How We Learn

by Benedict Carey

  The reinstatement seemed to facilitate better memory but Godden and Baddeley weren’t convinced. They wanted to test-drive reinstatement theory in an environment that was unusual but found in nature, not created by imaginative psychologists. So they had a group of eighteen scuba divers study a list of thirty-six words while submerged twenty feet underwater. The researchers split the divers into two groups. An hour later, one group took a test on the words on dry land, while the others strapped on their equipment and took the test back down under, using a waterproof mike to communicate with those on land doing the scoring. The results indeed depended strongly on test location. The divers who took the test underwater did better than those who took it on land, remembering about 30 percent more words. That’s a lot, and the two psychologists concluded that, “recall is better if the environment of the original learning is reinstated.”

  Maybe the bubbles streaming past the diving mask acted as a cue, accentuating the vowels in the studied words. Maybe it was the rhythmic bellows of the breath in the mouthpiece, or the weight of the tank, plus the sight of swarming nudibranchs. Or the fact that those semantic memories became part of an episodic one (learning while diving). Perhaps all of the above. Reinstatement seemed to work, anyway—for underwater learning.

  The Oban experiment lent comfort and encouragement to what would become a somewhat haphazard exploration of the influence of context on memory. The study materials in these experiments were almost always word lists, or word pairs, and the tests were usually on free recall. In one investigation, for example, people who studied a list of nonsense syllables on blue-gray cards remembered 20 percent more of them on a later test when the test cards were also blue-gray (as opposed to, say, red). In another, students who got exam questions from the same instructor who taught the material did 10 percent better than getting them from a neutral test proctor.

  A psychologist named Steven M. Smith performed some of the most interesting experiments in this area, and it’s worth looking at one of his in detail to see how scientists measure and think about so-called contextual cues. In 1985 Smith, at Texas A&M University, convened a group of fifty-four Psych 101 students—psychologists’ standard guinea pigs—and had them study a list of forty words. He divided the students into three groups. One group studied in silence. Another had a jazz number, Milt Jackson’s “People Make the World Go Around,” playing in the background. The third had Mozart’s Piano Concerto Number 24 in C Minor. The music was on when the subjects arrived in their assigned rooms, and they had no reason to believe it was relevant to the experiment. They spent ten minutes memorizing, and left.

  The students returned to the study room two days later and, without warning, they were given a test to see how many words they could freely recall. This time, Smith changed the tune for many of them. He subdivided the three groups. Some who’d studied to jazz took the test with jazz again; others took it with the Mozart; and others in silence. Likewise for those who studied with Mozart or in silence: They tested either in the same condition, or one of the other two. Nothing else changed.

  Nothing, that is, except their scores.

  Smith found that those who studied with Milt Jackson playing and took the test with the same music recalled twenty-one words on average—twice as many as those who studied with Jackson and took the test to Mozart, or in silence. Similarly, those who studied with Mozart recalled nearly twice as many words with Mozart playing than in silence or with the jazz in the background.

  The punch line: Of those who studied and tested in the same condition, the silence-silence group did the worst. They recalled, on average, about half the words that the jazz-jazz or classical-classical groups did (eleven versus twenty). This is bizarre, and it raised an unexpected question: Could quiet somehow be inhibiting memory? The answer was no. If it had, then those who’d studied with jazz would have done worse taking the test in silence than with Mozart (vice versa, for those who’d studied with classical). They hadn’t.

  What to make of this, then? The higher test scores square with reinstatement theory: The background music weaves itself subconsciously into the fabric of stored memory. Cue up the same music, and more of those words are likely to resurface. The lower scores in the quiet room (after quiet study) are harder to explain. Smith argued that they may be due to an absence of cues to reinstate. The students “do not encode the absence of sound any more than they might encode the absence of any type of stimulus, such as pain or food,” he wrote. As a result the study environment is impoverished, compared to one with music in the background.

  By themselves, experiments like Smith’s and the others don’t tell us how to study, of course. We can’t cue up our own personal soundtrack for an exam, and we certainly can’t retrofit the exam room with the same furniture, wallpaper, and ambience as where we studied. Even if we could, it’s not clear which cues are important or how strong they really are. Still, this research establishes a couple of points that are valuable in developing a study strategy. The first is that our assumptions about learning are suspect, if not wrong. Having something going on in the study environment, like music, is better than nothing (so much for sanctity of the quiet study room).

  The second point is that the experience of studying has more dimensions than we notice, some of which can have an impact on retention. The contextual cues scientists describe—music, light, background colors—are annoyingly ephemeral, it’s true. They’re subconscious, usually untraceable. Nonetheless, it is possible to recognize them at work in our own lives. Think of an instance in which you do remember exactly where and when you learned something. I’m not talking about hearing you made the high school all-star team or got chosen prom queen, either. I mean a factual, academic, semantic memory, like who assassinated Archduke Franz Ferdinand, or how Socrates died and why.

  For me, it’s a late night in 1982, when I was studying for a test in the university’s math building. The buildings were open all night back then, and you could walk in and take a classroom for yourself, spread out, use the blackboard, and no roommates bursting in with beer or other temptations. I did it all the time, and sometimes the only other person in the place was an old guy roaming the halls, disheveled but kindly, a former physics teacher. He would wander into my classroom occasionally and say something like, “Do you know why quartz is used in watches?” I would say no, and he would explain. He was legit, he knew his stuff, and one night he strolled in and asked whether I knew how to derive the Pythagorean theorem using geometric figures. I did not. The Pythagorean theorem, the most famous equation in math, states that adding the square of the two short sides of a right triangle equals the square of the longest side. It existed in my head as a2 + b2 = c2, and I have no idea where I was when I learned that.

  On that night, however, I learned a simple way to derive it—a beautiful thing it is, too—and I still can see what the guy was wearing (blue slacks, up to his chest), hear his voice (barely, he mumbled), and recall precisely where on the board he drew the figure (lower left corner):

  The proof is done by calculating the area of the large square (c squared) and making it equal to the sum of the figures inside: four triangles (area: ½ b x c times 4) plus the area of the little box ((a—b) squared). Try it. Simplify the right side of that equation and watch what you get. I remember it any time I sit alone in some classroom or conference room under dimmed fluorescent lights, like if I’ve arrived first for a meeting. Those cues bring back the memory of that night and the proof itself (although it takes some futzing to get the triangles in place).

  Those are contextual cues, when they’re conscious and visible. The reason I can recall them is that they’re also part of a scene, an autobiographical memory. The science tells us that, at least when it comes to retention of new facts, the subconscious ones are valuable, too. Not always—when we’re submerged in analytical work, they’re negligible—and not necessarily all of them. Only sometimes. So what, though? When it comes to learning, we’ll take any edge we can get.


  I recall something else about that night, too. Normally, when visited by the Ghost of Physics Past, I was not entirely patient. I had work to do. I could do without the lecture about the properties of quartz. That night, though, I’d finished most of my studying and was in an open, expansive mood. I was happy to sit and listen and even hear about how “physics students today, they don’t learn any of this …”

  That mood was part of my “environment,” too, wasn’t it? It had to be—I remember it. I wouldn’t have sat still for the lesson otherwise. If psychologists’ theory about reinstating sights and sounds was correct, then they’d have to show that it applied to internal mental states as well—to jealousy, anxiety, grumpiness, confidence—the entire mix-tape of emotions running through our heads.

  The question was, how?

  • • •

  No one who’s gone through a bad breakup while trying to be a student will doubt the impact of mood on learning. Moods color everything we do, and when they’re extreme they can determine what we remember. The clearest demonstration comes from psychiatry, and the study of bipolar disorder. People with this condition are the extreme athletes of the emotional realm. Their moods cycle between weeks or months of buoyant, manic activity and periods of dark, paralyzing depression, and they know too well that those cycles determine what they remember and what they don’t. “There is a particular kind of pain, elation, loneliness, and terror involved in this kind of madness,” wrote the psychologist Kay Redfield Jamison, who has a diagnosis of bipolar. “When you’re high it’s tremendous. The ideas and feelings are fast and frequent like shooting stars, and you follow them until you find better and brighter ones.… But, somewhere, this changes. The fast ideas are far too fast, and there are far too many; overwhelming confusion replaces clarity. Memory goes.”

  Indeed, researchers showed in a 1974 study that people with bipolar disorder have state-dependent memory: They remember best what happened during manic phases when they’re again manic. And vice versa: When depressed, they recall events and concepts they’d learned when they were down. As the study’s authors put it, “associations or episodic events … can be regenerated more completely in a similar mood state than they can in a different mood state.”

  Yet bipolar is an extraordinary condition, and learning scientists could hardly rely on it to measure the effects of emotion on the rest of us. For most people, moods come and go, coloring our experience rather than defining it. Their impact on memory, if significant at all, would be far weaker than for those with bipolar. And to measure this impact in a rigorous way would mean inducing the same mood in groups of people, reliably and continuously. That’s a tall order, so learning scientists began to focus not on moods per se but on the influence of differing “internal mental states.” Altered states.

  This was the 1970s, after all, when hundreds of thousands of young people were experimenting with consciousness-altering drugs, primarily LSD and marijuana. These recreational users, many of them college students, weren’t interested in the effect of the drugs on their grades—they were enjoying themselves. Yet there were all sorts of rumors about the possible benefits of such substances on learning. Hallucinogens were said to be “mind-expanding,” capable of opening up new ways of thinking about the world. Pot allowed the brain to see connections it hadn’t before (often too many, resulting in late night sessions full of perfect nonsense). Clearly, altered states intensified experience; might they intensify memory?

  The rigorous research into our inner study environment would begin with drugs—the recreational kind. And its primary sponsor was the U.S. government, which, beginning in the early 1970s, funded a string of experiments that might be called the Studying Under the Influence series. By then, a scattering of research reports had already appeared, suggesting that some drugs, like barbiturates and alcohol, could produce so-called state-dependent learning in modest amounts—the “Study Aid” effect. The government-backed researchers wanted to clarify the picture.

  These experiments tended to follow a similar blueprint: Get people high and have them study something; then give them a test hours later—either after getting high again or after ingesting a placebo. We’ll take a close look at one of these studies, to show what serious scientists and serious stoners can do when they put their heads together. In 1975, a research team led by James Eric Eich of the National Institute of Mental Health set out to test the effect of pot on retention (word lists again), as well as learn something about how the drug alters what the brain does with newly studied information. The researchers recruited thirty college students and recent graduates, brought them into their lab, and gave each a joint. Half of the group got a real one and half got a “placebo marijuana cigarette,” which looked and smelled real but delivered no THC, the active drug. “The subjects took deep inhalations, maintained them for 15 seconds, and repeated this process every 60 seconds,” the authors wrote. “The entire cigarette was smoked, with the aid of a holder, usually in about eight minutes.” These were not novices. On average, the participants smoked pot about five times a week. Within twenty minutes, those who smoked the full-strength joint were moderately high, based on their own ratings and physical measures, like pulse rate. Those who smoked the placebo did not show the same physiological changes.

  At this point, all thirty studied.

  They were handed sheets of paper and given a minute and a half to try to commit to memory forty-eight words. The words appeared grouped by category—for example, “A type of vehicle—streetcar, bus, helicopter, train,” or “A musical instrument—cello, organ, trumpet, banjo.” The categories were part of the experimental manipulation. We all look for patterns when trying to memorize a long list of items, bunching together those that look or sound the same, or are somehow related. The scientists wanted to see whether smoking pot influenced these “higher-order” cues we use to retrieve information later on, so they provided the categories. When the ninety seconds were up, the papers were taken away.

  Four hours later, when the effects of the drug had worn off, the participants returned to the lab and had another smoke. Some who’d been given a real joint the first time got a placebo this time around, and vice versa. Others smoked the same type both times. Twenty minutes later, without further study, they took a test.

  Some got a free recall test, writing down as many of the words as they could remember in six minutes. Others took a “cued recall” test, in which they saw the list of categories (“A type of vehicle”) and filled in as many of the words in that category as they could. And sure enough—on the free recall—those who’d smoked a real joint on both occasions remembered 40 percent more than those who got a real one to study and a placebo for the test. The reverse was also true to a lesser extent: Those who initially studied on the placebo joint did better after smoking another placebo, compared to a real joint. The participants’ memories functioned best when their brain was in the same state during study as during testing, high or not high.

  Why? The cued-recall test (the one with the categories) helped provide an answer. The scores on this test were uniformly high, no matter what the students smoked or when. This finding suggests that the brain stores roughly the same number of words when moderately high as when not—the words are in there, either way. Yet it must organize them in a different way for later retrieval. That “retrieval key” comes back most clearly when the brain is in the same state, stoned or sober. The key becomes superfluous, however, when the categories are printed right there on the page. There’s no need for it, because an external one is handy. As the authors wrote, “The accessibility of retrieval cues which have been encoded in drug associated state—such as that produced by a moderate dose of marijuana—appears to depend, in part, on restoration of that state at the time of desired recall.”

  The joint-placebo study also gives us an idea how strong these internal, drug-induced memory cues are. Not so strong. Give someone a real hint—like a category name—and it easily trumps the internal cues. The same thin
g turned out to be true for alcohol and other drugs that these researchers and others eventually studied: Internal and external cues can be good reminders, but they pale next to strong hints.

  The personality of the learning brain that emerges from all this work on external and internal cues is of a shifty-eyed dinner companion. It is tracking the main conversation (the homework assignment, the music notation, the hard facts) and occasionally becoming engaged in it. At the same time, it’s also periodically having a quick look around, taking in the room, sketching in sights and sounds and smells, as well as noting its internal reactions, its feelings and sensations. These features—the background music, a flickering candle, a pang of hunger—help our companion recall points made during the conversation later on, especially when the topic is a new one. Still, a strong hint is better.

  I think about this, again, in terms of the geometric proof of the Pythagorean theorem. Summoning up that late night scene in the math building three decades ago, I can begin to reconstruct the proof, but as I said it takes some futzing to get the triangles in place. However, if someone sketches out just part of the drawing, it all comes back immediately. The strong hint provided by a partial drawing trumps the weaker ones provided by reinstating my learning environment.

  In a world that provided strong hints when needed, this system would be ideal. Just as it would be wonderful if, whenever we had to perform on some test, we could easily re-create the precise environment in which we studied, piping in the same music that was playing, dialing up the same afternoon light, the same mental state—all of the internal and external features that were present when the brain stored the material in the first place.