The Invisible Gorilla: And Other Ways Our Intuitions Deceive Us

by Christopher Chabris

  People involved in the case of Neil Reed and Bobby Knight had sharply different recollections of what happened, but by the time they told their stories to the media in 2000, several years had already passed since the incident. It’s not unreasonable to think that memories can fade and morph over the years, and that they can be influenced by the motives and goals of the rememberer. But what if two people witness the exact same incident, and the delay before they have to describe it is no more than the length of time spent on hold waiting for a 911 operator?9

  Leslie Meltzer and Tyce Palmaffy, a young couple who had met as undergraduates at the University of Virginia, were on their way home from dinner on a summer night in 2002 in Washington, D.C. They drove their Camry north on Fourteenth Street and stopped at a traffic light at the intersection of Rhode Island Avenue.10 Today, it costs upward of $300,000 to buy a small apartment near the Whole Foods supermarket in this area, but then, the neighborhood was still recovering from the effects of race riots and arson that took place in the 1960s. Tyce, a writer on education policy, was driving. His wife, Leslie, who had recently earned a law degree at Yale, was in the passenger seat. To her right, Leslie saw a man riding a bicycle down the sidewalk in their direction. Suddenly, seemingly out of nowhere, another man approached the cyclist, pulled him off the bicycle, and began stabbing him repeatedly. Leslie heard the victim scream. She grabbed her cell phone and dialed 911, only to be greeted by a voice saying, “You have reached the emergency 911 service, all lines are busy, please hold.”

  By the time the 911 operator got on the line, less than a minute had passed, but the assault was over and the light had turned green. Leslie described what she saw as they continued driving with the traffic down Fourteenth Street. The victim was a man in his twenties or thirties riding a bicycle. What about the assailant? He was dressed in jeans, she said. Overhearing her, Tyce interrupted to say that he was wearing sweatpants. They also disagreed about the kind of shirt he was wearing, how tall he was, and even whether he was black or Hispanic. They soon realized that they could agree only on the attacker’s age (twenties), on his weapon (a knife), and on the fact that they were not painting the clearest picture for the operator.

  It is rare to witness exactly the same event, from the same vantage point as someone else, and then try to recall it in the presence of the other witness so soon afterward. Normally, when we observe an event, we store some memory of it. When we later recall the event, we do our best to retrieve our memory and report its contents. The memory seems vivid to us, and we typically lack any specific reason to doubt its accuracy. Had Tyce not been there to hear and correct—or at least contradict—Leslie’s report to the 911 operator, neither would have discovered the stark contradictions between their separate memories. Both were surprised by the extent of the differences. Tyce later recalled realizing right after the unnerving experience “how unbelievably untrustworthy witnesses must be,” an issue we’ll return to later in this book.

  Didn’t They Just Shoot Up His Windshield?

  In a famous scene in the movie Pretty Woman, Julia Roberts is having breakfast with Richard Gere in his hotel room. She picks up a croissant but then takes a bite out of a pancake. In Jagged Edge, Glenn Close’s outfit changes three times during a single courtroom scene. In The Godfather, Sonny’s car is riddled with bullets at a tollbooth, but seconds later its windshield is miraculously repaired. Did you know about these mistakes or others like them? These inadvertent changes, known as continuity errors, are common in movies, in part because of how movies are created. Rarely are movies shot in sequence and in real time from start to finish. They are completed piecemeal, with scenes filmed in an order determined by the actors’ schedules, the availability of physical locations for filming, the cost of hiring the crew at different times, the weather conditions, and many other factors. Each scene is filmed from many different angles, and the final movie is spliced together and put in order in the editing room.

  Just one person on the set is responsible for making sure that everything in each scene matches from one shot to the next.11 That person, known as the script supervisor, is charged with remembering all of the details: what people were wearing, where they were standing, which foot was forward, whether a hand was on a waist or in a pocket, whether an actress was eating a croissant or a pancake, and whether the windshield should be intact or bullet-ridden. If the script supervisor makes a mistake during filming, it’s often impossible to go back and reshoot the scene. And the editor may decide to ignore the error because other aspects of the shot are more important. As a result, some mistakes almost inevitably make it into the final product. That’s why some of the slaves in Spartacus, set during the Roman Empire, can occasionally be seen wearing wristwatches.

  Dozens of books and websites are devoted to cataloging such errors for the curious and obsessed.12 For The Godfather, one site lists forty-two separate continuity errors (plus dozens of other mistakes and anomalies). In part, such lists appeal because of the irony involved: Hollywood, despite spending tens of millions of dollars on a movie, makes clear mistakes that anyone can see. Finding such errors gives the amateur continuity sleuth a feeling of superiority—the filmmakers must have been sloppy not to notice what I can see clearly. And indeed, when you see an error in a movie, it suddenly seems obvious.

  Several years ago, Dateline NBC ran a story on film flubs in movies like Shakespeare in Love and Saving Private Ryan, which had both won Academy Awards and been acclaimed for their editing. Correspondent Josh Mankiewicz revealed an error in Saving Private Ryan in which eight soldiers walked across a field in the distance, even though one had been killed a few minutes earlier in the film, so by then there should have been only seven soldiers. In a disbelieving voice, he said, “This is Steven Spielberg, one of the most talented and most careful moviemakers out there. You’ve got to figure he watched the film several times before it actually got to the theaters. And he didn’t see it?” Later, Mankiewicz asked, “What is it about filmmakers that they can shoot so carefully, many takes, and still miss something so obvious, something the audience can see clearly?” The questions are nearly perfect illustrations of how the illusion of memory operates. Mankiewicz (and his producers) assumed that people have an accurate memory of everything that has happened and that they will automatically notice any discrepancies.13

  When they were in graduate school together at Cornell, Dan and his friend Daniel Levin (now a professor at Vanderbilt University) decided to explore experimentally how well people actually notice such errors in movies.14 With this project “the two Dans” began a long, productive, and ongoing collaboration. For their first study, they made a brief movie of a conversation between two friends, Sabina and Andrea, about a surprise party for their mutual friend Jerome. Sabina sat at a table when Andrea entered the scene. As they talked about the party, the camera cut back and forth between them, sometimes showing a close-up of one of them, and other times showing both of them. After about a minute, the conversation ended and the screen faded to black.

  Imagine being a subject in their experiment. You come to a laboratory room and are told that before you do another task, the experimenters would like you to watch a brief movie and then to answer some detailed questions about it. They advise you to pay close attention and they start the movie. As soon as the movie ends, they hand you a piece of paper that asks, “Did you notice any unusual differences from one shot to the next where objects, body positions, or clothing suddenly changed?” If you are like almost all of the subjects in this experiment, you would answer no—you would not have noticed any of the nine editing mistakes the two Dans intentionally made!15

  These “errors,” which were of the same type that end up in books and websites on film flubs, included plates on the table changing color and a scarf disappearing and reappearing. They were much more obvious than the ones Josh Mankiewicz disparaged in his Dateline report. Yet even when subjects watched the film a second time, now looking for changes, they still noticed, on avera
ge, just two of the deliberate errors. This phenomenon, the surprising failure to notice seemingly obvious changes from one moment to the next, is now known as change blindness—people are “blind” to the changes between what was in view moments before and what is in view now.16 This phenomenon is related to the inattentional blindness we discussed in the last chapter, but it is not the same. Inattentional blindness usually happens when we fail to notice the appearance of something we weren’t expecting to see. The thing we miss, such as a gorilla, is fully visible, right in front of us the entire time. For change blindness, unless we remember that Julia Roberts was eating a croissant, the fact that she is now eating a pancake is unremarkable. Change blindness occurs when we fail to compare what’s there now with what was there before. Of course, in the real world, objects don’t abruptly change into other objects, so checking all the visual details from moment to moment to make sure they haven’t changed would be a spectacular waste of brainpower.

  What is in some ways even more important than a failure to notice changes is the mistaken belief that we should notice them. Daniel Levin cheekily named this misbelief change blindness blindness, because people are blind to the extent of their own change blindness. In one experiment, Levin showed photographs from the Sabina/Andrea conversation to a group of undergraduates, described the film, and pointed out that the plates were red in one shot and white in another. That is, rather than run the change blindness experiment, he explained everything about it, including the intentional “flub.” He then asked these subjects to decide whether or not they would have noticed the change if they had just watched the film without being alerted to its presence. More than 70 percent confidently said that they would have spotted the change, even though in the original study no one actually did! For the disappearing scarf, more than 90 percent said they would have noticed, when again, in the original experiment, no one actually did.17 This is the illusion of memory at work: Most people firmly believe that they will notice unexpected changes, when in fact almost nobody does.

  Now imagine you are in another experiment conducted by the two Dans. You come to the lab and again you are asked to watch a brief silent movie. You are warned that it is really short and that you should pay close attention. The movie shows a person sitting at a desk who gets up and walks toward the camera. The shot then cuts to the hallway and shows a person exiting the door and answering a phone on the wall. He stands still, holding the phone to his ear and facing the camera for about five seconds before the scene fades to black. As soon as the movie ends, you are asked to write a detailed description of what you saw.

  Having just read about the Sabina/Andrea film, you’ve probably guessed that there’s more to this one than just the simple action of answering a phone. When the camera cut from a view of the actor walking toward the doorway to a shot of the actor entering the hall and answering the phone, the original actor was replaced by a different person! Wouldn’t you notice the only actor in a scene being replaced by a different person wearing different clothes, parting his hair the opposite way, and wearing different glasses?

  If you answered yes, you’re still under the illusion of memory. Here is what two subjects wrote after seeing the film:

  Subject 1: A young man with slightly long blond hair and large glasses turned around from the chair at a desk, got up, walked past the camera to a phone in the hallway, spoke into the phone and listened and looked at the camera.

  Subject 2: There was a blond guy with glasses sitting at a desk … not too cluttered but not exactly neat. He looked at the camera, rose, and walked out to the front right of the screen, his blue shirt billowing out a bit on his right over his white with light pattern tee-shirt … went into hallway, picked up phone, said something that didn’t seem to be “hello,” and then stood there looking kind of foolish for a bit.18

  Not a single subject who viewed this video spontaneously reported anything different before and after the change. Even when prompted more specifically with the question, “Did you notice anything unusual about the video?” no subjects reported the change in the actor’s identity or even his clothes from the first shot to the second. In a separate experiment, subjects watched the same video, but with the person-change pointed out to them. They were then asked whether they would have noticed the change had they viewed the video without the warning; 70 percent said they would have, compared with 0 percent who actually did. In this case, when people know about the change in advance, it becomes obvious and they all see it.19 But when they don’t expect the change, they completely miss it.

  Professional Change Detectors

  In most cases, we have almost no feedback about the limits on our ability to spot changes. We are aware only of the changes we do detect, and, by definition, changes we don’t notice cannot modify our beliefs about our change-detection acumen. One group, though, has extensive experience looking for changes to scenes: script supervisors, the professionals responsible for detecting continuity errors when making movies.20 Are they immune to change blindness? If not, do they at least have above-average awareness of the limits on their ability to retain and compare visual information from one moment to the next?

  Trudy Ramirez has been a Hollywood script supervisor for nearly thirty years. She got her start working on commercials and quickly moved up to feature films. She has been the script supervisor on dozens of major movies and television programs, including Total Recall, Basic Instinct, Terminator 2, and Spider-Man 3. Dan spoke with Trudy Ramirez while she was working on the set of Iron Man 2.21 “I have a very good visual memory, but I also take copious notes,” she said. “I know that writing something down that I want to remember will often cement it into my memory.” The key, according to Ramirez, is that script supervisors realize they don’t need to remember everything. They focus on those details and aspects of a scene that matter, and ignore the rest.

  “Most of the time, I will remember what is important to the scene,” she continued. “We know what to look for. We know how to look.” Everyone on a film set has their own area of focus when watching a scene, but script supervisors are trained to look for those aspects of the scene that are central to facilitating the editing of the film. Ramirez noted, “There are points in the action of a scene where you know the editor will most likely cut: when someone sits or stands up, when someone turns around, or when someone comes into or goes out of a room…. You start to develop a sense of how things will cut together, and therefore what is important to notice.” Script supervisors also learn what is important from experience, often painfully: “Over time, we all make tragic continuity errors which train us what to look for—whatever you didn’t notice that you later wished you had trains you to notice that thing or action next time.”22

  So script supervisors are not immune to change blindness. The difference between them and everyone else is that script supervisors get direct feedback that they can and do miss changes. Through their experience of searching for errors and learning about their mistakes, they become less prone to the illusion that they can notice and retain everything around them. Ramirez said, “The one thing this has taught me is that my memory is very fallible. It’s shockingly fallible. You wouldn’t necessarily have any reason to think about how your memory was working unless you were doing something such as script supervising where it’s such an important part of it.” Critically, though, she knows that other people have similar limits. “When I am watching a movie, the more into the story I am, the less I notice things that are out of continuity. If I’m being swept along by the story and I’m involved with the characters, I am much less apt to notice something out of visual continuity. If you’re really into the story, huge continuity errors will go right by you—you’re not looking for those kinds of details…. You can get away with a lot.”

  What does that say about people who make a habit of searching for continuity errors? If people spot continuity errors when watching a film, then the movie may have a bigger problem: It doesn’t engage viewers’ attention enoug
h to keep them from searching for minor changes! Of course, some people will watch a movie multiple times just to look for errors. And if they do that, they are likely to find some. The impossibility of noticing everything is what guarantees the business prospects for books and websites on film flubs.

  Do You Have Any Idea Who You’re Talking To?

  Professor Ulric Neisser, whose research inspired our gorilla experiment, watched the change blindness demonstration in which an actor changed into another person while answering a phone, and he pointed out a possible limitation of all of these studies: They all used videos. He commented that watching video is an inherently passive activity: The action unfolds in front of us, but we do not actively engage with it the way we do when we interact socially with other people. Neisser argued that change blindness might not occur if a person were changed in the middle of a real-world encounter rather than across a cut in a passively viewed motion picture. The two Dans thought Neisser probably was right, that people would notice such a change in the real world, but they decided to run an experiment to test Neisser’s prediction anyway.