The Invisible Gorilla: And Other Ways Our Intuitions Deceive Us

by Christopher Chabris

  Autism is a pervasive developmental disorder that currently affects about 1 in 110 children. The diagnosis of autism has become more common over the past decade in the United States.20 The symptoms of autism include delayed or impaired language and social skills. Prior to age two, most children engage in “parallel play”—doing the same things as other children they play with, but not interacting directly with them. And many kids are not very verbal before age two. Autism is most frequently diagnosed during preschool, when typically developing children start playing interactively and their language development accelerates. Many parents of autistic children begin noticing that something isn’t quite right with their kids around age two, and in some relatively rare cases, a child who had been developing normally starts to regress and loses the ability to communicate. These symptoms tend to be most noticeable to parents not long after their children have been vaccinated for measles, mumps, and rubella (MMR). In other words, the most clear-cut symptoms of autism become much more pronounced after childhood vaccinations.

  By now, you should recognize the harbingers of the illusion of cause. Parents and scientists seeking a cause for the increase in autism rates spotted this association and inferred a causal relationship. Parents who saw no symptoms before the vaccinations noticed them afterward, a chronological pattern consistent with a causal narrative. They also noticed that increases in vaccination rates roughly coincided with increases in the diagnosis of autism. All three of the major contributors to the illusion of cause—pattern, correlation, and chronology—converged in this case. Of course, the increase in the frequency of the autism diagnosis also coincided with an increase in piracy off the coast of Somalia, but nobody argues that autism causes piracy (or that pirates cause autism, for that matter). The association has to have a plausible causal link, a connection that makes intuitive sense on its surface. It needs to provide an “Aha!” experience, one that taps our pattern perception mechanisms and triggers the illusion of cause. It needs more than the perception of an intuitive causal link to become a popular movement, though. It needs a credible authority to validate the causal link. In the case of vaccines and autism, it needed Dr. Andrew Wakefield.21

  Andrew Wakefield was a prominent London physician who in 1998 announced the discovery of a link between autism and the MMR vaccine. He and a group of colleagues published an article in the medical journal The Lancet that suggested a link between the MMR vaccine and several cases of autism.22 At a press conference on the day his paper was released, Wakefield explained how he came to this belief: “In 1995, I was approached by parents—articulate, well-educated, and concerned—who told me the stories of their children’s deterioration into autism … Their children had developed normally for the first fifteen to eighteen months of life when they received the MMR vaccination. But after a variable period the children regressed, losing speech, language, social skills, and imaginative play, descending into autism.”23 Wakefield’s announced link between autism and the so-called “triple jab” received extensive popular media attention, which likely led some parents to begin refusing MMR vaccination for their children, in turn contributing to reduced population immunity to measles in Great Britain.

  Wakefield’s report was based on claims by parents of eight of the twelve children in the study that their children developed autism after receiving the MMR vaccine. The article acknowledged that the study had not proven an association between the vaccine and autism. To do that, you would need to conduct a large-scale epidemiological study to examine rates of autism in children who had and who had not received the vaccine. Wakefield’s promotion of an association in his press conferences prompted Paul Offit, a pediatrics professor at the University of Pennsylvania and a noted virologist, to comment sardonically in his book Autism’s False Prophets, “It would have been more accurate if he had said he hadn’t provided any evidence that MMR caused autism and had merely reported the convictions of the parents of eight autistic children.”24 Even if Wakefield had conducted a large-scale epidemiological study showing that vaccinated children had higher rates of autism, he still would not have demonstrated a causal link. Recall that to demonstrate causation, an experimenter must use random assignment to conditions. To make such an inference, Wakefield would have had to run a clinical trial in which some children were randomly assigned to receive a vaccine and others to receive a placebo, and he then would have had to show that the rates of autism differed significantly between these two groups.

  Not only has no such clinical trial ever been conducted—nor could it ethically be conducted—but extensive epidemiological studies with hundreds of thousands of children have shown no association whatsoever. The rates of autism are no higher among children who have been vaccinated than among those who haven’t. The link between vaccines and autism is illusory—there’s actually no association whatsoever, let alone a causal one. People perceive a pattern that fits their beliefs and expectations, and they infer a causal relationship from a sequence of events. Yet the anecdotal evidence provided by a few patients inspired an international fear of a highly effective vaccine.25

  What Mother Teresa, Quentin Tarantino, and Jenny McCarthy All Know

  The extensive epidemiological evidence against a link between vaccines and autism and the lack of any experiments showing such a link establish that any inference of causality is an illusion. Vaccines can’t cause autism if vaccinations aren’t even statistically associated with autism. Given such incontrovertible evidence, rates of vaccination should return to the levels that effectively eliminated measles as a common disease. The vaccine is safe and effective in preventing measles, and it is entirely unrelated to autism. Game over, right?

  Not exactly. As authors Chip and Dan Heath note in their engaging book Made to Stick, personal anecdotes are more memorable and stick in our minds much longer than abstract data.26 They quote Mother Teresa: “If I look at the mass, I will never act. If I look at the one, I will.” Anecdotes are inherently more persuasive than statistics. Precisely because anecdotes capitalize on the power of narrative, they hold considerable sway over all of us. You might know from reading Consumer Reports that Hondas and Toyotas have excellent reliability. Consumers Union, the publisher of Consumer Reports, surveys thousands of car owners and compiles their responses to generate their reliability ratings. But your one friend who complains that his Toyota is perpetually in the shop and insists that he would never buy another one can have more power than the aggregated reports of thousands of strangers. We can relate to the experiences—especially the suffering—of a single car owner. We can’t relate to the statistical facts about thousands. And for a story to be powerful, persuasive, and memorable, we need to be able to empathize. Quentin Tarantino, maker of ultraviolent films, explains the importance of empathy this way: “A beheading in a movie doesn’t make me wince. But when somebody gets a paper cut in a movie, you go, ‘Ooh!’”27

  It can be difficult to overcome a belief that is formed from compelling anecdotes. Recall the experiment in which people remembered pairs of sentences better when they had to infer a cause than when the cause was stated explicitly. Anecdotes work in much the same way—we naturally generalize from one example to the population as a whole, and our memories for such inferences are inherently sticky. Individual examples lodge in our minds, but statistics and averages do not. And it makes sense that anecdotes are compelling to us. Our brains evolved under conditions in which the only evidence available to us was what we experienced ourselves and what we heard from trusted others. Our ancestors lacked access to huge data sets, statistics, and experimental methods. By necessity, we learned from specific examples, not by compiling data from many people across a wide range of situations.

  Prominent neuroscientist V. S. Ramachandran uses the following analogy to explain the power of examples: “Imagine that I cart a pig into your living room and tell you that it can talk. You might say ‘Oh, really? Show me.’ I then wave my wand and the pig starts talking. You might respond, ‘My God! That�
�s amazing!’ You are not likely to say, ‘Ah, but that’s just one pig. Show me a few more and then I might believe you.’”28 If you’re convinced that you’ve seen a talking pig, no amount of scientific evidence that pigs are incapable of talking would convince you. Instead, scientists would need to prove to you that the pig you saw didn’t actually talk—that Ramachandran used smoke and mirrors to create an illusion of a talking pig. And the more people circulate similar anecdotes, all equally fooled into believing the magic is real, the more science will struggle.

  If a friend tells you, “I tried this new diet supplement and I now have more energy and fewer headaches,” you will infer that the diet supplement caused those benefits. And having drawn that inference yourself (or trusting your friend who did), you will remember it better. A parent’s story about how her son deteriorated after receiving the MMR vaccine and her expressed belief that the vaccine caused her son’s autism is compelling, memorable, and hard to dismiss from our thoughts. Even in the face of overwhelming scientific evidence and statistics culled from studies of hundreds of thousands of people, that one personalized case carries undue influence. Parents know what they’ve experienced, but they usually don’t know the science in the same way. Much as we intuitively think we know how a zipper works but never test that intuition, nothing impels us to test our anecdote-driven ideas. Like the illusion of knowledge, the illusion of cause can only be revealed by systematically testing our understanding, exploring the logical bases of our beliefs, and acknowledging that inferences of causality might derive from evidence that cannot really support them. That level of self-examination is one that we seldom reach.

  Enter Jenny McCarthy, former Playboy centerfold, star of a hit MTV series, actress, and mother of a boy diagnosed with autism. With the best of intentions and a desire to help children like her own, she has inadvertently become a spokesmodel for an illusion. When McCarthy’s son, Evan, was diagnosed with autism, she, like many parents, began looking for a cause. And despite overwhelming scientific evidence against a link between vaccinations and autism, she locked onto that false lead as the explanation: “It’s an infection and/or toxins and/or funguses on top of vaccines that push children into this neurological downslide which we call autism.” So convinced was she by her personal experience that she stated baldly in response to a question about whether parents should vaccinate their children, “If I had another child, there’s no way in hell.”29 She made similar claims on the Oprah Winfrey Show, lending support to the unfounded fears of a vast audience of parents worried that vaccines can cause autism. Unfortunately, her advocacy, coupled with frequent media coverage of the illusory link, has been effective. The sad result is lowered population immunity to diseases like measles, which makes possible outbreaks like the one we described at the beginning of this chapter.

  The powerful story of a mother who is convinced she understands the true reason for her son’s illness is far more influential than literally dozens of studies with hundreds of thousands of children showing that her reason is bunk. (It also makes for more engaging television.) Just as Jennifer Thompson’s powerful testimony about being raped led to the conviction of Ronald Cotton, the story of one mother’s experience overwhelms our ability to properly weigh the evidence. It appeals to emotion, to our natural tendency to empathize with a person in pain, and to our tendency to give undue influence to anecdotes. Unfortunately, as we empathize with someone’s experiences, we become less critical of the message those experiences convey. We also remember the message better. That is the basis of many advertising campaigns—if you can make the viewer empathize with the actors in the advertisement, people become less critical of what they have to say. In the case of autism, the consequences have been catastrophic.

  If people want to eschew vaccinating their children, thereby putting them at risk of devastating illnesses, current law essentially gives them that right. However, that choice is not made in a vacuum. By not vaccinating your own children, you put other children at risk of exposure during an outbreak. As virus expert Paul Offit notes, “There are 500,000 people in the United States who can’t be vaccinated. They can’t be vaccinated because they’re on cancer chemotherapy, or they’ve had a bone marrow transplant, or a solid organ transplant, or they’re receiving steroids because they have severe asthma. They depend on those around them being vaccinated.”30 When such children come into contact with measles, they can die.

  Vaccination enables a barricade against the rapid spread of the disease by making it possible to effectively quarantine a small number of people. The more unvaccinated people in a population, the greater the likelihood that an infection in one person will snowball into a broad outbreak. The relatively high levels of vaccination that still prevail in the United States are the reason why the outbreak in Indiana was easily halted. In Britain, where the media gave more coverage to Wakefield’s publicity campaign, widespread outbreaks are increasingly common and measles is again considered endemic. That is what happens when the media gives airtime and weight to anecdotal claims of causality rather than proper epidemiological studies.

  To some extent, we all must rely on secondary sources. We all put our trust in experts and the advice they give. Scientists, too, are affected by anecdotes and empathy. We tend to be more trusting of ideas from people close to us and more dismissive of those we know less well. Yet science has a way of filtering out unfounded conclusions: determine whether the studies behind them can be replicated. Anecdotes don’t cumulate in the way that large scientific studies can. And scientific training does help in determining which sources to trust. McCarthy, for all of her good intentions, has devoted her energy and charm to attracting media coverage to a scientifically debunked explanation for autism, effectively diverting attention and resources from more promising research on the condition.

  McCarthy’s reliance on anecdotes over the scientific method and more rigorous statistical analysis has also fueled her belief in false cures for autism. She is convinced that she cured her son’s autism through “a gluten-free, casein-free diet, vitamin supplementation, detox of metals, and anti-fungals for yeast overgrowth that plagued his intestines.”31 But she’s astonished that the medical and scientific communities haven’t jumped to investigate her son’s miraculous recovery: “What might surprise a lot of you is that we’ve never been contacted by a single member of the CDC, the American Academy of Pediatrics, or any other health authority to evaluate and understand how Evan recovered from autism.”

  Could McCarthy be right that her special diet cured her son? Possibly. Is it likely? Not at all. Her regimen is just the latest in a long list of alleged cures for autism. Given the overwhelming scientific evidence that autism has strong genetic bases and that brain development in people with autism differs markedly from brain development in typical children, it’s more likely that Evan’s improvements resulted from extensive behavioral modification therapy that does help some children with autism. Or perhaps his symptoms just became less pronounced as he matured. It’s even possible that Evan did not have autism in the first place, but instead had another disorder with similar symptoms that could have improved in response to medicines he was given for seizures.32

  The tools of scientific reasoning can resolve questions like whether or not vaccines are linked to autism, but people do not necessarily accept the results of scientific studies, even when the data are overwhelming. An earlier false lead in the search for an autism cure focused on the hormone secretin, which plays a role in the digestive system. Anecdotal evidence from a small number of cases suggested that injection of secretin obtained from pigs led to the elimination of autistic symptoms. Yet more than a dozen small clinical trials showed it to be no more effective than a placebo injection of salt water. And a large-scale clinical trial examining multiple doses of synthetic secretin, sponsored by a drug company seeking FDA approval to market the synthetic hormone as an autism treatment, found no benefit.33 That is science at work: Researchers test the hypothesis that a drug is e
ffective by randomly assigning some people to receive the treatment and others to receive a placebo, and then they measure the outcome. The problem comes when people must reason about the outcome—do they trust the science, or do they trust their often flawed intuitions? Do they believe that they know better?

  Adrian Sandler and his colleagues conducted one of these clinical trials. They randomly assigned 28 children to receive a dose of secretin and another 28 to receive a placebo. Not surprisingly (at least in hindsight), they found no benefit whatsoever from the secretin. The more interesting finding of this study came from interviews conducted afterward with the parents of the children: Even after learning that secretin had no benefit at all, 69 percent of them remained interested in injecting their child with it. In another double-blind study, parents were asked to guess whether their child had received secretin or the placebo. Parents often believe that they can detect effects that are missed by the more objective measures used in studies, and they use that belief to justify their continued faith in the efficacy of the treatment. In this case, though, the parents could not even guess successfully whether or not their child had received secretin—they had no idea whether their child had received the drug precisely because the drug had no detectable effect.

  A central problem in combating medical anecdotes with hard data is that in any clinical trial, some people receiving the treatment will improve and some won’t. Our tendency is to remember the cases where people improved and to assume that the treatment caused the improvement. What we usually fail to do is to compare the rates of improvement with the treatment and without the treatment. If the treatment has a causal effect, then a greater proportion of those who received the treatment should improve than those who didn’t. If the treatment doesn’t have a causal effect, then other, uncontrolled factors probably led some people to improve anyhow.