The Invisible Gorilla: And Other Ways Our Intuitions Deceive Us

by Christopher Chabris

  Just as business authors rarely consider how many companies follow the ideas they champion but still fail, or how many companies succeed with other approaches, people thinking about stories of vaccination and autism do not tally up the number of children who receive vaccines and do not develop autism, who show symptoms before vaccination, or who show symptoms without having been vaccinated. When these numbers are properly taken into account, it becomes clear that children tend to be diagnosed with autism at the same rates and at the same ages regardless of whether or not they received vaccinations.34 The problem is exacerbated by the typical developmental trajectory of cognition and behavior. As any parent knows, development is not a continuous, gradual process. Just as children grow physically in spurts, they develop cognitively in spurts as well. Children with autism are much the same. For long stretches, they might show no improvement, only to show a big change in a short time frame. If parents happen to notice improvement while they are trying some new miracle cure, they will readily associate the treatment with the improvement.35

  Accepting that a perceived cause is illusory can be difficult, and overcoming anecdotes with science and statistics can be even harder. Perhaps the best indication of the powerful hold of these anecdotal hypotheses comes from the emotions they inspire. Offit’s authoritative book on the lack of scientific links between autism and vaccination has an average customer rating of 3.9 on the 1–5 scale at Amazon.com. However, in this case, the average is not typical of the individual reviews. Of the 102 reviewers at the time of this writing, not a single one gives the book the middle rating (three stars), whereas 70 give it the highest possible rating and 25 give it the lowest possible rating!36

  Despite the now-overwhelming evidence that vaccinations are not at all associated with autism, 29 percent of people in our national survey agreed with the statement “vaccines given to children are partly responsible for causing autism.”37 It’s a bit reassuring that all the media attention to this illusory cause hasn’t influenced more people, but science can only claim a partial victory at best. If 29 percent of parents follow through on such beliefs and do not vaccinate their children, population immunity could drop precipitously, leading to widespread measles outbreaks. Moreover, new autism “cures” relying on anecdotal evidence rather than careful experimentation continue to surface and lead parents down dangerous paths. We hope that reading this chapter has given you some immunity to these attempts to exploit the illusion of cause.

  We have explored three ways the illusion of cause can affect us. First, we perceive patterns in randomness, and we interpret these repeating patterns as predictions of future events. Second, we look at events that happen together as having a causal relationship. Finally, we tend to interpret events that happened earlier as the causes of events that happened or appeared to happen later. The illusion of cause has deep roots. We humans are distinct from other primates in our ability to perform “causal inference.” Even young children realize that when one object hits another, it can make the other object move. They can reason about hypothetical causes as well: If an object moved, something must have caused it to move. Our primate relatives generally do not make these inferences, and consequently, they have trouble learning about causes that they can’t see.38 On the timeline of evolution, therefore, the ability to infer the existence of hidden causes is quite recent, and new mechanisms often need refinement. We have no trouble inferring causes—the real trouble is that we are sometimes too good at inferring causes for our own good.

  get smart quick!

  BEFORE THE 2007 NATIONAL FOOTBALL League season, as before every season, the New York Jets made several adjustments to the team. Rookies arrived at training camp, some veterans left the team, other players had to compete for positions on the starting roster, and the playbook was updated. But one change was more unusual: Head Coach Eric Mangini ordered that the stadium loudspeakers play classical music—specifically, compositions by Wolfgang Amadeus Mozart—during team practices. “Mozart’s music and brain waves are very similar, and it stimulates learning,” explained Mangini, a coach known for meticulously preparing his team.1

  Eric Mangini has much company in believing that listening to Mozart can make you smarter. An entrepreneur named Don Campbell trademarked the phrase “The Mozart Effect” and used it to market a series of books and CDs for adults and children alike. Campbell even consults with hospitals on the optimal design of sound systems to maximize the healing powers of music.2 In 1998, Governor Zell Miller persuaded the Georgia legislature to spend public money to issue classical music tapes to all parents of newborn babies in the state. As part of his state-of-the-state speech, he played Beethoven’s “Ode to Joy” to the legislators and asked, “Don’t you feel smarter already?”3 A hospital in Slovakia puts headphones on all of the infants in its nursery, within hours of their birth, to give them a true head start on building their brainpower. “Mozart’s music has a very good effect on the development of the intelligence quotient,” said the doctor who started the practice there.4

  So far, we have discussed several everyday illusions that expose errors in the way people think about their own minds, and we have tried to convince you that these errors can have dramatic consequences for human affairs. We have also suggested ways to minimize the impact these illusions have on your own life. With our understanding of these illusions, we have found it possible—though far from easy—to change our mindset so as to recognize and escape them at least some of the time. But we would all be better off if there were a simple way to overcome everyday illusions, a way to increase our brainpower enough to make the illusions just disappear.

  The illusion of potential leads us to think that vast reservoirs of untapped mental ability exist in our brains, just waiting to be accessed—if only we knew how. The illusion combines two beliefs: first, that beneath the surface, the human mind and brain harbor the potential to perform at much higher levels, in a wide range of situations and contexts, than they typically do; and second, that this potential can be released with simple techniques that are easily and rapidly implemented. The story of the Mozart effect is a perfect illustration of how this illusion can transform a claim with almost no scientific support into a popular legend that fuels multimillion-dollar businesses, so we will begin this chapter by going into it in depth.

  “The Magic Genius of Mozart”

  The Mozart effect burst into public consciousness in October 1993 when Nature, one of the top two scientific journals (the other being Science), published a one-page article by Frances Rauscher, Gordon Shaw, and Katherine Ky under the innocuous title, “Music and Spatial Task Performance.”5 Shaw, a physics professor who had shifted his interests to neuroscience, together with his student Xiaodan Leng, had developed a mathematical theory of how neurons in the brain work together. As a classical music enthusiast, Shaw noticed some similarities between the mathematical structure of classical pieces and the patterns his theory predicted would be found in the electrical activity of neurons. From this perceived similarity, he made the prediction that merely listening to music could enhance the function of one’s brain—but only the right kind of music.6 Shaw believed that Mozart had composed music that would “optimally resonate with the inherent internal neural language,” and that it would have the greatest enhancing effect. As he later wrote, “The magic genius of Mozart perhaps displayed a supreme use of the inherent cortical language in his music.”7

  To help him test his theory, Shaw hired Frances Rauscher, a former concert cellist who had switched her profession to psychology, and together they conducted a simple experiment. Each of thirty-six college students performed three tests taken from a standard IQ test battery: “pattern analysis,” “matrix reasoning,” and “paper folding and cutting.” In the pattern-analysis task, subjects constructed objects out of blocks according to patterns they were given. In the matrix-reasoning task, subjects selected which of several shapes would complete a pattern composed of other abstract shapes. In the paper-folding-and-cutt
ing task, subjects viewed a picture of an origami-like design, with dashed and solid lines showing where one would fold and cut the pattern. Then the subjects chose which of several pictures accurately showed what the paper would look like after being unfolded.

  Before taking these tests, the subjects listened to one of the following recordings: ten minutes of Mozart’s “Sonata for Two Pianos in D Major (K.448),” ten minutes of “relaxation instructions designed to lower blood pressure,” or ten minutes of silence. The sonata is described as “gallant from beginning to end … one of the most profound and mature of all Mozart’s compositions.”8 According to the article, the subjects who did well on one of the tests did well on the others: There were significant correlations among all the tests, just as would be expected for the subparts of an IQ test, or any test of general cognitive ability like the SAT. So Shaw and colleagues combined the three tests into a single measure of what they called “abstract reasoning ability” and transformed it to the scale of IQ scores, which have an average of 100 points for the general population. Then they compared the three listening conditions, and found that the scores after sitting in silence were 110, after listening to relaxation instructions they were 111, and after listening to the Mozart sonata they were 119.

  Thus, listening to Mozart appeared to make the students smarter, by eight to nine IQ points. Although nine points might seem small, it’s not: An average person, who is by definition more intelligent than 50 percent of other people, would be more intelligent than 70 percent of other people after listening to the Mozart sonata. The simple tonic of ten minutes of classical music, if its effects could be harnessed, would propel a typical student past 20 percent of his or her relaxing or silence-enjoying peers, potentially turning Bs into As and failing grades into passing ones.

  The media reported this new scientific finding with enthusiasm. “Mozart Makes You Smarter” read the headline in the Boston Globe. “Listening to Mozart is not only a music lover’s pleasure. It’s a brain tonic,” the article began.9 Less than a year after Rauscher, Shaw, and Ky published their article, music companies started creating new CDs to exploit the publicity, with titles such as Mozart for Your Mind, Mozart Makes You Smarter, and Tune Your Brain with Mozart. Ironically, most of these did not include the K.448 piano sonata that was used in the experiment, but it didn’t matter. Sales ran into the millions.10 In his address to the Georgia state legislature, Zell Miller cited the Rauscher article: “There’s even a study that showed that after college students listened to a Mozart piano sonata for ten minutes, their IQ scores increased by nine points … no one doubts that listening to music, especially at a very early age, affects the spatial-temporal reasoning that underlies math, engineering, and chess.”11

  Subsequent research reports from the Mozart effect team also were covered extensively in the press. Just like the original, these new experiments found dramatic improvements in mental task performance immediately after the Mozart sonata, but not after silence or relaxation.12 Meanwhile, psychologists interested in music and cognition began to examine this discovery, which was intriguing because no previous research had shown that merely listening to music could have such a large effect on mental ability.

  The first independent research group to publish its findings was headed by Con Stough of the University of Auckland in New Zealand.13 They used the same Mozart sonata and silence conditions as in the original study, and added a new one: dance music, specifically ten minutes of “Fake 88 (House Mix)” and “What Can I Say to Make You Love Me? (Hateful Club Mix)” by Alexander O’Neal. Thirty subjects participated in each listening condition and worked on part of the Raven’s Advanced Progressive Matrices test after each one. This test is considered an excellent measure of general intelligence. Stough’s team found that the Mozart group outperformed the control groups by only about one IQ point, not even close to the eight to nine points reported by Rauscher. A one-point difference is small enough that it could easily have arisen just from the random variations in the measures of cognitive abilities, or from accidental differences among the subjects assigned to the Mozart and control groups. Other researchers reported similar experiences.14

  Along with two of his students, Kenneth Steele, a psychology professor at Appalachian State University in North Carolina, tried a Mozart experiment in 1997. They used a “digit span” test, which measures the longest list of digits that you can hold in short-term memory accurately enough to repeat it back, either forward or backward. This test is strongly associated with general intelligence: the smarter you are, the longer your backward digit span. But listening to Mozart had no effect on digit span. Steele tried again the next year, this time copying the design of Rauscher and Shaw’s 1995 follow-up study, which had also produced a large Mozart effect. Steele used the paper-folding task rather than digit span, but again he found no benefits of Mozart.15 The next year the American Psychological Society’s flagship journal, Psychological Science, published these new results under the title “Mystery of the Mozart Effect: Failure to Replicate,” and the society issued a press release headlined “‘Mozart Effect’ De-Bunked.” Almost immediately, the headline was changed to “‘Mozart Effect’ Challenged” after Gordon Shaw threatened the APS with a lawsuit.16

  Steele wrote later that when he started his experiments, he expected to replicate the Mozart effect.17 Indeed, researchers rarely conduct experiments that they think will fail! Experiments can fail for many reasons even when the theory that motivated them is correct. In this case, the theory that listening to Mozart increases cognitive performance could be true, but any particular experiment intended to test the theory could fail to support it because of a variety of errors in design or execution, none of which have anything to do with the correctness of the theory. But after repeated failures to find any cognitive improvement after listening to Mozart, Steele came to believe that there was no Mozart effect to be found.

  The Media and the Aftermath

  The studies by Stough, Steele, and others received little notice, but the publications of the original discoverers continued to influence public perceptions and even public policy—Rauscher even testified about her findings before a committee of the U.S. Congress. The media gives tremendous weight and coverage to the first study published on a research question, and essentially ignores all of those that come later. This bias is unsurprising—fame goes to the discoverer, not to the person who got there a few months later, or who just followed up on the original work. But even in science, the judgment of greatness is a retrospective one that only history can render, and journalism is well known to be only the first draft of history. When a new finding is announced, journalists and other observers might be hard-pressed to say, “I won’t report this story until I see it replicated by at least two other laboratories.” And restraint is all the less likely when the impact might be as great as nine IQ points in ten minutes. The first report of a new scientific finding is analogous to the front-page coverage granted to a high-profile criminal indictment; the news that the results didn’t hold up winds up in the back pages (if it is covered at all), next to the story about the suspect’s eventual exoneration.

  As the Mozart effect story evolved, it became even more fantastical. Even though all of the relevant studies had been conducted with college students or adults, the legend spread that Mozart was great for children, babies, and even fetuses. A Chinese newspaper columnist wrote, “According to studies conducted in the West, babies who hear Cosi Fan Tutte or the Mass in C Minor during gestation are likely to come out of the womb smarter than their peers.”18

  Social psychologists Adrian Bangerter and Chip Heath measured the news coverage devoted to the initial Rauscher-Shaw study and found that in 1993, the year of its publication, it received plenty of media attention, but no more than the other widely covered research studies published in Nature around the same time. (These concerned topics like schizophrenia, the orbit of Pluto, skin cancer, and even how many sexual partners men and women claim to have.) In th
e ensuing eight years, though, the Mozart effect paper received more than ten times as much coverage as those studies. The media’s interest in the others diminished sharply after the initial reports, but coverage of the Mozart effect only grew.19

  Chris’s interest in the Mozart effect was piqued in early 1998 when he was writing an article about the concept of intelligence. The enthusiastic public reaction to the Mozart effect stems partly from the way that the concept of intelligence is presented in the media. Intelligence tests are thought by many to be a simplistic, arbitrary, inaccurate, and even racist way of understanding human cognition.20 What better way to debunk IQ tests than to show that just listening to a few minutes of music can dramatically change your score? The reception of the Mozart effect among experts on cognition was different. Chris noticed that the failures to replicate the original Rauscher, Shaw, and Ky finding were piling up, and that almost all of the successful replications came from the original team, not from independent researchers. In science, whenever just one or a few labs can produce an effect, and others cannot (as in the celebrated case of cold fusion), scientists and skeptics begin to doubt the effect itself. Was the Mozart effect real, or just a myth?

  Chris decided to conduct a meta-analysis, a statistical procedure that combines all of the available data from all of the studies on a research question to determine the best answer. The value of meta-analysis can perhaps be best understood by analogy to the classic carnival game of guessing the number of jelly beans in a jar that we discussed in Chapter 3. If you have a large group of people who want to come up with their best collective estimate of an unknown quantity, the way to do it is to have everyone make his or her guess privately, and then average together all the guesses. Each person’s guess is unlikely to be right, but it is equally likely to be too high or too low. As a result, if you average all of the independent guesses, the estimates that are too large will cancel out the ones that are too small, and you will end up with a more accurate estimate of the actual total.21