The results of an IQ test can depend on the gender of the person who's conducting the test. Likewise, studies of pain medication can be completely thrown off by the gender of the experimenter. This underappreciated problem is one reason that some scientific findings don't stand the test of time.
Colin Chapman found out about this problem the hard way. He had traveled to Sweden on a Fulbright scholarship to launch his career in neuroscience. And he decided to study whether a nasal spray containing a hormone called oxytocin would help control obesity. The hormone influences appetite and impulsive behavior in obese men.
"I was really excited about this project, from what I understood about how the brain works, I thought it was kind of a slam dunk," he says.
Chapman set up the experiment and then left for a few years to attend Harvard Law School. When he returned, the findings were not at all what he expected, "and I was really disappointed because this was my baby, it was my big project going into neuroscience."
But Chapman, who is now a graduate student at the University of Uppsala, says his idea turned out to be right after all.
"There was another research group that around the same time came up with the same idea," he says. "And they ran basically the same project and they got exactly the results I was expecting to get."
That led him to wonder what had gone wrong with his experiment. One possibility was that the hormone he was using, oxytocin, can waft through the air and affect social interactions, especially between heterosexual people of opposite sexes. So he started to worry that hormones from the experimenters could have been messing up his results.
One man and two women had been conducting the actual experiments. But when he went to find out which scientist did which test, "nobody had kept track of that, because that's not something that's commonly kept track of in science, just in general."
Chapman and two colleagues now argue that's a huge mistake. In this review paper in the online journal Science Advances, they dig back through the history of science and find many, many other examples of studies that are influenced by whether the experimenter and subject are the same gender.
"Even something that's supposedly as stable as IQ can be affected by the gender of the experimenter," he says, referring to a study conducted in the 1970s. "If you have a female experimenter with a male student, for instance, you're going to see higher IQ scores."
It's also a big problem in pain research. A heterosexual man participating in a pain experiment will report more pain to a male tester than to a female, Chapman says. He suspects this is partly because the man is — subconsciously or otherwise — trying to impress the woman, and partly because the biochemistry of sexual attraction is at work.
"If you're testing out a new drug for pain, and you're getting these kinds of great results, you might want to look at [the genders of] who's running the experiment and who's participating in the experiment, because that could explain it more than the drug itself," Chapman says.
And the subjects of these experiments don't even have to be humans.
In 2014, researchers discovered that the sex of a laboratory worker could completely screw up the results of pain experiments in rats and mice.
A sweaty man's T-shirt in the room, or a swab from a man's armpit, was enough to skew the results. Even scents from other animals affected the rodent studies.
"If you present the male chemical in front of rats or mice, they are stressed and that stress ends up killing pain," says Dr. Jeffery Mogil, a neuroscience professor at McGill University who led that study.
The study got a fair amount of attention at the time. But since? "Scientists change their practices very, very slowly," Mogil says, "and I think it would be fair to say very little has happened" as a result of those eye-opening observations.
In the meantime, scientists continue to come across these problems. Mogil saw a report at the Society for Neuroscience meeting last November, in which mice responded to an antidepressant only when it was administered by men, not women.
He and Chapman agree that a simple first step would be for scientists to report the gender of the people who run these kinds of experiments. Editors of scientific journals could require that, or funding agencies. But at the moment, they don't.