The Chronicle Review

The Trouble With Intuition

Figure © Daniel Simons

When subjects were shown a video and told to count the number of times a basketball was passed, half of the viewers failed to notice a research assistant in a gorilla suit strolling through the scene.
May 30, 2010

"How do I love thee? Let me count the ways." Those lines by Elizabeth Barrett Browning, written while she was being courted by Robert Browning, and among the most famous in all of poetry, open one of 44 of her love poems that are collectively known as Sonnets From the Portuguese. The Sonnets were first published in book form in 1850, as part of the second edition of her collected poems. At least that's what poetry scholars and bibliophiles thought for several decades until Thomas J. Wise announced that he had discovered a previously unknown earlier printing.

Wise was a celebrated British collector of rare books and manuscripts in the late 19th and early 20th centuries; the catalog of his private library filled 11 volumes. In 1885 an author named W.C. Bennett showed Wise several copies of a 47-page, privately printed pamphlet of the Sonnets dated 1847 and marked "not for publication." Private printings of literature were not unusual in that era. What was unusual was the discovery of a previously unknown collection of such important poetry that predated the first known public printing. Wise immediately recognized the rarity and value of the pamphlets, and bought one for £10 (about $1,200 today). Over the ensuing years, Wise discovered other previously unknown collections of minor works by major authors, including some by Alfred Tennyson, Charles Dickens, and Robert Louis Stevenson. Collectors and libraries snapped up those volumes, and Wise's fame and wealth grew.

At first glance, Wise's items seemed authentic, especially to a buyer considering just one pamphlet at a time. Each one fit nicely with the rest of its author's body of work. For example, the date of Browning's private printing of the Sonnets corresponded with a gap of four years between when the poems were finished and when they were officially published. The pamphlets also appeared to be authentic in format and typography—just how an expert would expect them to look and feel. Although the steady stream of new discoveries by Wise did raise isolated suspicions that something might be amiss, the pamphlets he distributed were broadly respected as genuine for decades.

Some 45 years after Wise found the private edition of the Sonnets, two British book dealers, named John Carter and Graham Pollard, decided to investigate his finds. They re-examined the Browning volume and identified eight reasons why its existence was inconsistent with typical practices of the era. For example, none of the copies had been inscribed by the author, none were trimmed and bound in the customary way, and the Brownings never mentioned the special private printing in any letters, memoirs, or other documents.

The array of circumstantial evidence was impressive, but not conclusive. Carter and Pollard found their smoking gun with scientific analysis. First they documented the fact that all paper used for printing before the 1860s was created from rags, straw, or a strawlike material called esparto. Carter and Pollard then examined one of Wise's Sonnets pamphlets under a microscope and found that the paper was made from chemically treated wood pulp, a technique that wasn't used in Britain until the 1870s. The 1847 edition had to be a fake. The two dealers proved that nearly half of the other Wise pamphlets they examined were also fraudulent, and published their findings in a 412-page book. Wise denied the charges until he died, in 1937, but subsequent investigations confirmed Carter and Pollard's work. Today Wise is celebrated as one of the greatest forgers of all time.

If you have read Malcolm Gladwell's 2005 book, Blink, which is subtitled The Power of Thinking Without Thinking, this tale might seem familiar. Blink begins with a similar story, about an ancient Greek statue known as a kouros, that was offered to the Getty Museum, in Los Angeles. The curators believed the kouros to be genuine, and, relying on scientific tests of its authenticity, they bought it for nearly $10-million. But other art historians, upon first viewing the statue, instantly thought that it was hinky. The former director of the Metropolitan Museum of Art said his first reaction was "fresh"—as in, too fresh-looking to be so old. A Greek archaeologist "saw the statue and immediately felt cold." According to Gladwell, those experts' intuitions proved correct, and the initial scientific tests that authenticated the statue turned out to have been faulty.

Gladwell uses the kouros forgery to launch his case for the surprising power of intuitive snap judgments and instinctive gut feelings, which he calls "rapid cognition." As he puts it, "there can be as much value in the blink of an eye as in months of rational analysis." Gladwell goes on to argue that rapid intuitions often outperform rational analyses, and that excessive thinking can lead us to mistakenly second-guess what we know in our gut to be true. Is that conclusion merited? In the case of Wise's pamphlets, the top experts and collectors of the time trusted in their authenticity, but their rapid judgments were wrong, and only painstaking systematic analysis, which integrated multiple types of information from a variety of sources, uncovered the truth. And even for the kouros, expert intuition was divided: The Getty's curators must have initially thought the statue looked authentic, or they wouldn't have considered buying it in the first place. In fact, some experts still believe the kouros to be authentic, and the Getty today labels it "Greek, about 530 B.C., or modern forgery."

Cases in which forgeries that intuitively appear real but later are discovered through analysis to be frauds are fairly common in the art world. Many of the master forger Han van Meegeren's paintings hung in galleries around the world before scientific analysis showed that they were not authentic Vermeers. Indeed, the skill of the forger is precisely in creating works that appear at first glance, even to experts, to be genuine, and that can be exposed as fakes only through lengthy, expensive study. Like Wise's pamphlets, the infamous "Hitler Diaries" were declared authentic and made public in the 1980s before paper-testing proved that they had been created after the end of World War II.

Gladwell's message in Blink has been interpreted by some readers as a broad license to rely on intuition and dispense with analysis, which can lead to flawed decisions. In his book, Too Big to Fail: The Inside Story of How Wall Street and Washington Fought to Save the Financial System From Crisis—and Themselves (Viking, 2009), the New York Times journalist Andrew Ross Sorkin notes that the former Lehman Brothers president Joseph Gregory was a devotee of Blink who even hired Gladwell to lecture his employees "on trusting their instincts when making difficult decisions." (Gregory was removed from power as his firm circled the bankruptcy drain in 2008.)

Intuition means different things to different people. To some it refers to a sudden flash of insight, or even the spiritual experience of discovering a previously hidden truth. In its more mundane form, intuition refers to a way of knowing and deciding that is distinct from and complements logical analysis. The psychologist Daniel Kahneman nicely contrasts the two: "Intuitive thinking is perception-like, rapid, effortless. ... Deliberate thinking is reasoning-like, critical, and analytic; it is also slow, effortful, controlled, and rule-governed." Intuition can help us make good decisions without expending the time and effort needed to calculate the optimal decision, but shortcuts sometimes lead to dead ends. Kahneman received the Nobel Memorial Prize in Economic Science in 2002 for his work with the late Amos Tversky that showed how people often rely on intuitive heuristics (rules of thumb) rather than rational analysis, and how those mental shortcuts often lead us to make decisions that are systematically biased and suboptimal.

Gerd Gigerenzer, director of the Max Planck Institute for Human Development and author of Gut Feelings: The Intelligence of the Unconscious (Viking, 2007), takes a more benign view of intuition: Intuitive heuristics are often well adapted to the environments in which the human mind evolved, and they yield surprisingly good results even in the modern world. For example, he argues, choosing to invest in companies based on whether you recognize their names can produce reasonably good returns. The same holds for picking which tennis player is likely to win a match. Recognition is a prime example of intuitive, rapid, effortless cognition. Gigerenzer's book jacket describes his research as a "major source for Malcolm Gladwell's Blink," but the popular veneration of intuitive decision-making that sprang from Blink and similar works lacks the nuance of Gigerenzer's claims or those of other experimental psychologists who have studied the strengths and limits of intuition.

The idea that hunches can outperform reason is neither unique nor original to Malcolm Gladwell, of course. Most students and professors have long believed that, when in doubt, test-takers should stick with their first answers and "go with their gut." But data show that test-takers are more than twice as likely to change an incorrect answer to a correct one than vice versa.

Intuition does have its uses, but it should not be exalted above analysis. Intuition can't be beat when we are deciding which ice cream we like more, which songs are catchier, which politician is most charismatic. The essence of those examples is the absence of any objective standard of quality—there's no method of analysis that will decisively determine which supermodel is more attractive or which orchestra audition was superior. The key to successful decision making is knowing when to trust your intuition and when to be wary of it. And that's a message that has been drowned out in the recent celebration of intuition, gut feelings, and rapid cognition.

There is, moreover, one class of intuitions that consistently leads us astray—dangerously astray. These intuitions are stubbornly resistant to analysis, and it is exactly these intuitions that we shouldn't trust. Unfortunately, they are also the intuitions that we find the most compelling: mistaken intuitions about how our own minds work.

We met in the late 1990s at Harvard University, where Dan was a new psychology professor and Chris was a graduate student. As part of an undergraduate laboratory course Dan was teaching, we decided to re-examine some landmark studies the cognitive psychologist Ulric Neisser conducted in the 1970s. In one of those experiments, observers counted the number of times a group of three people wearing white shirts passed a basketball to one another while ignoring three people wearing black shirts who were also passing a ball. In the middle of the video, a woman carrying an open umbrella walked through the scene. Surprisingly, many of the observers didn't notice her. Some psychologists assumed that this failure was a side effect of the unusual video displays Neisser used—the players and the umbrella woman were all partially transparent and looked ghostly, making them somewhat harder to see. As a class project, we decided to test whether people could miss something that was opaque and fully visible.

We filmed the basketball-passing game with a single camera and, like Neisser, we had a female research assistant stroll through the game with an open umbrella. We also made a version in which we replaced the umbrella woman with a woman in a full-body gorilla suit, even having her stop in the middle of the game, turn toward the camera, thump her chest, and exit on the other side of the display nine seconds later. People might miss a woman, we thought, but they would definitely see a gorilla.

We were wrong. Fifty percent of the subjects in our study failed to notice the gorilla! Later research by others, with equipment that tracks subjects' eye movements, showed that people can miss the gorilla even when they look right at it. We were stunned, and so were the subjects themselves. When they viewed the video a second time without counting the passes, they often expressed shock: "I missed that?!" A few even accused us of sneakily replacing the "first tape" with a "second tape" that had a gorilla added in.

The finding that people fail to notice unexpected events when their attention is otherwise engaged is interesting. What is doubly intriguing is the mismatch between what we notice and what we think we will notice. In a separate study, Daniel Levin, of Vanderbilt University, and Bonnie Angelone, of Rowan University, read subjects a brief description of the gorilla experiment and asked them whether they would see the gorilla. Ninety percent said yes. Intuition told those research subjects (and us) that unexpected and distinctive events should draw attention, but our gorilla experiment revealed that intuition to be wrong. There are many cases in which this type of intuition—a strong belief about how our own minds work—can be consistently, persistently, and even dangerously wrong.

The existence of this class of faulty intuitions would just be an academic curiosity if it did not have such significant practical consequences. If you believe you will notice unexpected events regardless of how much of your attention is devoted to other tasks, you won't be vigilant enough for possible risks. Consider talking or texting on a cellphone while driving. Most people who do this believe, or act as though they believe, that as long as they keep their eyes on the road, they will notice anything important that happens, like a car suddenly braking or a child chasing a ball into the street. Cellphones, however, impair our driving not because holding one takes a hand off the wheel, but because holding a conversation with someone we can't see—and often can't even hear well—uses up a considerable amount of our finite capacity for paying attention.

Flawed intuitions about the mind extend to virtually every other domain of cognition. Consider eyewitness memory. In the vast majority of cases in which DNA evidence exonerated a death-row inmate, the original conviction was based largely on the testimony of a confident eyewitness with a vivid memory of the crime. Jurors (and everyone else) tend to intuitively trust that when people are certain, they are likely to be right. Almost all of us have precise memories of how we heard about the attacks of 9/11 or, if we're old enough, the Challenger explosion or President John F. Kennedy's assassination. But you should not be certain that your detailed memories of those events are accurate. Study after study has shown that memories of important events like those are no more accurate than run-of-the-mill memories. They are more vivid, and we are therefore more confident about their accuracy, but that confidence is largely an illusion.

Other intuitions about the mind's workings fail in the same way. For example, it's easy to fall prey to the belief that you understand complex systems better than you really do. This instinct played a role in the financial crisis, especially among investors who bought newfangled mortgage-related bonds whose risks they did not truly appreciate.

The most troublesome aspect of intuition may be the misleading role it plays in how we perceive patterns and identify causal relationships. When two events occur in close temporal proximity, and the first one plausibly could have caused the second one, we tend to infer that this is what must have happened. A tendency to jump to that conclusion is not a bad "default setting" for the human mind, especially in light of the circumstances in which it evolved. In a nonindustrialized society, with no computers, Internet, Google, or even public libraries to access information, the only ways to infer cause and effect were personal experiences and the stories told by others. If a friend ate berries from a particular bush and soon became sick, you might wisely avoid those berries yourself. But your friend's illness might have had nothing to do with the berries.

To determine whether two events are truly associated, we must consider how frequently each one occurs by itself, and how frequently they occur together. With just one or a few anecdotes, that's impossible, so it pays to err on the side of caution when inferring the existence of an association from a small number of examples. Verifying the existence of a genuine association becomes trivial, though, when we can rely on the accumulated experience of hundreds, thousands, or even millions of people. We can decide which car to buy based on the compiled ratings in Consumer Reports rather than on the rantings of a disgruntled owner who happens to be a cousin (or on the manufacturer's slick ad campaign). We can rely on accumulated data, but too often we don't. Why not? Because our intuitions respond to vivid stories, not abstract statistics.

Imagine that your 2-year-old child is diagnosed with an ear infection. Your pediatrician prescribes an antibiotic and, within 48 hours, your child feels better and the infection is gone. Did the antibiotic work? There is no evidence that it did. The infection might have resolved on its own without antibiotics. The first step in demonstrating the efficacy of a drug is to see whether taking it leads to greater improvements than not taking it. To do that, you would first need to show that improvement rates are higher for people who receive the drug than for those who do not. Showing that association is a necessary first step, but it still does not show that the drug caused the improvement. A crucial second step is to randomly assign some patients to receive the antibiotic for their ear infections and others to receive a placebo. Only if the antibiotic group healed faster than the placebo group could you conclude that the antibiotic caused the improvement.

Compared with epidemiological studies and clinical trials, anecdotes—with their lack of control groups—look downright pitiful. Yet we rely on anecdotal causal reasoning all the time, without even realizing the giant leaps of logic we are making. In a recent issue of The New Yorker, John Cassidy writes about U.S. Treasury Secretary Timothy Geithner's efforts to combat the financial crisis. "It is inarguable," writes Cassidy, "that Geithner's stabilization plan has proved more effective than many observers expected, this one included." It's easy for even a highly educated reader to pass over a sentence like that one and miss its unjustified inference about causation. The problem lies with the word "effective." How do we know what effect Geithner's plan had? History gives us a sample size of only one—in essence, a very long anecdote. We know what financial conditions were before the plan and what they are now (in each case, only to the extent that we can measure them reliably—another pitfall in assessing causality), but how do we know that things wouldn't have improved on their own had the plan never been adopted? Perhaps they would have improved even more without Geithner's intervention, or much less. The "data" are consistent with all of those possibilities, but Cassidy and most of his readers are drawn to the most intuitive conclusion: that Geithner's 2009 plan caused the improvements seen in 2010.

We are not naïvely arguing that people should trust only double-blind studies with random assignment when inferring cause. If a man points a loaded gun at us, we don't doubt the outcome of his pulling the trigger, and we won't wait to be shown a peer-reviewed journal article about the appropriate controlled experiment before we start running. There is a plausible and well-established mechanism by which bullets fired from a gun kill people. In simple situations like that, we can safely generalize from a set of principles (Newtonian mechanics) that are well understood and that do involve causal tests that long ago proved the principles correct. In the Geithner example, though, and in many, many other situations, there is no simple analogy with well-understood causal relationships. For complex systems like the global economy, human physiology, or the human mind itself, inferring cause from single examples is not logically justified, because we do not have a complete enough understanding of the internal workings of the system.

Take the case of the perceived link between childhood vaccinations and autism. Nowadays children receive several vaccines before age 2, and autism is often diagnosed in 2- and 3-year-olds. When a child is diagnosed with autism, parents naturally and understandably seek possible causes. Vaccination involves the introduction of unusual foreign substances (dead viruses, attenuated live viruses, and preservative chemicals) into the body, so it's easy to imagine that those things could profoundly affect a child's behavior. But more than a dozen large-scale epidemiological studies, involving hundreds of thousands of subjects, have shown that children who were vaccinated are no more likely to be diagnosed with autism than are children who were not vaccinated. In other words, there is no association between vaccination and autism. And in the absence of an association, there cannot be a causal link.

Many people who believe that vaccination can cause autism are aware of those data. But the intuitive cause-detector in our minds is driven by stories, not statistics, and once a compelling story leads us to ascribe an effect to a cause, we can hold to that belief as stubbornly as when we trust in our ability to talk on a phone while driving—or to spot a person wearing a gorilla suit. In a way, intuition and statistics are like oil and water: They can easily coexist in our minds without ever interacting. That's one reason some in the media continue to treat the vaccine-autism link as a "controversy"—the emotional stories of parents have a constant tug on our beliefs because their effects can't be wiped away by knowing the statistics, no matter how solid they are.

Malcolm Gladwell is regarded as an exceptional science writer in part because of the effective way he uses stories. But it's not just that Gladwell is a better storyteller than his peers. He deploys his stories—anecdotes, really—as part of a compelling rhetorical strategy. Gladwell surrounds his arguments with examples that suggest an association, letting his readers infer the causal relationships he wants to convey. In Blink, he begins his argument with a case in which intuition revealed the kouros fraud, and readers conclude for themselves that putting more trust in their intuition can make them better thinkers. Indeed, experiments have shown that the more mental work readers have to do to infer a cause from a set of facts, the more memorable the causal inference will be. Gladwell, like most good writers, is a master of letting readers "discover" his argument rather than hitting them over the head with it.

There is nothing wrong with using Gladwell's rhetorical technique, as long as the examples are truly illustrative of a valid causal relationship. Charities do this when they highlight the plight of a single individual, with a name and a face, rather than the numerical magnitude of a problem: The stories bring in more money than the statistics. The danger comes from the fact that we promiscuously infer cause from such positive anecdotes in the absence of proper evidence, or even in the face of contradictory evidence. Most people aren't inveterate skeptics vigilantly testing each anecdote to make sure it is representative of an overall pattern.

The actress Jenny McCarthy has used her celebrity to promote proposed cures for autism, such as a special diet she designed for her own autistic son. She often talks about the thousands of parents who have let her know that her regimen helped their children. McCarthy believes, and wants her audience to believe, that those parents have made a valid inference about the effects of the diet. The accumulation of examples in which a possible cause and effect co-occur, no matter how emotionally compelling, provides no evidence of a true association. In McCarthy's case, parents who tried her cure and had no success are unlikely to write to her. Parents who didn't try the cure at all are even less likely to drop her a note reporting that their children got better without trying it—especially if they are among the millions of parents who have never even heard of her proposal.

To know whether intuition should trump analysis, we need more than case studies of initial impressions that were later vindicated. What we need to know is how often experts intuitively identify a forgery despite preliminary scientific analysis suggesting that it was genuine (the kouros case), and how often experts intuitively believe a piece to be genuine only to be proven wrong (the Thomas J. Wise forgeries). Conversely, how often do experts make the mistake of intuiting a forgery when scientific analysis later proves the work to be authentic? Without comparing how frequently intuitions outperform analysis for both genuine and fake items, there is no way to draw general lessons about the power of intuition.

The kouros example is effective because it capitalizes on our tendency to generalize from a single positive association, leading to the conclusion that intuition trumps reason. But in this case, a bit of thought would show that conclusion to be unlikely, even within the confined realm of art fakery. Think about how often experts throughout history have been duped by forgers because intuition told them that they were looking at the real thing. It is ironic that Gladwell (knowingly or not) exploits one of the greatest weaknesses of intuition—our tendency to blithely infer cause from anecdotes—in making his case for intuition's extraordinary power.

Intuition is not always wrong, but neither is it a shortcut around the hard work of logical analysis and rational choice. The trouble with intuition is that while intuitive modes of thought are easier to use than analytical modes, they are poorly adapted to many circumstances and decisions we face in the modern world. If we follow our gut instincts, we will talk on the telephone while we drive, have too much trust in eyewitnesses, and believe we know what causes what—in health care, finance, politics, and every other domain—without even realizing that we haven't considered the right evidence, let alone come to the right conclusions.

Daniel J. Simons is a professor of psychology at the University of Illinois at Urbana-Champaign. Christopher F. Chabris is an assistant professor of psychology at Union College in New York. They are the authors of the new book The Invisible Gorilla, and Other Ways Our Intuitions Deceive Us (Crown Publishers).