side of a conversation proved more distracting than overhearing both sides, and the content of those partial conversations was better recalled later by the unintentional eavesdroppers. Why was this? Presumably, those overhearing half a conversation were strongly compelled to try to infer the missing half in a way that made for a complete narrative. As the authors point out, the study may help explain why we find one-sided cell phone conversations in public spaces so intrusive, but it also reveals the ineluctable way we are drawn to imbue the events around us with rational explanations.

The psychologists Larry Jacoby, Bob Bjork, and Colleen Kelley, summing up studies on illusions of comprehension, competence, and remembering, write that it is nearly impossible to avoid basing one’s judgments on subjective experience. Humans do not give greater credence to an objective record of a past event than to their subjective remembering of it, and we are surprisingly insensitive to the ways our particular construals of a situation are unique to ourselves.

The curse-of-knowledge effect is close kin to hindsight bias, or what is often called the knew-it-all-along effect, in which we view events after the fact as having been more predictable than they were before they occurred.

Respondents’ most emotional memories of their personal details at the time they learned of the attacks are also those of which they are most confident and, paradoxically, the ones that have most changed over the years relative to other memories about 9/11.14

Incompetent people lack the skills to improve because they are unable to distinguish between incompetence and competence. This phenomenon, of particular interest for metacognition, has been named the Dunning-Kruger effect after the psychologists David Dunning and Justin Kruger. Their research showed that incompetent people overestimate their own competence and, failing to sense a mismatch between their performance and what is desirable, see no need to try to improve.

In one series of studies that demonstrate this finding, they gave students a test of logic and asked them to rate their own performance. In the first experiment the results confirmed expectations that the least competent students were the most out of touch with their performance: students who scored at the twelfth percentile on average believed that their general logical reasoning ability fell at the sixty-eighth percentile. In a second experiment, after taking an initial test and rating their own performance, the students were shown the other students’ answers and then their own answers and asked to reestimate the number of test questions they had answered correctly. The students whose performance was in the bottom quartile failed to judge their own performance more accurately after seeing the more competent choices of their peers and in fact tended to raise their already inflated estimates of their own ability.

Don’t make the mistake of dropping material from your testing regime once you’ve gotten it correct a couple of times. If it’s important, it needs to be practiced, and practiced again. And don’t put stock in momentary gains that result from massed practice. Space your testing, vary your practice, keep the long view.

The apprentice model is a very old one in human experience, as novices (whether cobblers or attorneys) have traditionally learned their craft from experienced practitioners. In other settings, teams are formed of people with complementary areas of expertise.

Take police work, where many different forms of simulation are used in training. For firearms training it’s often video-based scenarios, with a large screen set up at one end of a room where a number of props have been placed to imitate the situation confronting the officer, who enters the scene armed with a gun that has been modified to interact with the video.

The training regime had violated the cardinal rule that you should practice like you play, because you will play like you practice.

On any list of differences that matter most for learning, the level of language fluency and reading ability will be at or near the top.

A report on a 2004 survey conducted for Britain’s Learning and Skills Research Centre compares more than seventy distinct learning styles theories currently being offered in the marketplace, each with its companion assessment instruments to diagnose a person’s particular style. The report’s authors characterize the purveyors of these instruments as an industry bedeviled by vested interests that tout “a bedlam of contradictory claims” and express concerns about the temptation to classify, label, and stereotype individuals.

An approach called VARK, advocated by Neil Fleming, differentiates people according to whether they prefer to learn through experiences that are primarily visual, auditory, reading, or kinesthetic (i.e., moving, touching, and active exploration).

In 2008 the cognitive psychologists Harold Pashler, Mark McDaniel, Doug Rohrer, and Bob Bjork were commissioned to conduct a review to determine whether this critical claim is supported by scientific evidence. The team set out to answer two questions. First, what forms of evidence are needed for institutions to justify basing their instructional styles on assessments of students’ or employees’ learning styles?

The second question the team asked was whether this kind of evidence existed. The answer was no.

Psychologists today generally accept that individuals possess at least two kinds of intelligence. Fluid intelligence is the ability to reason, see relationships, think abstractly, and hold information in mind while working on a problem; crystallized intelligence is one’s accumulated knowledge of the world and the procedures or mental models one has developed from past learning and experience.

What both theories lack is an underpinning of empirical validation, a problem Gardner himself recognizes, acknowledging that determining one’s particular mix of intelligences is more an art than a science.8

Rather than eight intelligences, Sternberg’s model proposes three: analytical, creative, and practical. Further, unlike Gardner’s theory, Sternberg’s is supported by empirical research.9

Children whose environments prized one kind of learning over another (practical over academic, in the case of the families who taught their children about herbs) were at a lower level of knowledge in the academic areas not emphasized by their environment. Other families placed more value on the analytic (school-based) information and less on the practical herbal knowledge.

The study found that IQ is unrelated to handicapping ability, and “whatever it is that an IQ test measures, it is not the ability to engage in cognitively complex forms of multivariate reasoning.”11

Analytical intelligence is our ability to complete problem-solving tasks such as those typically contained in tests; creative intelligence is our ability to synthesize and apply existing knowledge and skills to deal with new and unusual situations; practical intelligence is our ability to adapt to everyday life—to understand what needs to be done in a specific setting and then do it; what we call street smarts.

Robert Sternberg and Elena Grigorenko have proposed the idea of using testing to assess ability in a dynamic manner. Sternberg’s concept of developing expertise holds that with continued experience in a field we are always moving from a lower state of competence to a higher one. His concept also holds that standardized tests can’t accurately rate our potential because what they reveal is limited to a static report of where we are on the learning continuum at the time the test is given.

In tandem with Sternberg’s three-part model of intelligence, he and Grigorenko have proposed a shift away from static tests and replacing them with what they call dynamic testing: determining the state of one’s expertise; refocusing learning on areas of low performance; follow-up testing to measure the improvement and to refocus learning so as to keep raising expertise.

Dynamic testing has three steps. Step 1: a test of some kind—perhaps an experience or a paper exam—shows me where I come up short in knowledge or a skill. Step 2: I dedicate myself to becoming more competent, using reflection, practice, spacing, and the other techniques of effective learning. Step 3: I test myself again, paying attention to what works better now but also, and especially, to where I still need more work.

One of these differences is the idea mentioned earlier that psychologists call structure building: the act, as we encounter new material, of extracting the salient ideas and constructing a coherent mental framework out of them.

These frameworks are sometimes called mental models or mental maps. High structure-builders learn new material better than low structure-builders.

High structure-builders develop the skill to identify foundational concepts and their key building blocks and to sort new information based on whether it adds to the larger structure and one’s knowledge or is extraneous and can be put aside. By contrast, low structure-builders struggle in figuring out and sticking with an overarching structure and knowing what information needs to fit into it and what ought to be discarded. Structure building is a form of conscious and subconscious discipline: stuff fits or it doesn’t; it adds nuance, capacity and meaning, or it obscures and overfreights.

What’s happening in this situation remains an open question for now, but the implication for learners seems to reinforce a notion offered earlier by the neurosurgeon Mike Ebersold and the pediatric neurologist Doug Larsen: that cultivating the habit of reflecting on one’s experiences, of making them into a story, strengthens learning.

Another cognitive difference that appears to matter is whether you are a “rule learner” or “example learner,” and the distinction is somewhat akin to the one we just discussed. When studying different kinds of problems in a chemistry class, or specimens in a course on birds and how to identify them, rule learners tend to abstract the underlying principles or “rules” that differentiate the examples being studied. Later, when they encounter a new chemistry problem or bird specimen, they apply the rules as a means to classify it and select the appropriate solution or specimen box.

However, example learners may improve at extracting underlying rules when they are asked to compare two different examples rather than focus on studying one example at a time.

Here’s the payoff: after figuring out this common, underlying solution, students are then able to go on to solve a variety of different convergence problems.14

Knowledge is not knowhow until you understand the underlying principles at work and can fit them together into a structure larger than the sum of its parts. Knowhow is learning that enables you to go do.

Describe what you want to know, do, or accomplish. Then list the competencies required, what you need to learn, and where you can find the knowledge or skill. Then go get it.

Of more than six hundred children who took part in the experiment, only one-third succeeded in resisting temptation long enough to get the second marshmallow.

This study showed what others have suggested, that the speed of our mental abilities is determined by the robustness of our neural connections; that this robustness, at the initial stages, is largely determined by our genes, but that our neural circuitry does not mature as early as our physical development and instead continues to change and grow through our forties, fifties, and sixties.

The thickness of the myelin coating correlates with ability, and research strongly suggests that increased practice builds greater myelin along the related pathways, improving the strength and speed of the electrical signals and, as a result, performance. Increases in piano practice, for example, have shown correlated increases in the myelination of nerve fibers associated with finger movements and the cognitive processes that are involved in making music, changes that do not appear in nonmusicians.8

The study of habit formation provides an interesting view into neuroplasticity. The neural circuits we use when we take conscious action toward a goal are not the same ones we use when our actions have become automatic, the result of habit. The actions we take by habit are directed from a region located deeper in the brain, the basal ganglia. When we engage in extended training and repetition of some kinds of learning, notably motor skills and sequential tasks, our learning is thought to be recoded in this deeper region, the same area that controls subconscious actions such as eye movements.

Some researchers have used the word “macro” (a simple computer app) to describe how this chunking functions as a form of highly efficient, consolidated learning.

Another fundamental sign of the brain’s enduring mutability is the discovery that the hippocampus, where we consolidate learning and memory, is able to generate new neurons throughout life.

This rise in neurogenesis starts before the new learning activity is undertaken, suggesting the brain’s intention to learn, and continues for a period after the learning activity, suggesting that neurogenesis plays a role in the consolidation of memory and the beneficial effects that spaced and effortful retrieval practice have on long-term retention.9

Learning and memory strategies such as retrieval practice and the building of mental models are effective for enhancing intellectual abilities in the material or skills practiced, but the benefits don’t extend to mastery of other material or skills. Studies of the brains of experts show enhanced myelination of the axons related to the area of expertise but not elsewhere in the brain.

Richard Nisbett writes of environmental “multipliers” that can deliver a disproportionate effect from a small genetic predisposition—the kid who is genetically just a little bit more curious becomes significantly smarter if she’s in an environment that feeds curiosity. Now stand that notion on its head. Since it’s unlikely I’ll be raising my IQ anytime soon, are there strategies or behaviors that can serve as cognitive “multipliers” to amp up the performance of the intelligence I’ve already got? Yes. Here are three: embracing a growth mindset, practicing like an expert, and constructing memory cues.

Those who interpret failure as the result of insufficient effort or an ineffective strategy dig deeper and try different approaches.

When you praise for intelligence, kids get the message that being seen as smart is the name of the game. “Emphasizing effort gives a child a rare variable they can control,” Dweck says. But “emphasizing natural intelligence takes it out of a child’s control, and it provides no good recipe for responding to a failure.”18