The Invisible Gorilla: And Other Ways Our Intuitions Deceive Us
Page 3
Why do drivers turn in front of motorcyclists? We favor, at least for some cases, an explanation that appeals to the illusion of attention. People don’t see the motorcyclists because they aren’t looking for motorcyclists. If you are trying to make a difficult left turn across traffic, most of the vehicles blocking your path are cars, not motorcycles (or bicycles, or horses, or rickshaws …). To some extent, then, motorcycles are unexpected. Much like the subjects in our gorilla experiment, drivers often fail to notice unexpected events, even ones that are important. Critically, though, they assume they will notice—that as long as they are looking in the right direction, unexpected objects and events will grab their attention.
How can we remedy this situation? Motorcycle safety advocates propose a number of solutions, most of which we think are doomed to fail. Posting signs that implore people to “look for motorcycles” might lead drivers to adjust their expectations and become more likely to notice a motorcycle appearing shortly after the sign. Yet, after a few minutes of not seeing any motorcycles, their visual expectations will reset, leading them to again expect what they see most commonly—cars. Such advertising campaigns assume that the mechanisms of attention are permeable, subject to influence from our intentions and thoughts. Yet, the wiring of our visual expectations is almost entirely insulated from our conscious control. As we will discuss extensively in Chapter 4, our brains are built to detect patterns automatically, and the pattern we experience when driving features a preponderance of cars and a dearth of motorcycles. In other words, the ad campaign itself falls prey to the illusion of attention.
Suppose that one morning, we told you to watch for gorillas. Then, at some point a week later, you participated in our gorilla experiment. Do you think our warning would have any effect? Most likely not; in the time between the warning and the experiment, your expectations would have been reset by your daily experience of seeing no gorillas. The warning would only be useful if we gave it shortly before showing you the video.
Only when people regularly look for and expect motorcycles will they be more likely to notice. In fact, a detailed analysis of sixty-two accident reports involving cars and motorcycles found that none of the car drivers had any experience riding motorcycles themselves.26 Perhaps the experience of riding a motorcycle can mitigate the effects of inattentional blindness for motorcycles. Or, put another way, the experience of being unexpected yourself might make you better able to notice similar unexpected events.
Another common recommendation to improve the safety of motorcycles is for riders to wear bright clothing rather than the typical attire of leather jacket, dark pants, and boots. The intuition seems right: A yellow jumpsuit should make the rider more visually distinctive and easier to notice. But as we’ve noted, looking is not the same as seeing. You can look right at the gorilla—or at a motorcycle—without seeing it. If the gorilla or motorcycle were physically imperceptible, that would be trivially true—nobody would be surprised if you failed to see a gorilla that was perfectly camouflaged in a scene. What makes the evidence for inattentional blindness important and counterintuitive is that the gorilla is so obvious once you know it is there. So looking is necessary for seeing—if you don’t look at it, you can’t possibly see it. But looking is not sufficient for seeing—looking at something doesn’t guarantee that you will notice it. Wearing conspicuous clothing and riding a brightly colored motorcycle will increase your visibility, making it easier for people who are looking for you to see you. Such bright clothing doesn’t guarantee that you will be noticed, though.
We did not always realize this ourselves. When we first designed the gorilla experiment, we assumed that making the “gorilla” more distinctive would lead to greater detection—of course people would notice a bright red gorilla. Given the rarity of red gorilla suits, we and our colleagues Steve Most (then a graduate student in Dan’s lab and now a professor at the University of Delaware) and Brian Scholl (then a postdoctoral fellow in the psychology department and now a professor at Yale) created a computerized version of the “gorilla” video in which the players were replaced by letters and the gorilla was replaced by a red cross (+) that unexpectedly traversed the display.27 Subjects counted how many times the white letters touched the sides of the display window while ignoring the black letters.
Even jaded researchers like us were surprised by the result: 30 percent of viewers missed the bright red cross, even though it was the only cross, the only colored object, and the only object that moved in a straight path through the display. We thought the gorilla had gone unnoticed, at least in part, because it didn’t really stand out: It was dark-colored, like the players wearing black. Our belief that a distinctive object should “pop out” overrode our knowledge of the phenomenon of inattentional blindness. This “red gorilla” experiment shows that when something is unexpected, distinctiveness does not at all guarantee that we will notice it.
Reflective clothing helps increase visibility for motorcyclists, but it doesn’t override our expectations. Motorcyclists are analogous to the cross in this experiment. People fail to see them, but not just because they are smaller or less distinctive than the other vehicles on the road. They fail to see the motorcycles precisely because they stand out. Wearing highly visible clothing is better than wearing invisible clothing (and less of a technological challenge), but increasing the visual distinctiveness of the rider might be of limited use in helping drivers notice motorcyclists. Ironically, what likely would work to increase detection of motorcycles is to make them look more like cars. For example, giving motorcycles two headlights separated as much as possible, to resemble the visual pattern of a car’s headlights, could well increase their detectability.
There is one proven way to eliminate inattentional blindness, though: Make the unexpected object or event less unexpected. Accidents with bicyclists and pedestrians are much like motorcycle accidents in that car drivers often hit the bikers or walkers without ever seeing them. Peter Jacobsen, a public health consultant in California, examined the rates of accidents involving cars and either pedestrians or bicyclists across a range of cities in California and in a number of European countries.28 For each city, he collected data on the number of injuries or fatalities per million kilometers people traveled by biking and by walking in the year 2000. The pattern was clear, and surprising: Walking and biking were the least dangerous in the cities where they were done the most, and the most dangerous where they were done the least.
Why are motorists less likely to hit pedestrians or bicyclists where there are more people bicycling or walking? Because they are more used to seeing pedestrians. Think of it this way: Would you be safer crossing the pedestrian-clogged streets of London, where drivers are used to seeing people swarm around cars, or the wide, almost suburban boulevards of Los Angeles, where drivers are less accustomed to people popping up right in front of their cars without warning? Jacobsen’s data show that if you were to move to a town with twice as many pedestrians, you would reduce your chance of being hit by a car while walking by one-third.
In one of the most striking demonstrations of the power of expectations,29 Steve Most, who led the “red gorilla” study, and his colleague Robert Astur of the Olin Neuropsychiatry Research Center in Hartford, Connecticut, conducted an experiment using a driving simulator. Just before arriving at each intersection, subjects looked for a blue arrow that indicated which way they should turn, and they ignored yellow arrows. Just as subjects entered one of the intersections, a motorcycle unexpectedly drove right into their path and stopped. When the motorcycle was blue, the same color as the attended direction arrows, almost all of the drivers noticed it. When it was yellow, matching the ignored direction arrows, 36 percent of them hit the motorcycle, and two of them failed to apply their brakes at all! Your moment-to-moment expectations, more than the visual distinctiveness of the object, determine what you see—and what you miss.
Of course, not every automobile-versus-motorcycle collision is entirely the fault of the pe
rson driving the car. In the Ben Roethlisberger accident, the driver and the rider both had green lights, but Roethlisberger was going straight and had the right-of-way. A witness at the scene quoted Martha Fleishman, the driver of the car, as saying, “I was watching him approach but he was not looking at me.”30 Roethlisberger might never have seen Fleishman’s car, even though it was right in front of him. Had he seen it, he might have been able to avoid the accident.
A Hard Landing
NASA research scientist Richard Haines spent much of his career at Ames Research Center, a space and aeronautics think tank in northern California. He is best known publicly for his attempts to document UFO experiences. But in the late 1970s and early 1980s, he and his colleagues Edith Fischer and Toni Price conducted a pioneering study on pilots and information display technologies using a flight simulator.31 Their experiment is important because it is one of the most dramatic demonstrations of looking without seeing. They tested commercial airline pilots who were rated to fly the Boeing 727, one of the most common planes of the time. Commercial airline pilots tend to be among the most experienced and expert pilots—many flew in the military for years, and only the top pilots get to fly the larger commercial planes, where they have responsibility for hundreds of passengers on every flight. The subjects in this study were either first officers or captains who had flown 727s commercially for over one thousand hours.
During the experiment, the pilots underwent extensive training on the use of a “head-up display.” This technology, which was relatively new at the time, displayed much of the critical instrumentation needed to fly and land the simulated 727—altitude, bearing, speed, fuel status, and so on—in video form directly on the windshield in front of the pilots, rather than below or around it as in an ordinary cockpit. Over the course of multiple sessions, the pilots flew a number of simulated landings under a wide range of weather conditions, either with or without the head-up display. Once they were practiced with the simulator, Haines inserted a surprise into one of the landing trials. As the pilots broke through the cloud ceiling and the runway came into view, they prepared for landing as they had on all of the previous trials, monitoring their instruments and the weather conditions to decide whether or not to abort. In this case, however, some of them never saw the large jet on the ground turning onto the runway right in front of them.
Such “runway incursions”—which happen when planes enter runways when they shouldn’t—are among the more common causes of airplane accidents. More than half of the incursions result from pilot error—a pilot taxis into the path of another aircraft. Just as the USS Greeneville was exceptionally unlikely to surface into another ship, most runway incursions present little or no risk of a collision. In fiscal year 2007, the Federal Aviation Administration recorded a total of 370 runway incursions at American airports. In only 24 of them was there a significant potential for a collision, and only 8 of those involved commercial flights. Over the four years from 2004 through 2007, there were a total of 1,353 runway incursions in the United States, 112 of which were classified as serious, and only 1 of which resulted in a collision. That said, the single worst accident in aviation history involved a runway incursion. In 1977, in the Canary Islands, KLM flight 4805 took off down the runway and collided at full speed with Pan Am flight 1736, which was taxiing in the other direction on the same runway. The collision of these two Boeing 747s resulted in 583 deaths.
Although runway incursions are relatively common compared with other aviation accidents, airplane collisions of every sort are exceptionally rare. With only eight runway incursions out of more than 25 million flights in 2007, you would need to take an average of one commercial round-trip flight every day for about three thousand years to have a more than even chance of encountering a serious runway incursion. These incidents are relatively common, with the key word being “relatively.” They are still exceedingly rare—and consequently, they are unexpected.32
What’s surprising about Haines’s flight simulator experiment is that the head-up display should—or at least our intuition suggests that it should—have kept the pilots’ attention on the place where the parked plane was going to appear. They never had to look away from the runway to see their instruments. But two of the pilots using the head-up display would have plowed right through the plane on the runway had the experimenter not aborted the trial. The plane was clearly visible just seconds after the pilots cleared the clouds, and they had about seven more seconds to safely abort their landing. The pilots using the head-up display were also slower to respond, and when they tried to execute a “missed approach” (by pulling up to go around and make a new landing attempt), they were late in doing so. The two who didn’t manage to abort their landings in time were both rated either good or excellent in their simulator flying performance. When the trial was over, Haines asked them whether they saw anything, and both said no. After the experiment, Haines showed the pilots a videotape of the landing with the airplane stationed in their path, and both expressed surprise and concern that they had missed something so obvious. One said, “If I didn’t see [the tape], I wouldn’t believe it. I honestly didn’t see anything on that runway.”33 The plane on the runway was their invisible gorilla—they didn’t expect it to be there, so they never saw it.
Now that we understand that looking is not seeing, we can see that the intuition that a head-up display will enhance our ability to detect unexpected events is wrong. Head-up displays can help in some respects: Pilots get faster access to relevant information from their instruments and need to spend less time searching for that information. In fact, flight performance can be somewhat better with a well-designed head-up display than without one. Using a so-called conformational display, which superimposes a graphical indication of the runway on top of the physical runway visible through the windshield, pilots can fly more precisely.34 Although the head-up display helps pilots perform the task they are trying to accomplish (like landing a plane), it doesn’t help them see what they are not expecting to see, and it might even impair their ability to notice important events in the world around them.
How is it possible that spending more time with the world in view actually reduces our ability to see what is right in front of us? The answer, it seems, stems from our mistaken beliefs about how attention works. Although the plane on the runway was right in front of the pilots, fully in view, the pilots were focusing their attention on the task of landing the plane and not on the possibility of objects on the runway. Unless pilots inspect the runway to see if there are any obstructions, they are unlikely to see something unexpected, such as a plane taxiing onto their landing strip. Air traffic controllers are, after all, supposed to control the traffic to make sure that this doesn’t happen. If a failure to inspect the runway were the only factor in play, though, a head-up display would be no worse than looking away at your instruments and then back to the windshield. After all, in both cases, you could spend the same amount of time ignoring the runway. You either focus attention on the readings on the windshield or focus attention on the instruments surrounding the windshield. But as Haines’s study showed, pilots are slower to notice unexpected events when they are using a head-up display. The problem has to do not as much with the limits on attention—which are in effect regardless of whether the readings are displayed on the windshield or around it—as with our mistaken beliefs about attention.
Hold All Calls, Please
Imagine that you are driving home from work, thinking about what you will do when you get there and everything you left unfinished at the office. Just as you begin to make a left turn across a lane of oncoming traffic, a boy chases a ball into the road in front of you. Would you notice him? Maybe not, you should now be thinking. What if, rather than being lost in thought while you were driving, you were talking on a cell phone? Would you notice then? Most people believe that as long as their eyes are on the road and their hands are on the wheel, they will see and react appropriately to any contingency. Yet extensive research h
as documented the dangers of driving while talking on a phone. Both experimental and epidemiological studies show that the driving impairments caused by talking on a cell phone are comparable to the effects of driving while legally intoxicated.35 When talking on a cell phone, drivers react more slowly to stoplights, take longer to initiate evasive maneuvers, and suffer from generally reduced awareness of their surroundings. In most cases, neither drunk driving nor driving while talking on a cell phone lead to accidents. In part, that is because most driving is predictable and lawful, and even if you aren’t driving perfectly, the other drivers are trying not to hit you. The situations in which such impairments are catastrophic, though, are those that require an emergency reaction to an unexpected event. A slight delay in braking might make the difference between stopping short of the boy in the street and running him over.
For the most part, people are at least familiar with the dangers of talking on a cell phone while driving. We’ve all seen distracted drivers run a stop sign, obliviously veer into another lane, or drive at 30 mph in a 45 mph zone. As columnist Ellen Goodman wrote, “The very same people who use cell phones … are convinced that they should be taken out of the hands of (other) idiots who use them.”36