How Scientists, Too, Can Be Stubborn and Wrong
No one is immune to biases that may cloud judgment.
Posted Sep 15, 2016
All of us would like to believe in the courageous scientist whose insight turns decades of dogma on its head, forging brilliant new truths. But in fact Copernicus and Einstein are rare phenomena in science, and scientists are as prone as anyone to the effects of what psychologists and behavioral economists call “confirmation bias."
Confirmation bias is the tendency to look for information that simply confirms what we already believe is true. As Yogi Berra supposedly put it: “I wouldn’t have seen it if I didn’t believe it.” This can, of course, make new learning and changes in opinion very difficult. It is a natural human phenomenon, and scientists are certainly not immune to it.
I remember well my own battle with confirmation bias. In the late 1970s when I trained in psychiatric research, I belonged to a camp that was enthralled with the relatively recent discovery that medications could treat illnesses like depression, panic disorder, and schizophrenia. Our belief, based on some of the first randomized controlled trials ever done in psychiatry, was that psychiatric illnesses are purely “biological” and inherited and therefore treatable only by chemical means. There was another school of thought, however, that insisted that psychotherapy was equally effective for many disorders. That camp produced increasing amounts of data showing that interventions like cognitive behavioral therapy (CBT) are effective for depression and panic disorder, but those of us who believed firmly in the superiority of medication to therapy always found “fatal flaws” in their experiments.
In an attempt to prove once and for all that we were right about medication's superiority, I contacted one of the most prominent CBT researchers, David Barlow of Boston University. I proposed that we conduct a study together, comparing medication to CBT for the treatment of panic disorder. Neither camp could reject the study’s findings, I suggested, because we would design it together and not start until we were both satisfied that the design offered a true test. It turned out that Dave had already been planning such a study with Scott Woods of Yale University, a distinguished researcher known then primarily for his work in psychopharmacology. We decided to ask Katherine Shear, then of the University of Pittsburgh and best known for psychotherapy research, to join us in what would turn out to be one of the first randomized controlled comparisons of psychotherapy and medication for a behavioral disorder.
After an exhausting several years of grant writing, study design, and implementation, the results were finally in. And I did not like them. I stared at those findings endlessly, trying to twist them to fit my fixed belief that medication works better than psychotherapy. I tried and tried to torture the data until they would confess.
Only after some time did it become quite clear to me that CBT had soundly beaten medication. There was simply no way around the fact that CBT yielded better outcomes for patients than medication in this study. Over the next several years I became a proponent of using CBT as the first line treatment for panic disorder and most other anxiety disorders. Medications, while effective, should be reserved for those who do not respond to CBT in most cases. But it took me a long time to shrug off my confirmation bias and admit that I was wrong.
We are actually surrounded by many cases of confirmation bias and the inability to shift opinions in biological sciences and medical practice. As Gina Kolata explained in a recent New York Times article, surgeons continue to perform spinal fusion operations even though study after study has shown they are no better than nonsurgical alternatives. Doctors continue to prescribe antibiotics for viral illnesses like acute bronchitis, despite evidence that they offer no benefit. Dentists tell their patients to floss, despite an almost complete lack of evidence that flossing offers any meaningful benefit to oral health. For years, scientists looked at data and asserted that dietary fat intake is related to coronary heart disease, when in fact the data show no such relationship. In each of these cases, fixed beliefs were tenaciously held despite either a lack of evidence to support them or direct evidence to contradict them.
The good news is that, ultimately, science seems to correct itself. Many studies have now shown that CBT is at least as effective and certainly safer than medication for anxiety disorders and depression; dietary recommendations are changing from eliminating fat to restricting sugar; and slowly the message about unnecessary antibiotic prescribing and back surgeries seems to be gaining some traction. But any time scientists accuse the general public of holding firmly to anti-science views, we can remind them that they are hardly immune to the very forces, such as confirmation bias, that shape those incorrect notions. The more we can all recognize that and be open about it, the more civil and less contentious the conversation around scientific error and changes of opinion will be in this country.
 Siri-Tarino PW, et al, Am J Clin Nutr 2010;91:535-546