How We Learn

by Benedict Carey

  In essence, Spitzer conducted what was, and probably still is, the largest pop quiz experiment in history. The students had no idea that the quizzes were coming, or when. And each group got hit with quizzes at different times. Group 1 got one right after studying, then another a day later, and a third three weeks later. Group 6 didn’t take their first quiz until three weeks after reading the passage. Again, the time the students had to study was identical. So were the questions on the quizzes.

  Yet the groups’ scores varied widely, and a pattern emerged.

  The groups that took pop quizzes soon after reading the passage—once or twice within the first week—did the best on a final exam given at the end of two months, getting about 50 percent of the questions correct. (Remember, they’d studied their peanut or bamboo article only once.) By contrast, the groups who took their first pop quiz two weeks or more after studying scored much lower, below 30 percent on the final. Spitzer showed not only that testing is a powerful study technique, he showed it’s one that should be deployed sooner rather than later.

  “Immediate recall in the form of a test is an effective method of aiding the retention of learning and should, therefore, be employed more frequently,” he concluded. “Achievement tests or examinations are learning devices and should not be considered only as tools for measuring achievement of pupils.”

  For lab researchers focused on improving retention, this finding should have rung a bell, and loudly. Recall, for a moment, Ballard’s “reminiscence” from chapter 2. The schoolchildren in his “Wreck of the Hesperus” experiment studied the poem only once but continued to improve on subsequent tests given days later, remembering more and more of the poem as time passed. Those intervals between studying (memorizing) the poem and taking the tests—a day later, two days, a week—are exactly the ones that Spitzer found most helpful for retention. Between them, Gates and Spitzer had demonstrated that Ballard’s young students improved not by some miracle but because each test was an additional study session. Even then, after Spitzer published his findings in The Journal of Educational Psychology, the bell didn’t sound.

  “We can only speculate as to why,” wrote Henry Roediger III and Jeffrey Karpicke, also then at Washington University, in a landmark 2006 review of the testing effect, as they called it. One possible reason, they argued, is that psychologists were still primarily focused on the dynamics of forgetting: “For the purpose of measuring forgetting, repeated testing was deemed a confound, to be avoided.” It “contaminated” forgetting, in the words of one of Spitzer’s contemporaries.

  Indeed it did, and does. And, as it happens, that contamination induces improvements in thinking and performance that no one predicted at the time. More than thirty years passed before someone picked up the ball again, finally seeing the possibilities of what Gates and Spitzer had found.

  That piece of foolscap Winston Churchill turned in, with the smudges and blots? It was far from a failure, scientists now know—even if he scored a flat zero.

  • • •

  Let’s take a breather from this academic parsing of ideas and do a simple experiment, shall we? Something light, something that gets this point across without feeling like homework. I’ve chosen two short passages from one author for your reading pleasure—and pleasure it should be, because they’re from, in my estimation, one of the most savage humorists who ever strode the earth, however unsteadily. Brian O’Nolan, late of Dublin, was a longtime civil servant, crank, and pub-crawler who between 1930 and 1960 wrote novels, plays, and a much beloved satirical column for The Irish Times. Now, your assignment: Read the two selections below, four or five times. Spend five minutes on each, then put them aside and carry on with your chores and shirking of same. Both come from a chapter called “Bores” in O’Nolan’s book The Best of Myles:

  Passage 1: The Man Who Can Pack

  This monster watches you try to stuff the contents of two wardrobes into an attaché case. You succeed, of course, but have forgotten to put in your golf clubs. You curse grimly but your “friend” is delighted. He knew this would happen. He approaches, offers consolation and advises you to go downstairs and take things easy while he “puts things right.” Some days later, when you unpack your things in Glengariff, you find that he has not only got your golf clubs in but has included your bedroom carpet, the kit of the Gas Company man who has been working in your room, two ornamental vases and a card-table. Everything in view, in fact, except your razor. You have to wire 7 pounds to Cork to get a new leather bag (made of cardboard) to get all this junk home.

  Passage 2: The Man Who Soles His Own Shoes

  Quite innocently you complain about the quality of present-day footwear. You wryly exhibit a broken sole. “Must take them in tomorrow,” you say vaguely. The monster is flabbergasted at this passive attitude, has already forced you into an armchair, pulled your shoes off and vanished with them into the scullery. He is back in an incredibly short space of time and restored your property to you announcing that the shoes are now “as good as new.” You notice his own for the first time and instantly understand why his feet are deformed. You hobble home, apparently on stilts. Nailed to each shoe is an inch-thick slab of “leather” made from Shellac, saw-dust and cement.

  Got all that? It’s not The Faerie Queene, but it’ll suffice for our purposes. Later in the day—an hour from now, if you’re going with the program—restudy Passage 1. Sit down for five minutes and reread it a few more times, as if preparing to recite it from memory (which you are). When the five minutes are up, take a break, have a snack, and come back to Passage 2. This time, instead of restudying, test yourself on it. Without looking, write down as much of it as you can remember. If it’s ten words, great. Three sentences? Even better. Then put it away without looking at it again.

  The next day, test yourself on both passages. Give yourself, say, five minutes on each to recall as much as you can.

  So: Which was better?

  Eyeball the results, counting the words and phrases you remembered. Without being there to look over your shoulder and grade your work, I’m going to hazard a guess that you did markedly better on the second passage.

  That is essentially the experimental protocol that a pair of psychologists—Karpicke, now at Purdue, and Roediger—have used in a series of studies over the past decade or so. They’ve used it repeatedly, with students of all ages, and across a broad spectrum of material—prose passages, word pairs, scientific subjects, medical topics. We’ll review one of their experiments, briefly, just to be clear about the impact of self-examination. In a 2006 study, Karpicke and Roediger recruited 120 undergraduates and had them study two science-related passages, one on the sun and the other on sea otters. They studied one of the two passages twice, in separate seven-minute sessions. They studied the other one once, for seven minutes, and in the next seven-minute session were instructed to write down as much of the passage as they could recall without looking. (That was the “test,” like we just did above with the O’Nolan passages.) Each student, then, had studied one passage two times—either the sea otters, or the sun—and the other just once, followed by a free recall test on it.

  Karpicke and Roediger split the students into three groups, one of which took a test five minutes after the study sessions, one that got a test two days later, and one that tested a week later. The results are easily read off the following graph:

  There are two key things to take away from this experiment. First, Karpicke and Roediger kept preparation time equal; the students got the same amount of time to try to learn both passages. Second, the “testing” prep buried the “study” prep when it really mattered, on the one-week test. In short, testing does not = studying, after all. In fact, testing > studying, and by a country mile, on delayed tests.

  “Did we find something no one had ever found before? No, not really,” Roediger told me. Other psychologists, most notably Chizuko Izawa, had shown similar effects in the 1960s and ’70s at Stanford University. “People had noticed testi
ng effects and gotten excited about them. But we did it with different material than before—the prose passages, in this case—and I think that’s what caught people’s attention. We showed that this could be applied to real classrooms, and showed how strong it could be. That’s when the research started to take off.”

  Roediger, who’s contributed an enormous body of work to learning science, both in experiments and theory, also happens to be one of the field’s working historians. In a review paper published in 2006, he and Karpicke analyzed a century’s worth of experiments, on all types of retention strategies (like spacing, repeated study, and context), and showed that the testing effect has been there all along, a strong, consistent “contaminant,” slowing down forgetting. To measure any type of learning, after all, you have to administer a test. Yet if you’re using the test only for measurement, like some physical education push-up contest, you fail to see it as an added workout—itself making contestants’ memory muscles stronger.

  The word “testing” is loaded, in ways that have nothing to do with learning science. Educators and experts have debated the value of standardized testing for decades, and reforms instituted by President George W. Bush in 2001—increasing the use of such exams—only inflamed the argument. Many teachers complain of having to “teach to the test,” limiting their ability to fully explore subjects with their students. Others attack such tests as incomplete measures of learning, blind to all varieties of creative thinking. This debate, though unrelated to work like Karpicke and Roediger’s, has effectively prevented their findings and those of others from being applied in classrooms as part of standard curricula. “When teachers hear the word ‘testing,’ because of all the negative connotations, all this baggage, they say, ‘We don’t need more tests, we need less,’ ” Robert Bjork, the UCLA psychologist, told me.

  In part to soften this resistance, researchers have begun to call testing “retrieval practice.” That phrase is a good one for theoretical reasons, too. If self-examination is more effective than straight studying (once we’re familiar with the material), there must be reasons for it. One follows directly from the Bjorks’ desirable difficulty principle. When the brain is retrieving studied text, names, formulas, skills, or anything else, it’s doing something different, and harder, than when it sees the information again, or restudies. That extra effort deepens the resulting storage and retrieval strength. We know the facts or skills better because we retrieved them ourselves, we didn’t merely review them.

  Roediger goes further still. When we successfully retrieve a fact, he argues, we then re-store it in memory in a different way than we did before. Not only has storage level spiked; the memory itself has new and different connections. It’s now linked to other related facts that we’ve also retrieved. The network of cells holding the memory has itself been altered. Using our memory changes our memory in ways we don’t anticipate.

  And that’s where the research into testing takes an odd turn indeed.

  • • •

  What if you somehow got hold of the final exam for a course on Day 1, before you’d even studied a thing? Imagine it just appeared in your inbox, sent mistakenly by the teacher. Would having that test matter? Would it help you prepare for taking the final at the end of the course?

  Of course it would. You’d read the questions carefully. You’d know what to pay attention to and what to study in your notes. Your ears would perk up anytime the teacher mentioned something relevant to a specific question. If you were thorough, you’d have memorized the correct answer to every item before the course ended. On the day of that final, you’d be the first to finish, sauntering out with an A+ in your pocket.

  And you’d be cheating.

  But what if, instead, you took a test on Day 1 that was comprehensive but not a replica of the final exam? You’d bomb the thing, to be sure. You might not be able to understand a single question. And yet that experience, given what we’ve just learned about testing, might alter how you subsequently tune into the course itself during the rest of the term.

  This is the idea behind pretesting, the latest permutation of the testing effect. In a series of experiments, psychologists like Roediger, Karpicke, the Bjorks, and Kornell have found that, in some circumstances, unsuccessful retrieval attempts—i.e., wrong answers—aren’t merely random failures. Rather, the attempts themselves alter how we think about, and store, the information contained in the questions. On some kinds of tests, particularly multiple-choice, we learn from answering incorrectly—especially when given the correct answer soon afterward.

  That is, guessing wrongly increases a person’s likelihood of nailing that question, or a related one, on a later test.

  That’s a sketchy-sounding proposition on its face, it’s true. Bombing tests on stuff you don’t know sounds more like a recipe for discouragement and failure than an effective learning strategy. The best way to appreciate this is to try it yourself. That means taking another test. It’ll be a short one, on something you don’t know well—in my case, let’s make it the capital cities of African nations. Choose any twelve and have a friend make up a simple multiple-choice quiz, with five possible answers for each nation. Give yourself ten seconds on each question; after each one, have your friend tell you the correct answer.

  Ready? Put the smartphone down, close the computer, and give it a shot. Here are a few samples:

  BOTSWANA:

  • Gaborone

  • Dar es Salaam

  • Hargeisa

  • Oran

  • Zaria

  (Friend: “Gaborone”)

  GHANA:

  • Huambo

  • Benin

  • Accra

  • Maputo

  • Kumasi

  (Friend: “Accra”)

  LESOTHO:

  • Lusaka

  • Juba

  • Maseru

  • Cotonou

  • N’Djamena

  (Friend: “Maseru”)

  And so on. You’ve just taken a test on which you’ve guessed, if you’re anything like me, mostly wrong. Has taking that test improved your knowledge of those twelve capitals? Of course it has. Your friend gave you the answers after each question. Nothing surprising there.

  We’re not quite done, though. That was Phase 1 of our experiment, pretesting. Phase 2 will be what we think of as traditional studying. For that, you will need to choose another twelve unfamiliar nations, with the correct answer listed alongside, and then sit down and try to memorize them. Nigeria—Abuja. Eritrea—Asmara. Gambia—Banjul. Take the same amount of time—two minutes—as you took on the multiple-choice test. That’s it. You’re done for the day.

  You have now effectively studied the capital cities of twenty-four African nations. You studied the first half by taking a multiple-choice pretest. You studied the other half the old-fashioned way, by straight memorization. We’re going to compare your knowledge of the first twelve to your knowledge of the second twelve.

  Tomorrow, take a multiple-choice test on all twenty-four of those nations, also with five possible choices under each nation. When you’re done, compare the results. If you’re like most people, you scored 10 to 20 percent higher on the countries in that first group, the ones where you guessed before hearing the correct answer. In the jargon of the field, your “unsuccessful retrieval attempts potentiated learning, increasing successful retrieval attempts on subsequent tests.”

  In plain English: The act of guessing engaged your mind in a different and more demanding way than straight memorization did, deepening the imprint of the correct answers. In even plainer English, the pretest drove home the information in a way that studying-as-usual did not.

  Why? No one knows for sure. One possible explanation is that pretesting is another manifestation of desirable difficulty. You work a little harder by guessing first than by studying directly. A second possibility is that the wrong guesses eliminate the fluency illusion, the false impression that you knew the capital of Eritrea because yo
u just saw or studied it. A third is that, in simply memorizing, you saw only the correct answer and weren’t thrown off by the other four alternatives—the way you would be on a test. “Let’s say you’re studying capitals and you see that Australia’s is Canberra,” Robert Bjork told me. “Okay, that seems easy enough. But when the exam question appears, you see all sorts of other possibilities—Sydney, Melbourne, Adelaide—and suddenly you’re not so sure. If you’re studying just the correct answer, you don’t appreciate all the other possible answers that could come to mind or appear on the test.”

  Taking a practice test provides us something else as well—a glimpse of the teacher’s hand. “Even when you get wrong answers, it seems to improve subsequent study,” Robert Bjork added, “because the test adjusts our thinking in some way to the kind of material we need to know.”

  That’s a good thing, and not just for us. It’s in the teacher’s interest, too. You can teach facts and concepts all you want, but what’s most important in the end is how students think about that material—how they organize it, mentally, and use it to make judgments about what’s important and what’s less so. To Elizabeth Bjork, that seemed the best explanation for why a pretest would promote more effective subsequent studying—it primes students to notice important concepts later on. To find out, she decided to run a pretesting trial in one of her own classes.

  Bjork decided to start small, in her Psychology 100B class at UCLA, on research methods. She wouldn’t give a comprehensive prefinal on the first day of class. “It was a pilot study, really, and I decided to give the pretests for three individual lectures,” she said. “The students would take each pretest a day or two before each of those lectures; we wanted to see whether they remembered the material better later.”

  She and Nicholas Soderstrom, a postdoctoral fellow, designed the three short pretests to have forty questions each, all multiple-choice. They also put together a cumulative exam to be given after the three lectures. The crucial question they wanted to answer was: Do students comprehend and retain pretested material better and longer than they do material that’s not on a pretest but is in the lectures? To answer that, Bjork and Soderstrom did something clever on the final exam. They filled it with two kinds of questions: those that were related to the pretest questions and those that were not. “If pretesting helps, then students should do better on related questions during a later exam than on material we covered in the lectures but was not pretested,” Bjork said. This is analogous to the African nation test we devised above. The first twelve capitals were “pretested”; the second twelve were not—they were studied in the usual way. By comparing our scores on the first twelve to the second twelve, on a comprehensive test of all twenty-four, we could judge whether pretesting made any difference.