A Blog by

When Experts Go Blind

The picture above is an X-ray computed tomography (CT) scan of a human lung. Go ahead and take a few seconds to look at it carefully.

How long did it take you to spot the gorilla?

The image takes a starring role in a fun study in press at Psychological Science. Trafton Drew and colleagues at the Brigham and Women’s Hospital showed that when people focus on searching these images for bright white cancer nodules, they never notice the gorilla. More shocking, radiologists — who are trained to read CT scans — usually miss it, too.

“It’s a vivid example that looking at something and seeing it are different,” says Drew, a postdoctoral fellow in Jeremy Wolf’s lab. “You can put your eyes on something, but if you’re not looking for it, then you’re functionally blind to it.”

On Tuesday, science writer Wray Herbert wrote about the work for the Huffington Post, calling the data “really scary” with “life-threatening implications.” I have to respectfully disagree. On the contrary, I’d argue that it’s because of this hyper-focused, selectively blind attention that expert radiologists are useful.

The study draws inspiration from the famous ‘invisible gorilla’ experiment done in 1999. In that work, researchers asked participants to watch a short video of people passing a basketball and count the number of passes made by those wearing white. About halfway through the video, somebody walks through the scene wearing a ridiculous black gorilla suit. The gorilla even does a little jig. Yet half of the study’s participants never noticed.

“We just wondered, how far could we push this?” Drew says. Would the eyes of trained experts also be vulnerable to this so-called ‘inattentional blindness’?

There were signs they might be. In one study, for example, researchers asked radiologists to review chest X-rays with a pretty big anomaly — a missing collarbone. About 60 percent of the experts didn’t notice it was gone.

In the new study, Drew asked 24 radiologists to do a typical lung cancer screening using the CT scans of five patients. This requires the radiologist to sit at a computer and look for small white blobs on hundreds of X-rays, each showing a slightly different slice of the patient’s lung. “It’s incredible to watch them do this,” Drew says. “They go through these things in under three minutes.”

The researchers didn’t do anything tricky to the images from the first four patients, which included on the order of 1,000 scans. But hidden in the stack of 239 images from the fifth patient, the researchers inserted 5 consecutive scans showing the cartoon gorilla. They were sneaky about it, too. On first appearance, the gorilla was 50 percent transparent. On the second it was 75 percent, and on the third fully visible. Then it faded back out on the last two scans.

Drew et al., Psych. Sci. (in press)

Just 4 of the 24 radiologists reported seeing the gorilla. What’s more, the researchers had used eye-tracking technology to chart exactly where on the scans the participants had been looking. “The majority of them looked directly at the gorilla for extended periods of time. They just don’t see it,” he says.

Drew repeated the experiment with 25 adults who had no medical training. All of them missed the gorilla.

Like the 1999 gorilla study, this one worked because participants were intensely focused on a very difficult task. As you’d expect (and hope!), the radiologists in the study were far better at spotting the cancer nodules than were the non-experts, with success rates of 55 percent and 12 percent, respectively. And that’s why I doubt the findings have dire implications for medical science.

Let’s say more of the experts had noticed the gorilla. That would necessarily mean that they weren’t as focused on the task they were instructed to perform: find the cancer. If it had been a real clinical setting, a broader focus might very well have caused the doctors to miss a few life-threatening nodules.

In any case, the radiologists didn’t seem very concerned about their competence when they found out about the gorilla at the end of the experiment. “They just thought it was funny,” Drew says.

The bigger lesson here — not just for radiologists, but for scientists, journalists, and anybody else — is about testing your biases and assumptions, Drew says. “It’s important to be willing to look for more than one thing, to set yourself up for success.”

18 thoughts on “When Experts Go Blind

  1. Nice article! In college back in the 1970s I learned about this in the context of animal behavior as “search image.” Animals searching for prey tend to stick to one prey item even after it’s become less common. They’ve become extremely efficient at spotting that kind of prey and they do not see other prey. It’s why when I go for a walk with a non biologist, I nearly always see snakes, salamanders, and other stuff that my friends just pass right by.

  2. spotted the gorilla but was looking for it too. am trained to find the unusually hidden or difficult to spot such as reptiles. i HAVE to think this way to preserve sanity and to stay on this side. in my walks, i look for the early/late leaf, bug, bird, animal that is out of the mornal. i watch the signs of the wee creatures for the weather for they are as good as the forecasters.

  3. This is a very sensible take on the study and it’s nice to see someone taking the time to recognize that the sky isn’t always falling. Makes me curious to know whether there would be any relationship between radiologists’ experience and their likelihood of noticing the gorilla, or any correlation between their likelihood of finding cancer and finding the gorilla.

    I think where the study does have potential implications is in terms of how easy doctors, and others, can miss what they’re not looking for. Sure, that’s rarely going to be a gorilla, but still.

    I also wanted to note that I like that you’ve got the picture of your adorable puppy lined right up with where I make my comments — I bet it makes it harder for people to go off on angry rants.

  4. This makes me think of how people can’t find the ketchup or mustard in the fridge even when it’s right in front of them because they are looking behind things. I also don’t believe people didn’t really notice the missing collarbone, I suspect they just discarded that information from their immediate attention in favor of what they were focused on.

  5. Balanced report. Perhaps another point that should tie to other research on attention, is that should we want the radiologists to scan for some other disease or injury, it may mean necessary reframing of their task. Maybe retraining or a different mental set?

  6. Solitary gorillas found in the lung are almost always benign. A few cause minor irritation and are eventually coughed out or resorbed without sequelae.

    Bilateral gorillas, on the other hand, are cause for concern and should be watched carefully.

  7. Virginia, bear in mind that appropriate screening and early detection are the keys to successful treatment. Recent developments in the lab, some as yet unpublished, suggest we may soon have a sensitive, reasonably priced screening assay for protein-bound gorilla juice.

  8. The title made me wonder how a radiologist would spot a gorilla by its CT image, after all, an ape’s lungs look a lot like a humans. This errant thought led me quite far away from actually seeing the gorilla. I spent more than a minute looking for clues in the comparative anatomy – gorilla’s have more muscle mass, ‘leggy’ arms, broader chests… the image looked like a chip or bonobo’s CT but not a gorilla’s. How would non-vet radiologist notice?
    Then I read the article and instantly saw the gorilla.

  9. This illustrates the importance of the referring doctor explaining the reason for the exam (so the radiologist knows what to look for), a practice often neglected, and for which the requesting MD is not held accountable for, in spite of the fact that it is standard accepted procedure.

  10. Although i spotted the gorilla the first time around i agree with the results of the study. Radiologists often are afflicted with a tunnel vision. A case like this is a good reminder that all possibilities should be born in mind and all corners of the film noticed when interpreting films.

  11. I had the same experience in looking at a tiny Triassic fossil reptile from Spain named Cosesaurus. I overlooked the prepubis and quadrant-shaped coracoid because I wasn’t expecting it. Once I nested Cosesaurus close to pterosaurs in a phylogenetic analysis, these two very pterosaurian bones popped right out. Preconceptions do rule. Google: “Cosesaurus” for more on this.


Leave a Reply

Your email address will not be published. Required fields are marked *