ROBERT SIEGEL, host:
Now, a riddle: What do airport baggage screeners and radiologists looking for breast cancer have in common? Well, in a roundabout way, a recent study in the Journal of Current Biology considers that very question. In both jobs, the study explains, trained professionals spend huge amounts of time visually searching images for rare threats. And in both jobs, these people come up against inherent limits on their ability to look for things that are hard to find.
NPR's Alix Spiegel explores the implications, both for medicine and for national security.
ALIX SPIEGEL: Julianne Greenberg spends much of her professional life in a dark room slightly larger than a closet. There in the blackness, for up to two hours at a time, she will sit staring earnestly into the room's only light source: three computer screens completely filled with images of women's breasts.
Greenberg is a radiologist who hunts for breast cancer. Hundreds of times a day, she visually searches digitized black and white X-rays filled with these beautiful, spidery images of breast tissue.
Recently, she invited me to join her.
Dr. JULIANNE GREENBERG (Radiologist): We're not seeing anything of concern, so far. And that's the case with 90 out of 100 mammograms that we look at.
SPIEGEL: Actually, the odds of finding a real case of breast cancer are much steeper than 90 out of 100. Greenberg tells me that out of 1,000 patients, only five will have actual breast cancer.
Now, typically, Greenberg looks at four images per patient. So for each 1,000 people, she views at least 4,000 pictures. As she tells me this, I do some quick math in my head. If only five people in 1,000 have cancer, that means that only 20 images in every 4,000 images she views will have pictures of real cancer in them.
Is that right?
Dr. GREENBERG: That's about right, yeah. Although 20 of those images may not show the cancer, the cancer may be initially only visible on one of those four images or two of them.
SPIEGEL: So you really are looking for a needle in a haystack.
Dr. GREENBERG: You are. Yeah, you're looking for a needle in a haystack.
SPIEGEL: Of course, the stakes of this work could not be higher. If Greenberg does not find her needle, a woman might die. This is why every time she goes to hit the move on button on her computer, there's this moment.
Dr. GREENBERG: Before I hit the button that gives the case a negative result and moves on to the next one, before I do that, I sit back for a second and say, is there anything else I need to do here? I ask myself once again.
SPIEGEL: Does your finger like literally kind of pause over the button?
Dr. GREENBERG: Yeah, it truly does. Yeah, it pauses.
SPIEGEL: Jeremy Wolfe is a professor at Harvard who studies people like Greenberg, people whose job is to search for things that are extremely hard to find. Because he does this, a lot of his recent work has actually been funded by the Department of Homeland Security. Because very much like breast cancer screeners, baggage screeners mostly spend their days visually searching for stuff that doesn't turn up.
And in his lab, Wolfe has uncovered an unsettling reality about how looking for rare things affects the people looking for them.
Mr.�JEREMY WOLFE: The one-liner version is that if you don't find it often, you often don't find it.
SPIEGEL: Basically, according to Wolfe's research, when what you are looking for is rare, for some reason, your ability to see it decreases. For example, in one experiment he did, Wolfe took 20 X-rayed images of luggage that had been stuffed with guns and knives, and he inserted those images of bags with guns and knives into stacks of images of X-rayed luggage which did not have guns and knives.
Mr.�WOLFE: If I stick those 20 bags into a stack of 40 bags, so on average there's a gun and knife on 50 percent of the bags, then in the experiment the way we did it, people missed about seven percent of those.
If you took exactly the same 20 bags and stuck them in a stack of 2,000 bags, so now the targets are only showing up two percent of the time, then people all of a sudden were missing about 30 percent of the bags.
SPIEGEL: Thirty percent. Obviously, this is a problem. We do not want either our breast cancer screeners or our baggage screeners to miss their targets 30 percent of the time.
Now, Wolfe doesn't know as much as he would like about why this happens. But one thing he is certain of is that even if you swear to yourself that you're going to be hyper-vigilant and not make any mistakes, it doesn't help. Wolfe points out that over the years, he himself has taken many of his own tests.
Mr.�WOLFE: I know what the probabilities are, and I know that I'm trying to be as good a subject as I can possibly be, and I still fall victim to the same effects because the chunks of my brain that are doing this are, well, the technical term is that they are cognitively impenetrable, you can't talk yourself out of it. It just happens essentially automatically.
SPIEGEL: Now, among people who study visual search, there's apparently fairly widespread acceptance of what's now known as the prevalence effect, this idea that people often miss rare things, at least when tested in a lab. The real question, says Kyle Cave, a cognitive researcher at Amherst University is whether or not that effect is as large in the real world, among full-time, well-trained experts like Julianne Greenberg.
Cave points to a famous research study from the 1970s where people were trained for weeks to look for certain letters in the alphabet among a garble of letters. And he says they seemed to get really good at it.
Mr.�KYLE CAVE (Cognitive Researcher, Amherst University): In fact, when they read the newspaper, I've heard accounts that they had trouble ignoring those letters, that they popped out even when they didn't want them to. It became very automatic.
SPIEGEL: Jeremy Wolfe is now working on another study with Homeland Security that is actually testing the prevalence effect directly in people currently working as baggage screeners. So an answer to this question is on the way.
Alix Spiegel, NPR News, Washington.
NPR transcripts are created on a rush deadline by Verb8tm, Inc., an NPR contractor, and produced using a proprietary transcription process developed with NPR. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.