Brain Buys

by Dean Buonomano

  The implicit-association effect is a product of both nature and nurture. Nature, because no matter what we learn, it is stored as a web of associations. Nurture, because the specific associations we learn are a product of our environment, culture, and education.

  To explore the effect of culture on our implicit associations Anthony Greeenwald and colleagues asked more than 500,000 people to perform a gender-science implicit association test online.26 The test required people to classify words as “science” words (for example, physics, chemistry) or “liberal arts” words (history, Latin), and intermixed male (boy, father) or female (girl, mother) words according to gender. During one-half of the test the response keys were grouped as science/male versus liberal arts/female, in the other half they were grouped as science/female versus liberal arts/male. In countries in which boys, on average, performed better than girls on a standardized math test, people tended to take longer to respond when the female and science words were assigned the same key—capturing the stereotypical association that men are better in math and physics. In a few countries, such as the Philippines and Jordan, it was the girls who out-scored boys on the standardized science test; in these countries reaction times were less dependent on whether “female” shared the response key with “science” or “liberal arts” (yet the reactions were still a bit slower in the female/science condition). The authors of the study suggest that implicit associations—which is to say how information is laid out in our neural circuits—contribute to the gender differences on standardized tests.

  The above studies raise the question of whether the way information is stored in our brain merely influences the speed with which we can access this information, or whether it actually influences the way we think and behave in the real world. The question is tricky if not impossible to answer. Research by the psychologist Bertram Gawronski and his colleagues explored this issue by testing Italian citizens living in the city of Vicenza, which is home to a U.S. military base.27 The volunteers were asked their views as to whether the government should allow the United States to expand the base, and given an implicit association test in which one-by-one “positive” (joy, lucky) or “negative” (pain, danger) words or pictures of the U.S. base were presented on a computer screen. When a word was presented, subjects had to decide if it was “positive” or “negative,” and when a picture of the U.S. base was shown, they simply had to respond by pressing the assigned key (in half the trials the picture key was shared with the positive words and in the other half with the negative words). For example, the positive word joy might require pressing the left button, and the negative word pain the right, and pictures of the American base could be assigned to the left (positive). The difference in reaction time between when the pictures shared the positive or negative button were taken as a measure of implicit associations—and so presumably reflected whether the military base was more strongly linked to “positive” or “negative” words within an individual’s neural networks. If someone had a strong implicit bias against the expansion of the base it would be expected that his reaction time to the photographs would be longer when the key assigned to pictures was shared with that of the positive words. Here is where things get interesting: some subjects fell into an undecided group during the initial questionnaire, but, one week later, during a second study session, had made up their minds. In this subset of subjects, the implicit association measured during the first test was a reasonable predictor of the opinion reported one week later. These results indicate that the unconscious and automatic associations within our neural networks could in effect reveal someone’s opinion before they were consciously aware of it. The results also support the notion that the structure of our associative nets may indeed influence our opinions and decisions.

  PRIMING BEHAVIOR

  Suppose I’m doing a crossword puzzle and I ask a friend for a thirteen-letter word for “nice,” and he offers “compassionate.” Could this innocuous exchange alter my friend’s behavior, making him a better person for the next few minutes? Would this be a good time to ask if he could lend me money? In short, is it possible to prime someone’s behavior? The cognitive psychologist John Bargh, currently at Yale University, has examined this question by surreptitiously studying people’s behavior after priming them with certain concepts.28 In one study, subjects were asked to perform a task that they thought was a test of language skills. The test consisted of making four-word sentences from five scrambled words. They, her, send, usually, and disturb, for example, could lead to, “They usually disturb her.” In one group the words were weighted toward sentences that reflected rude behavior; in the other group the sentences were biased toward polite phrases (they, her, encourage, see, and usually would lead to “They usually encourage her”). Because the subjects were engaged in making many different sentences, they were probably not consciously aware that they were being subliminally primed with rude or polite words.

  On finishing the word task, the subjects were instructed to seek out the experimenter in a nearby room for further instructions. Unbeknownst to the participants, this request was the key to the entire study. When the subjects approached the doorway of the office, they found the researcher engaged in a conversation. The measure Bargh and his colleagues were interested in was one we have all encountered: how long do we wait before interrupting? Presumably the answer depends on a complex mix of factors, including whether we are patient or impatient by nature, our mood that day, whether we have another appointment, and if we need to go pee. The fascinating results revealed that activating the neural nodes associated with the concept of rudeness or politeness altered the behavior of the participants in the experiment in a predictable fashion. Ten minutes was the maximum time the experimenter would carry on her rehearsed conversation while the subject waited. In the “polite” group, only about 20 percent of subjects interrupted the conversation before 10 minutes had elapsed, whereas 60 percent of subjects primed with “rude” sentences interrupted the conversation within 10 minutes. We have seen that words can prime what people think of (“Africa” primes “zebra”), but this study shows that words can prime the way people feel and behave. It seems that behavioral priming takes place when the activity of nodes not only spreads to other nodes within our semantic nets, but to the parts of the brain governing our decisions and behavior.

  The same adjectives that describe physical characteristics about the world are sometimes also used to characterize people’s personality traits. In English, warm and cold are used to describe whether someone is friendly or unfriendly, respectively. Because we associate hot temperatures with warmth, and in turn associate warmth with being friendly, John Bargh and his colleagues wondered whether hot temperatures might influence whether we judge people as being friendly. They asked volunteers to read a description of someone, and then rate that person’s personality on different traits including some related to being “warm” (generous, social, caring). There were two groups, and the only difference between them was that subjects were asked to hold a cup of hot or iced coffee in the elevator ride up to the experimental room. One would have hoped that our judgment of other people would not be so arbitrary as to be influenced by the physical temperature of a cup held for 20 seconds. Yet the subjects holding the hot coffee did indeed rate the person being profiled as friendlier than did those holding the iced cup.29

  I do not want to give the impression that our behavior and decisions are hopelessly at the mercy of irrelevant factors, such as whether we are holding a hot or cold cup of coffee. Behavioral priming effects are often fairly weak and unlikely to be the major determinant of how people behave. Nevertheless, at least in some situations the effects are reproducible and significant. This fact allows us to conclude that merely accessing information about some concept can influence someone’s behavior, which lends some support to the mainstay of self-help books: the importance of positive thinking and the contribution of attitude on performance.

  Our brain consists of an unimaginably compl
ex tangle of interconnected neurons. Like the links of the World Wide Web the patterns of connections between neurons is anything but random. If we could disentangle our neural circuitry we would see that it has been sculpted by a lifetime of experiences. The structure of these circuits stores our memories, and influences our thoughts and decisions. It follows that manipulating our experiences provides a means to influence our opinions and behavior. Long before Donald Hebb put forth the notion of associative synaptic plasticity, some people implicitly understood that the associative nature of human memory was a vulnerability to be exploited. The simple act of associating the name of a politician with a controversial or negative statement—through massive repetition and media exposure—remains one of the most abused and effective political campaign strategies.30 A single spurious and slanderous headline such as “Is Barack Obama a communist?” will certainly grab your attention but because you already have many links associated with your “Barack Obama” node, that single headline is unlikely to exert much influence on the structure of your neural nets and thus on your opinions—memories that have multiple strong associations in place are more robust. But imagine for a moment that the headline was about a politician you were unfamiliar with, perhaps a long-shot presidential candidate: “Is Jonathan Hazelton a pedophile?” You have no previous associations formed in your memory banks for Jonathan Hazelton, but now one of the first ones is with the word pedophile. Even if the report concluded that he was definitely not a pedophile, Hazelton’s long-shot presidential bid just got much longer. Slander by association—sometimes disguised as journalism—is commonly used to mold public opinion, and it works because it exploits the associative architecture of the human brain. That same architecture, though unparalleled in its ability to store and organize information about a diverse and dynamic world, also sets the stage for our susceptibility to marketing and propaganda, and our proclivity to make irrational decisions, as we will see in the upcoming chapters.

  2

  Memory Upgrade Needed

  She was the most placid, the most adrift in nature’s currents, of the women I have known, or perhaps that is the way I prefer to remember, memory being no less self-serving than our other faculties.

  —John Updike, Toward the End of Time

  On July 29, 1984, Jennifer Thompson, a 22-year-old college student, was raped in her home in the town of Burlington, North Carolina. During the ordeal she made a conscious effort to memorize the face of the man who was raping her; she vowed that if she survived she’d ensure her assailant was caught. Later that same day, she picked out a man named Ronald Cotton from a selection of six photographs. Understandably, immediately after the photo lineup, she sought some feedback from the detective: “Did I do OK?” she asked. He responded, “You did great, Ms. Thompson.” Eleven days later, after picking out Ronald from a physical lineup, she again wondered how she did; the detective told her “We thought that might be the guy. It’s the same person you picked from the photos.” At trial, based almost exclusively on Jennifer’s eyewitness testimony, Ronald was sentenced to life in prison.

  In prison Ronald crossed paths with another African American man who by some accounts resembled Ronald in appearance. The man, Bobby Poole, was from the same area and had also been convicted of rape. Ronald heard that Bobby boasted about raping Jennifer. A few years later Ronald’s case was retried. Based on Jennifer’s testimony, as well as that of an additional victim who had been raped the same night, Ronald was again sentenced to life in prison, despite testimony by another prisoner stating that Bobby had confessed to raping Jennifer. Thanks to Ronald’s persistence, a zealous attorney, and emerging DNA fingerprinting technology, genetic tests were eventually performed. DNA from the second victim matched Bobby Poole’s, and when confronted with the new evidence, he confessed to raping Jennifer, providing information about the case that only the rapist could have known. After an 11-year forced separation from his sick mother and the few loved ones who stood by his side during his ordeal, Ronald was finally released. Jennifer was sickened by the consequences of her mistake, and genuinely bewildered as to how her memory could have betrayed her. Eventually she sought forgiveness from Ronald Cotton. The two slowly became good friends and have campaigned together for reforms in witness interview procedures and the use of eyewitness testimony in trials.1

  CORRUPTED MEMORIES

  As we have seen the associative architecture of human memory makes us prone to certain mistakes, such as falsely remembering a word that was closely related to the actual words on a list. Other types of memory bugs, however, such as the one responsible for Ronald Cotton’s 11-year incarceration are different, in both their causes and consequences. Causes, because they are not solely a product of the associative nature of human memory; and consequences because they can result in tragic life-altering errors.

  Digital memory, whether in the form of a hard drive or a DVD, relies on distinct mechanisms for the storage and retrieval of information—the writing and reading operations are fundamentally different processes. On a hard drive there are separate read and write elements: the first can measure the polarity of a tiny dot of ferromagnetic material, whereas the second can alter the polarity of the magnetic granules. Similarly a DVD player can only retrieve the memory burned into a DVD. The read operation is performed by a laser beam directed onto the surface of the DVD; if the light is reflected back, a “1” has been stored; if not, a “0.” There is no danger whatsoever that retrieving information from a DVD will alter its content; for that, a DVD burner, which has a more powerful laser, is required. In the brain, on the other hand, the read and write operations are not independent; the act of retrieving a memory can alter its content. When Jennifer Thompson was looking at the picture of her potential assailant, she was not simply retrieving an established memory, but melding new and old ones. In particular, the positive feedback from the detective immediately after she picked a suspect likely contributed to the “updating” of her memory. By the time she got to trial, months after the rape, the memory of the rapist was the well-lit and clear image of the man in the photos and lineup, rather than the dark fragmented image from the night of the rape. Jennifer Thompson’s memory betrayed her because it overwrote Bobby Poole’s image with Ronald Cotton’s (Poole’s picture was not in the first lineup).

  Most of us have had the experience of not recognizing someone we have met, or the converse experience of incorrectly believing we’ve seen someone somewhere before. So it seems surprising that the American judicial system has traditionally relied heavily on the accuracy of memories of victims and witnesses. Memory errors that can derail the judicial process are not limited to mistaken identities, but also include incorrect recall of factual information and erroneous judgments about how long an event lasted or when it took place. Take the trial of Andrea Yates, the Texas woman who drowned her five children in a bathtub in 2001. In this case it was the testimony of a psychiatrist that proved incorrect. In court, Andrea Yates stated that voices in her head told her that her children would be tormented in hell forever; but, if she killed them, Satan would be destroyed. Hallucinations featuring Satan fit with the family’s devotion to Scripture, a fact reflected in the name of the five victims: Mary, Luke, Paul, John, and Noah. During her trial a psychiatrist for the prosecution testified that an episode of the TV program Law & Order may have been pertinent to the case, stating “there was a show of a woman with postpartum depression who drowned her children in the bathtub and was found insane, and it was aired shortly before the crime occurred,” implying that the murders may have been premeditated. This testimony may have contributed to the jurors’ rejection of Andrea’s insanity defense, and the sentence of life in prison. It later came to light that the episode the psychiatrist was thinking of was aired after the crime, and differed in some of the details. Trials often take place years after the crime; remembering an episode of a TV program is one thing, correctly remembering the “time stamp” of the memory is a different process all together. You may
recall events related to the O.J. Simpson trial, but did it occur before or after the Atlanta Olympic Games?2 Every computer file is stored along with the date it was created; there is no such time stamping with our memories. It is easy to see how even the most honest witness can generate false recollections that could ultimately prove critical to determining the course of someone else’s life. In the case of Andrea Yates, a retrial was granted on the basis of the erroneous testimony, and the new jury judged her to be insane at the time of the homicides.3

  The psychologist Elizabeth Loftus, now at the University of California in Irvine, has devoted her career to exposing the brain’s propensity to make errors of the type committed in the testimony of Jennifer Thompson and the trial of Andrea Yates. Studying such false memories in the real world is, of course, often impossible because it is difficult to verify what witnesses or victims actually experienced. Indeed, when courts rely on eyewitness testimony it is precisely because of the lack of incontrovertible evidence. To overcome this limitation Loftus and her colleagues developed experiments aimed at simulating some aspects of real-world eyewitness testimony. In a classic study Loftus and colleagues showed 200 students a sequence of 30 slides depicting an automobile accident involving a car at an intersection.4 All subjects saw the same images with one important difference: half saw a stop sign and half saw a yield sign at the intersection where the accident occurred. Immediately after the presentation subjects were asked several questions such as the color of the car. Among these questions one was key to the experiment because it was actually used to plant a false memory: for half the subjects in each of the two groups the question was, “Did another car pass the red car while it was stopped at the stop sign?” while the other half in each group was asked, “Did another car pass the red car while it was stopped at the yield sign?” In other words, half of the subjects were asked a question with misleading information about the sign; the misinformation was fairly subtle because it was not relevant to the question being asked. Twenty minutes after this questionnaire subjects were given a recognition test: pairs of slides were presented and the subjects had to indicate which picture of each pair they had seen before—the crucial pair being when they had to choose between slides with either a stop or yield sign. When the key question had contained consistent information, 75 percent of subjects correctly reported the image they had seen. But when the key question had contained misinformation, only 41 percent correctly chose the slide they actually saw during the initial presentation. Not only did a misleading question dramatically impair memory reliability, but it actually made performance worse than chance: an erroneous question about reality trumped reality.