How We Learn

by Benedict Carey

  The scientist who picked up the ball next would set the tone for a generation of research that barely moved forward an inch. Adolf Jost, an Austrian psychologist known mostly for advocating eugenics, did his own studies of spacing—also with nonsense syllables—and in 1897 formulated what became known as Jost’s Law: “If two associations are of equal strength but of different age, a new repetition has a greater value for the older one.” Translation: Studying a new concept right after you learn it doesn’t deepen the memory much, if at all; studying it an hour later, or a day later, does. Jost basically repeated one of Ebbinghaus’s experiments, found the very same thing, and got a law out of it, with his name attached. He managed to sound like he was extending the research without really doing so.

  Other psychologists followed suit, first testing more nonsense syllables and gradually graduating to lists of words or word pairs. In a way, the science went backward in the first half of the twentieth century. The psychologists who followed Jost launched scores of experiments with small numbers of people studying “grouped” or “spaced” items over intervals of minutes or even seconds, getting so lost in minutiae that by 1960 the research had succeeded mostly in showing that the spacing effect “worked” during very short time periods. If you’re told—three times, in succession—that James Monroe was the fifth president of the United States, you remember it for a while; if you’re told it three times, at ten-minute intervals, you remember it for longer.

  And it’s nice to know if you’re preparing for a trivia contest against your ten-year-old brother. But this focus on short intervals left a large question unanswered: Can spaced practice help people build and maintain a base of knowledge that’s useful in school and in life?

  In the 1970s a growing number of psychologists began asking just that, sensing that a big idea was being squandered. Some were questioning the field’s entire research tradition, including its faith in the methods of Ebbinghaus. “This all began happening during the Vietnam War protests, when students and young people were questioning authority across the board,” Harry P. Bahrick, a psychologist at Ohio Wesleyan University, told me. “That was what set these questions into motion, and people started speaking up. We spent all these years genuflecting to the giants in the field, and what did we have to show for it? Teachers and students don’t care about how many words you do or don’t remember in some ten-minute test taken in a lab. They want to know how spacing affects how well you learn French or German, how well you pick up math and science concepts. We couldn’t tell them. We had to do something completely different.”

  Bahrick wasn’t interested in extending lab findings. He wanted to blow the doors open and let in some air. He wanted to shake off the influence of Ebbinghaus, Jost, and the old guard and test long intervals, of weeks, months, years: the time periods relevant to lifetime learning. How does distributed learning contribute to building mastery of, say, auto mechanics, or music skills? Does it help at all, or are the benefits negligible? To answer that convincingly, he would have to test acquisition of the kind of knowledge that people couldn’t get casually, at work, by reading the paper, or from friends. He chose foreign language. For the experiment he had in mind, his test subjects couldn’t be just anyone, either. He had to find people who would stick with the experiment for years; who would not quit or fall out of touch; who would not misrepresent their effort; and who, ideally, could supervise their own studying.

  He settled on his wife and kids. The Bahricks are a family of psychologists. His wife, Phyllis, a therapist, and his daughters, Lorraine and Audrey, both academic researchers, would be ideal subjects. “I’m not sure it’s something they wanted to do, but I think they wanted to please me,” Bahrick, who included himself as participant number four, told me. “And over the years it became a fun family project. We always had something to talk about, and we talked about it a lot.”

  The ground rules were as follows. Phyllis, Audrey, and Lorraine would study French vocabulary words, and Harry would study German. He compiled lists of three hundred unfamiliar words per person, and each Bahrick split his or her list into six groups of fifty and studied each of those groups according to a different schedule. For one list, it was once every two weeks; for another, it was once every month; for a third, it was once every two months. They used flashcards, with French or German on one side and English on the other, and drilled themselves in each session until they remembered the meaning of all the words on that list. It was a chore much of the time. It was tedious. No one was being paid for all that study time. But it was also a start. The first truly long-term test of the spacing effect—the “Four Bahrick Study,” as they called it—was under way.

  • • •

  The best foreign language program in the world is what I call the James Method. To implement this program, simply follow the example of the American writers Henry and William James and grow up the child of wealthy, cultured parents who see to it that throughout your childhood you travel widely in Europe and the Americas and receive language tutoring along the way. The Jameses were determined that their sons have what Henry Sr. called a “sensuous education.” The most famous of the siblings, the novelist Henry, studied with tutors in Paris, Bologna, Geneva, and Bonn; he spent extended time in each place and returned periodically throughout his life. As a result, he became proficient in French, Italian, and German.

  The James Method integrates foreign language and first-rate instruction into childhood development. That’s not quite the same as growing up in a multilingual home, but it’s a pretty close facsimile. Children absorb a new language quickly when forced to speak and understand it—when living with it—and that is what the James children did to some extent. They had to memorize non-English verbs and nouns like the rest of us but did so at a time when the language modules in their brain were still developing.

  A nice gig if you can get it.

  If not—if you spent your childhood closer to Geneva, Ohio, or Paris, Texas, and want to learn Farsi—you’re at a spectacular disadvantage. You’ve got some not-so-sensual memorizing to do, and a lot of it, in relative isolation. There is no other way, no trick or secret code.

  Consider learning English as a foreign language, a challenge that millions of people around the world face if they want a certain type of job, in the sciences certainly, but also in government, in sectors of the digital economy, in tourism and trade. An educated English speaker knows twenty to thirty thousand words, along with hundreds of idioms and expressions. Stockpiling half that many words is a tall order when you’re starting from scratch. By one estimate, it takes roughly two hours of practice a day for five or so years to do so. And storing those words is only one part of the job. Remember, from the Forget to Learn theory, storage and retrieval are two different things. Just because you’ve studied (stored) the word “epitome” doesn’t mean it’s retrievable when you read or hear it. To build fluency—to keep this ever-expanding dictionary readily accessible, usable in the moment—it takes more than the time needed to store them.

  How much more?

  In 1982, about the time that Bahrick embarked on his family study, a nineteen-year-old Polish college student named Piotr Wozniak calculated an answer to that question based on his own experience: too much. At the rate he was going, Wozniak determined that he would have to study English four hours a day for years to become proficient enough to read scientific papers and converse with other scientists. He simply didn’t have the time, not while carrying a load of computer science and biology courses. He’d have to find a more efficient system, if one existed, and the only experimental subject he had was himself. He began by building a database of about three thousand words and 1,400 scientific facts in English that he was trying to absorb. He divided the total into three equal groups and started to study according to different schedules. He tried intervals of two days, four days, a week, two weeks, and so on. He kept detailed records to determine when newly learned words or facts began to defy recall.

  He began to see
a pattern. He found that, after a single study session, he could recall a new word for a couple days. But if restudied on the next day, the word was retrievable for about a week. After a third review session, a week after the second, the word was retrievable for nearly a month. He continued to refine the ideal intervals for keeping his English sharp, and programmed a computer to track his progress. “These optimum intervals are calculated on the basis of two contradictory criteria,” he wrote at the time. “Intervals should be as long as possible to obtain the minimum frequency of repetitions, and to make the best use of the so-called spacing effect … Intervals should be short enough to ensure that the knowledge is still remembered.”

  Before long, Wozniak was living and learning according to the rhythms of his system, applying it to all his subjects. The English experiment became an algorithm, then a personal mission, and finally, in 1987, he turned it into a software package called SuperMemo. SuperMemo teaches according to Wozniak’s calculations. It provides digital flashcards and a daily calendar for study, keeping track of when words were first studied and representing them according to the spacing effect. Each previously studied word pops up onscreen just before that word is about to drop out of reach of retrieval. It’s easy to use and—after Wozniak made it available as freeware in the 1990s—the program took off, especially among young people trying to learn English in places like China and Poland (it’s now a commercial website and an app).

  In effect, Wozniak had reinvented Ebbinghaus for the digital age. His algorithm answered a crucial question about the timing of intervals. To build and retain foreign vocabulary, scientific definitions, or other factual information, it’s best to review the material one or two days after initial study; then a week later; then about a month later. After that, the intervals are longer.

  By 1992, researchers saw that what began as a lab curiosity in fact had enormous potential in education. One group had shown that teaching third graders addition once a day for ten days was far more effective than twice a day for five days. Another had shown that middle school students learned biology definitions like cell, mitosis, and chromosome far better in spaced sessions than in a single class. And ever-expanding intervals—as per SuperMemo—indeed appeared to be the most effective way to build a knowledge base, making the spacing effect “one of the most remarkable phenomenon to emerge from laboratory research on learning,” one reviewer, psychologist Frank N. Dempster, of the University of Nevada, Las Vegas, wrote.

  The next year, in 1993, the Four Bahrick Study appeared in the journal Psychological Science. If Wozniak helped establish the minimum intervals required to keep newly learned facts accessible, the Bahricks provided insight into the maximum intervals for lifetime learning. After five years, the family scored highest on the list they’d reviewed according to the most widely spaced, longest-running schedule: once every two months, for twenty-six sessions. They got 76 percent of those words on a final test, compared to 56 percent on a test of words studied once every two weeks for twenty-six sessions.

  In the beginning of the study, the two-month wait meant they forgot a lot of words, compared to when they waited two weeks. That gap narrowed quickly; remember, they practiced until they knew all the words on their list during each study session. By the end, the two-month interval improved performance by 50 percent. “Who knew?” Bahrick said. “I had no idea. I thought, in two months, I might forget everything.”

  Why spaced study sessions have such a large impact on learning is still a matter of debate. Several factors are likely at work, depending on the interval. With very short intervals—seconds or minutes, as in the early studies—it may be that the brain becomes progressively less interested in a fact when it’s repeated multiple times in rapid succession. It has just heard, and stored, the fact that James Monroe was the fifth president. If the same fact is repeated again, and then a third time, the brain pays progressively less attention.

  For intermediate intervals of days or weeks, other factors might come into play. Recall the Forget to Learn theory, which holds that forgetting aids learning in two ways: actively, by filtering out competing facts, and passively, in that some forgetting allows subsequent practice to deepen learning, like an exercised muscle.

  The example we used in chapter 2 was meeting the new neighbors for the first time (“Justin and Maria, what great names”). You remember the names right after hearing them, as retrieval strength is high. Yet storage strength is low, and by tomorrow morning the names will be on the tip of your tongue. Until you hear, from over the hedges—“Justin! Maria!”—and you got ’em, at least for the next several days. That is to say: Hearing the names again triggers a mental act, retrieval—Oh that’s right, Justin as in Timberlake and Maria as in Sharapova—which boosts subsequent retrieval strength higher than it previously was. A day has passed between workouts, allowing strength to increase.

  Spaced study—in many circumstances, including the neighbor example—also adds contextual cues, of the kind discussed in Chapter 3. You initially learned the names at the party, surrounded by friends and chatter, a glass of wine in hand. The second time, you heard them yelled out, over the hedges. The names are now embedded in two contexts, not just one. The same thing happens when reviewing a list of words or facts the second time (although context will likely be negligible, of course, if you’re studying in the same place both days).

  The effects described above are largely subconscious, running under the radar. We don’t notice them. With longer intervals of a month or more, and especially with three or more sessions, we begin to notice some of the advantages that spacing allows, because they’re obvious. For the Bahricks, the longer intervals helped them identify words they were most likely to have trouble remembering. “With longer spaces, you’re forgetting more, but you find out what your weaknesses are and you correct for them,” Bahrick told me. “You find out which mediators—which cues, which associations, or hints you used for each word—are working and which aren’t. And if they’re not working, you come up with new ones.”

  When I first start studying difficult material that comes with a new set of vocabulary (new software, the details of health insurance, the genetics of psychiatric disorders), I can study for an hour and return the next day and remember a few terms. Practically nothing. The words and ideas are so strange at first that my brain has no way to categorize them, no place to put them. So be it. I now treat that first encounter as a casual walk-through, a meet-and-greet, and put in just twenty minutes. I know that in round two (twenty minutes) I’ll get more traction, not to mention round three (also twenty minutes). I haven’t used any more time, but I remember more.

  By the 1990s, after its long incubation period in the lab, the spacing effect had grown legs and filled out—and in the process showed that it had real muscle. Results from classroom studies continued to roll in: Spaced review improves test scores for multiplication tables, for scientific definitions, for vocabulary. The truth is, nothing in learning science comes close in terms of immediate, significant, and reliable improvements to learning. Still, “spacing out” had no operating manual. The same questions about timing remained: What is the optimal study interval given the test date? What’s the timing equation? Does one exist?

  • • •

  The people who have worked hardest to turn the spacing effect into a practical strategy for everyday learning have one thing in common: They’re teachers, as well as researchers. If students are cramming and not retaining anything, it’s not all their fault. A good class should make the material stick, and spaced review (in class) is one way to do that. Teachers already do some reviewing, of course, but usually according to instinct or as part of standard curriculum, not guided by memory science. “I get sick of people taking my psych intro class and coming back next year and not remembering anything,” Melody Wiseheart, a psychologist at York University in Toronto, told me. “It’s a waste of time and money; people pay a lot for college. As a teacher, too, you want to teach so that people learn and r
emember: That’s your job. You certainly want to know when it’s best to review key concepts—what’s the best time, given the spacing effect, to revisit material? What is the optimal schedule for students preparing for a test?”

  In 2008, a research team led by Wiseheart and Harold Pashler, a psychologist at the University of California, San Diego, conducted a large study that provided the first good answer to those questions. The team enrolled 1,354 people of all ages, drawn from a pool of volunteers in the United States and abroad who had signed up to be “remote” research subjects, working online. Wiseheart and Pashler’s group had them study thirty-two obscure facts: “What European nation consumes the most spicy Mexican food?”: Norway. “Who invented snow golf?”: Rudyard Kipling. “What day of the week did Columbus set sail for the New World in 1492?”: Friday. “What’s the name of the dog on the Cracker Jack box?”: Bingo. Each participant studied the facts twice, on two separate occasions. For some, the two sessions were only ten minutes apart. For others, the interval was a day. For still another group, it was a month. The longest interval was six months. The researchers also varied the timing of the final exam. In total, there were twenty-six different study-test schedules for the researchers to compare.

  The researchers compared all twenty-six different study schedules, and calculated the best intervals given different test dates. “To put it simply, if you want to know the optimal distribution of your study time, you need to decide how long you wish to remember something,” Wiseheart and Pashler’s group wrote. The optimal interval ranges can be read off a simple chart:

  Have a close look. These numbers aren’t exact; there’s wiggle room on either side. But they’re close. If the test is in a week, and you want to split your study time in two, then do a session today and tomorrow, or today and the day after tomorrow. If you want to add a third, study the day before the test (just under a week later). If the test is a month away, then the best option is today, a week from today (for two sessions); for a third, wait three more weeks or so, until a day before the test. The further away the exam—that is, the more the time you have to prepare—the larger the optimal interval between sessions one and two. That optimal first interval declines as a proportion of the time-to-test, the Internet study found. If the test is in a week, the best interval is a day or two (20 to 40 percent). If it’s in six months, the best interval is three to five weeks (10 to 20 percent). Wait any longer between study sessions, and performance goes down fairly quickly. For most students, in college, high school, or middle school, Wiseheart told me, “It basically means you’re working with intervals of one day, two days, or one week. That should take care of most situations.”