The Invisible Gorilla: And Other Ways Our Intuitions Deceive Us

by Christopher Chabris

  The Fault Lies Not in Our Confidence, But in Our Love of Confidence

  In the hit Fox TV show House, M.D., Dr. Gregory House and his medical fellows encounter one rare case after another, solving each one by the end of an episode, after first testing several false leads. House, like many other television doctors, is overbearingly confident and self-assured. He has an uncanny ability to diagnose rare disorders that others miss. Although the character of Dr. House is fictional, Dr. Jim Keating plays the same role in his work at St. Louis Children’s Hospital. Like House, he solves the cases that nobody else can. Unlike House, he is gregarious, friendly, quick to laugh, and willing to admit when he doesn’t know the answer. Dr. Keating runs a clinic for infants and children with undiagnosed (and often undiagnosable) problems. Dr. Keating typically sees patients only after they have been to many other doctors and specialists and undergone countless tests. He’s called in as a last resort—somebody who might be able to see what everyone else missed.

  As you might expect, Dr. Keating has an impressive educational pedigree—undergraduate and medical degrees from Harvard, specialties in pediatrics, pediatric critical care, and pediatric gastroenterology, a master’s in epidemiology and biostatistics from London, and a stint in Vietnam when he treated civilians during the war, even diagnosing one patient with bubonic plague. Only after accumulating decades of experience across a broad range of medical subspecialties did he start the diagnostic clinic that he’s been running for more than ten years now. Now that he is in his early seventies, he told Dan, “It was time not to be doing all of those things. The diagnostic center fits well because I have the breadth of experience with a whole range of problems and the confidence that comes from doing clinical medicine intensely with patients.”

  Keating recognizes the role that confidence plays in medicine. “Doctors need to have some level of confidence to be able to interact with patients and everybody else, the nurses … In the emergency room, when everything is happening at once and the patient’s in shock, I like to hear a voice that’s steady and calm.” Patients trust doctors, perhaps more than they should, and that trust reinforces the confidence that doctors already have. As Keating puts it, “When people go to the doctor, they often believe that the doctor has an ability to make the right decisions for them. That goes beyond the scientific reality. They trust your decision-making more than their own. That’s a problem because it encourages doctors to not be honest about what they know and what they don’t know. It builds your ego to have people think that you know.”

  In medicine, the confidence cycle is self-perpetuating. Doctors learn to speak with confidence as part of their training process (of course, there may also be a tendency for inherently confident people to become doctors). Then patients, mistaking confidence for competence, treat doctors more as priests with divine insight than as people who might not know as much as they profess to. This adulation in turn reinforces the behavior of doctors, leading them to be more confident. The danger comes when confidence gets too far ahead of knowledge and ability. As Keating notes, “Equanimity is something we should aspire to, but we ought to get there by building skills, and it should always have a ‘not sure’ component to it so you can continue to learn. There’s still a lot of room for humility in our profession.” Doctors have to be able to listen to the evidence, admit when they don’t know, and learn from their patients. Not all of them are able to overcome their overconfidence.

  Psychology professor Seth Roberts of Berkeley described the experience of being told by his doctor that he had a small hernia and that he needed surgery. Roberts asked the surgeon whether the risk of side effects from anesthesia and surgery, as well as the costs in time and money, justified the benefits of correcting a “problem” that wasn’t actually bothering him at all. Yes, he was told, there are clinical trials showing the value of the surgery, and you can find them easily online. Roberts couldn’t find them, nor could his mother, a former medical-school librarian. The surgeon insisted that the studies existed and promised to find and send them. They never arrived. We don’t have any special insight into whether the surgery was a good idea for Roberts—it might or might not have been. Our focus is on the surgeon’s extreme confidence that her decision was not only correct, but justified by clinical trials. Even after learning that an experienced medical researcher couldn’t find this evidence, she continued to insist on its existence.38

  An obdurate certainty in the face of conflicting evidence is perhaps the best indication that you need a different doctor. The best doctors show a range of confidence—they admit when they don’t know and are more confident when they do know. Doctors who willingly consult people with greater knowledge than their own are likely to provide much better care than those who think they can handle any situation on their own. When Dan met with potential pediatricians for his son, one of the first things he mentioned was that his own father is a pediatrician. He then gauged their reactions. Did they seem to be threatened by this fact? Did they express willingness to take input from other doctors, including Dan’s father? Dr. Keating advises looking for the following trait in a doctor: “They need to be able to say ‘I don’t know’ and mean it.”

  Adopting this strategy for evaluating doctors requires consciously overriding our tendency to trust that confidence corresponds to knowledge—to assume that doctors who express certainty in their knowledge are better than those who express doubt. A study conducted in 1986 at the University of Rochester demonstrates the power of this misguided assumption.39 The researchers asked patients who were waiting for their own appointments to view a videotape of a simulated meeting between a doctor and a patient and to rate their satisfaction with the doctor. The patient had a heart murmur and had been told by his dentist that he should talk to his doctor about possibly taking antibiotics before having oral surgery (taking antibiotics prior to dental surgery is a common step to prevent heart valve infections in people with heart disease).

  In the video, the doctor took a history, performed a physical exam, confirmed the existence of a heart problem, and wrote a prescription for antibiotics. In some versions of the tape, the doctor expressed no uncertainty whatsoever about the diagnosis or treatment. In other versions the doctor acknowledged uncertainty about the need for antibiotics but prescribed them nonetheless. In one of these tapes, the doctor just said, “You have nothing to lose,” and went ahead with the prescription. In another, he consulted a reference book before writing the prescription. The patients viewing these videos found the confident doctors most satisfying, and they rated the one who looked in a book to be the least satisfying of all. At least in medicine, an expert is evidently expected to have all relevant knowledge stored in memory; consulting a reference is even worse than effectively saying “what the hell” and charging ahead.

  Recall Chris’s encounter with the doctor who diagnosed and treated his Lyme disease. This doctor would have received the lowest rating from the subjects in the videotape study, and at the time, Chris probably would have given her a low rating too. But he filled his prescription, took all of the antibiotic as directed, and was cured in short order. Looking back, he realizes that the doctor had the self-awareness to know the limits of her knowledge and the true competence to look information up rather than charge ahead with a decision in a false show of bravado.

  Doctors who express doubt are probably more self-aware than those who don’t, but people rarely notice that sign of actual competence in an expert. Instead, we focus on personality and appearances. A number of studies show that patients are more likely to trust and confide in doctors who are dressed formally and wear a white lab coat than those who dress more casually.40 Yet the worst doctor is just as able to put on a lab coat as the best doctor, so what doctors wear should have no bearing on our estimation of their abilities.

  The self-help literature focuses extensively on the importance of appearing confident. Rightly so: You will persuade more people and consequently you will have more success (at least in the shor
t term) if you present your ideas confidently. If your goal is to convince patients to accept your diagnoses without questioning, by all means wear a lab coat. Faking confidence can be beneficial (although those who can convincingly feign confidence are likely to be fairly confident people to begin with). Unfortunately, if everyone takes the advice of the self-help books and “fakes it till they make it,” the already limited signaling value of confidence will be further eroded, which will make the illusion of confidence even more dangerous. In the extreme, we will be relying on something that has no predictive validity rather than something that—at present—at least occasionally does improve our judgments. Increasing your own confidence might help you, but at the cost of hurting all of us.

  One question still remains: Why do we tend to trust the pronouncements of confident doctors more than those of more hesitant doctors? One reason is self-knowledge. When we know more about a topic, we tend to be more confident in our judgments about it. (As we mentioned earlier, our confidence increases as we gain skill, but our over-confidence decreases.) When dealing with people we know well, we can judge whether their confidence is high or low for them. With knowledge of the range of confidence someone exhibits, you can use confidence as a reasonable predictor of that person’s knowledge; just like you, people generally act more confident when they know more about a topic and less confident when they know less. For example, if you observe that your close friend is more confident about his ability to write a good wedding toast than about his ability to fix a flat tire, you can reasonably infer that he is better at being a best man than at repairing cars.

  The problem, though, is that confidence is also a personality trait, which means that the baseline level of confidence people express can vary dramatically from one person to the next. If you don’t know how much confidence someone expresses across a range of situations, you have no way to judge whether the confidence you see at any particuiar moment reflects their knowledge or personality. If during your first encounter with someone, he expressed confidence in his ability to give wedding toasts, you would have no way to know whether he is truly skilled at giving toasts or whether he is just confident in general. If he is a confident person but an inexperienced toast-giver, then his confidence level would likely be even higher in a different area where he actually had some expertise.

  We all encounter hundreds or even thousands of people whom we don’t know well, but whose confidence we can observe—and draw conclusions from. For such casual acquaintances, confidence is a weak signal. But in a smaller-scale, more communal society, such as the sort in which our brains evolved, confidence would be a much more accurate signal of knowledge and abilities. When close-knit groups and families spend their entire lives together, people come to know almost everyone they ever interact with, and they can adjust for baseline differences in confidence when interpreting other people’s behavior. In these conditions it is entirely reasonable to rely on confidence; if your brother shows more confidence across a range of situations than your sister, you know to discount his bravado when assessing his true competence. Unfortunately, this otherwise useful mechanism becomes a potentially catastrophic everyday illusion when we deal with people we hardly know—like eyewitnesses testifying in court.

  Her Confidence and His Convictions

  In July 1984, Jennifer Thompson was a twenty-two-year-old student at Elon College in North Carolina. She lived in an apartment complex in Burlington, a town about five miles from the college. Late one night, Thompson was startled awake by a noise and saw a black man in her bedroom. He jumped on her and pinned her down by her arms. She screamed. He produced a knife, held it to her throat, and told her that if she made any more noise he would kill her.41

  At first Thompson thought this might be a joke played on her by a friend (a friend with an appalling sense of humor). But she realized it wasn’t once she got a look at the intruder’s face. She said he could take whatever he wanted from her apartment. The man pulled off her underwear, held her legs down, and performed oral sex. Thompson later recalled, “At that point I realized that I was going to be raped. And I didn’t know if this was going to be the end, if he was going to kill me, if he was going to hurt me, and I decided that what I needed to do was to outsmart him.” The attack went on for half an hour, and during that time Thompson turned on lights to get a better look at the rapist. Each time, he ordered her to turn them off right away. The rapist turned on her stereo, and a blue light illuminated his face. Gradually, Thompson assembled a sense of what he looked like. “It was just long enough for me to think, OK, his nose looks this way, his shirt is navy blue, not black.”

  At one point the rapist tried to kiss Thompson. She told him that she would “feel so much more at ease” if he would just put his knife outside the apartment. Surprisingly, he did. Then she asked to get a drink in the kitchen. Once there, she saw that the back door was open and realized that the rapist must have entered the apartment through it. She ran outside and found a neighbor—a professor at Elon who recognized her from campus—who let her in. She fainted and was taken to the hospital.

  Later that same night, less than a mile away, another rape took place. The attacker appeared in the victim’s bedroom, fondled her breasts, and briefly left before returning to rape her. The victim tried to telephone for help, but the line was cut (as it had been at Thompson’s home). The rapist spent as much as thirty minutes in the apartment and left by the front door. The police quickly inferred that the same man committed both crimes.

  Just hours after her ordeal, Jennifer Thompson described her attacker to a police composite artist. Detective Mike Gauldin, who investigated the case, said later that he “had great confidence in her ability to identify her assailant.” According to the bulletin the police issued, the suspect was a “black male with a light complexion, around six feet tall, 170 to 175 pounds … with short hair and a pencil-type mustache.” After publicizing the sketch, Gauldin received a tip that Ronald Cotton, who worked at a nearby seafood restaurant, resembled the person in the picture. Thompson readily picked a photograph of Cotton out of an array that included five other potential suspects, all black males, mentioned by tipsters. Only then did the police tell her that Cotton had a prior conviction for attempted rape. He’d also been convicted for breaking and entering and was said to have touched some of the waitresses at his workplace and made inappropriate comments to them. Thompson later identified Cotton in a “live” lineup, in which the suspects also spoke words she remembered her attacker saying. Ronald Cotton was arrested and imprisoned while he awaited trial.

  During the trial, which took place in January 1985, no definitive physical evidence was offered, nor was it mentioned that the victim of the other rape that night could not identify Cotton (and thus that he was not being tried for that crime). The case was decided on the contrast between Cotton’s shaky and inconsistent alibis for the night of the rape, and Thompson’s confident, consistent identification of Cotton, from the photo array, to the lineup, to the courtroom. Thompson proved to be a compelling witness: She told the jury that during the rape, she had the presence of mind to focus her efforts on memorizing “every single detail on the rapist’s face” in order to make sure he was caught later. “Jennifer, are you absolutely sure that Ronald Junior Cotton is the man?” asked the prosecutor. “Yes,” she replied. The jury convicted Cotton after four hours of deliberation. He was sentenced to life plus fifty years in prison.

  Two years later, Ronald Cotton received a new trial after another prisoner named Bobby Poole told other inmates that he, not Cotton, was the one who had raped Jennifer Thompson. Cotton and Poole looked similar, so much so that some prison workers mistook them for each other. Cotton tricked Poole into posing side by side with him for a photograph, which he sent to his lawyer with a letter describing his claim that Poole was the real rapist. But in court during Cotton’s second trial, Thompson looked at Bobby Poole and said, “I have never seen him in my life. I have no idea who he is.” A more categorical—a
nd confident—statement is hard to imagine. The jury was convinced, and Cotton went back to prison with an even harsher sentence, this time having been convicted of both rapes.

  As the years passed, Thompson gradually managed to put the entire matter behind her. In 1995, ten years after the first trial, she was contacted again by Mike Gauldin and the district attorney, who told her that lawyers for Cotton had requested DNA testing to determine whether he might have been wrongly convicted. DNA recovered from her body at the hospital would be compared with fresh samples provided by Ronald Cotton, Bobby Poole, and Thompson herself. She cooperated enthusiastically, convinced that the test “would allow me to move on once and for all.” But the test proved that Thompson, despite her inner and outward confidence in her memory, had been wrong all along. Cotton had been right in protesting his innocence, as had the jailhouse braggart Poole in boasting of his own guilt—his DNA matched that left by the rapist.

  Thompson accepted Cotton’s innocence, but she was racked with guilt over the responsibility she felt for taking away his freedom. She wrote later that “for so many years, the police officers and the prosecutors told me I was the ‘best witness’ they ever put on the stand; I was ‘textbook.’” Jurors believe confident witnesses, and investigators and prosecutors know this. The U.S. Supreme Court stated that the “level of certainty of the witness” was an important factor in a 1972 case where a victim expressed “no doubt” in court that she recognized her own rapist.42 By contrast, most psychologists who testify as experts on eyewitness memory say that “an eyewitness’s confidence is not a good predictor of his or her identification accuracy.”43 In fact, mistaken eyewitness identifications, and their confident presentation to the jury, are the main cause of over 75 percent of wrongful convictions that are later overturned by DNA evidence.44