Studying the genetics of behavior is often like riding a roller coaster. No sooner has one research group tied a gene to a behavior in a certain animal when along comes the next study proving that the link is spurious or even that the gene in question has exactly the opposite effect. Now, a study  in tomorrow's issue of Science shows just how easily such discrepancies can arise.
In part because they were frustrated by unreproducible results, behavioral geneticist John Crabbe of the Veterans Affairs' Portland Alcohol Research Center and Oregon Health Sciences University, together with colleagues from labs in Edmonton, Canada, and Albany, New York, decided to apply the same battery of behavioral tests to the same eight strains of mice in all three labs, to see how well the outcomes matched. All possible conditions, from the light-dark cycle to the brand of mouse feed, were painstakingly equated.
Yet the results were strikingly different. For instance, in a standard test for anxiety that uses a simple maze, all strains in Edmonton showed lower anxiety levels than in the other two labs. In addition, one strain, in which a receptor for the neurotransmitter molecule serotonin had been knocked out, gave different results in all three cities: In Portland, it showed more activity on the maze than in controls with intact serotonin receptors; in Albany, it was less active; and in Edmonton, lacking the receptor didn't seem to make any difference.
Crabbe says the discrepancies are probably the result of very subtle differences in lab conditions that the team couldn't control for, like chemical impurities in the drinking water, or the way the researchers handled the animals. If they want to make more progress, he says, behavioral geneticists should at least standardize their tests and perform them just as attentively as, say, a DNA extraction. But because even that won't eliminate outcome differences, every result should be replicated with a new batch of mice within the same lab, and perhaps even elsewhere, before it's published.
The work is "the kind of study that needs to be done, but nobody wants to be doing," says behavioral geneticist Elizabeth Simpson of the University of British Columbia in Vancouver. "You're looking into something that people would like to believe is not a problem."