In 1990 The New York Times published a front-page article by Lawrence Altman, a reporter with a medical degree, announcing that scientists had discovered "a link between alcoholism and a specific gene."

The evidence for the "feel-good gene," which supposedly reduces anxiety, is flimsy, just like the evidence linking specific genes to high intelligence, violent aggression, homosexuality, bipolar disorder and countless other complex human traits and ailments. Image: Wikimedia Commons.

That was merely one in a string of reports in which the Times and other major media hyped what turned out to be erroneous claims linking complex traits and disorders—from homosexuality and high intelligence to schizophrenia and bipolar disorder—to specific genes.

I thought those days were over, and that scientists and the media have learned to doubt extremely reductionist genetic accounts of complex traits and behaviors. I was wrong. Last Sunday, the "Opinion" section of the Times published an essay, "The Feel-Good Gene," which states:

"For the first time, scientists have demonstrated that a genetic variation in the brain makes some people inherently less anxious, and more able to forget fearful and unpleasant experiences. This lucky genetic mutation produces higher levels of anandamide--the so-called bliss molecule and our natural marijuana--in our brains. In short, some people are prone to be less anxious simply because they won the genetic sweepstakes and randomly got a genetic mutation that has nothing at all to do with strength of character."

This article, like the one touting the alcoholism gene 25 years ago, was written by a physician, Richard Friedman, professor of psychiatry at Weill Cornell Medical College. I emphasize this fact because scientific hype is often blamed on supposedly ignorant journalists like me rather than on physicians and other so-called experts.

Here is the scientific backstory. The "feel-good gene" is a variant, or allele, of a gene that produces an enzyme called fatty acid amide hydrolase, FAAH. FAAH is thought to regulate levels of the neurotransmitter anandamide, an endogenous cannabinoid structurally similar to tetrahydrocannabinol, THC, the primary psychoactive ingredient of marijuana. An allele of the FAAH gene, sometimes called 385A, has been associated with lower levels of FAAH, which result in higher levels of anandamide. An estimated 20 percent of Americans carry the 385A gene.

Friedman's article is, in effect, an extremely dumbed down, sensationalized press release for a highly technical article in Nature Communications by 16 researchers, including two of Friedman's colleagues at Weill Cornell Medical College, Iva Dincheva and Francis Lee.

The bulk of the article, "FAAH genetic variation enhances frontal amygdala function in mouse and human," describes elaborate experiments on mice into which the 385A allele has been inserted. The article also describes experiments on humans. I'm focusing on this human research, which is also the primary focus of Friedman's Times article.

Dincheva and Lee's group carried out a two-part experiment on 40 humans, including 18 carriers of the 385A allele. In the first stage of the experiment, called "fear acquisition," subjects were shown various neutral images, such as colored squares. Particular colors, such as yellow, were repeatedly followed by an "aversive" stimulus--such as a loud noise or scary image, like a snarling dog--so that subjects became conditioned to associate the first stimulus with the second.

In the second part of the experiment, called "fear extinction," subjects were shown the colored squares without subsequently being exposed to the "aversive" stimulus. During the experiments, researchers measured subjects' skin conductance by means of electrodes attached to their fingers. High skin conductance indicates more sweating which supposedly indicates more "fear" in response to the aversive stimulus.

So what were the results of Dincheva and Lee et al.? During the "fear extinction" trials, the 385A subjects had a more persistent skin-conductance response to colors that had previously been associated with a noise. The researchers also reported that 385A carriers scored higher on questionnaires that probed their levels of anxiety.

These latter findings are the basis for Friedman's claim that the researchers "demonstrated"—note the absence of qualification in that word—that 385A carriers are "inherently less anxious, and more able to forget fearful and unpleasant experiences."

Friedman does not question the assumptions of the fear acquisition and extinction experiments: first, that someone's response to a noise or image of a dog in a laboratory setting can serve as a proxy for "anxiety" in all its myriad manifestations; second, that skin conductance measures anxiety. (Fun facts: commercial lie detectors, which are notoriously unreliable, measure skin conductance, and so do the "E-meters" with which Church of Scientology "audits" people.)

Friedman only briefly mentions that the researchers found no significant difference between the 385A subjects and other subjects in their initial response to "aversive" stimuli, even though this finding undercuts the claim that 385A confers resistance to anxiety.

Here's another reason to view the "feel-good gene" with skepticism. A 2009 study of 82 subjects reported in Biological Psychiatry found correlations between 385A and neural activity previously associated with anxiety, but the authors add a caveat: "It is important to note that there were no direct associations between FAAH genotype and behavioral phenotypes (i.e., anxiety or impulsivity) in our study."

Friedman's article addresses not only anxiety but also substance abuse. He speculates that people with the feel-good gene may be "less likely to become addicted to marijuana and other drugs" because they have higher levels of the endogenous cannabinoid anandamide.

A 2002 article in Proceedings of the National Academy of Sciences, "Divergent Effects of Genetic Variation in Endocannabinoid Signaling on Human Threat- and Reward-Related Brain Function," reaches precisely the opposite conclusion. The PNAS article reports that the 385A allele "is strongly associated with street drug use and problem drug/alcohol use." The authors even suggest that the 385A allele "may prove to be a useful diagnostic predictor of individuals at risk for drug-related disorders."

Friedman predicts that research on the feel-good gene might leads to better drug treatments of anxiety. I predict that the feel-good gene will suffer the same fate as the "alcoholism gene," which was a dopamine-related gene called DRD2.

In his 1990 Times article on the alcoholism gene, Lawrence Altman noted the researchers' hope that the finding would "open new avenues of research for prevention and treatment of alcoholism." Over the next decade scientists kept announcing that they had found more evidence linking the DRD2 gene to alcoholism, but a 1999 review in Neuropsychopharmacology found "no effect of DRD2 polymorphisms on behavioral phenotypes related to alcohol dependence."

Last fall, I quoted from a 2012 editorial in Behavior Genetics: "The literature on candidate gene associations is full of reports that have not stood up to rigorous replication. This is the case both for straightforward main effects and for candidate gene-by-environment interactions… As a result, the psychiatric and behavior genetics literature has become confusing and it now seems likely that many of the published findings of the last decade are wrong or misleading and have not contributed to real advances in knowledge."

That warning should be appended to all reporting on behavioral genetics. People are desperate for genuine advances in understanding and treating disorders such as substance abuse and pathological anxiety. Prominent scientists such as Richard Friedman and media such as The New York Times do these people a disservice by offering false hope based on flimsy science.

Further Reading (see also comments below from scientists who do work related to that described above):

"Quest for Intelligence Genes Churns Out More Dubious Results."

"My Problem with “Taboo” Behavioral Genetics? The Science Stinks!"

“'Gene-whiz' science strikes again: Researchers discover a liberal gene."

"Have researchers really discovered any genes for behavior?"

"Code rage: The 'warrior gene' makes me mad! (Whether I have it or not)."

Comment from Dale Deutsch, professor of biochemistry and cell biology at Stony Brook University:

I read your blog regarding the "feel good gene." Good for you. I was the guy who identified the enzyme, now called FAAH, in 1993, and we called it anandamide amidase. Ben Cravatt cloned it in 1996 and identified the human gene, and then in collaboration with Dr. Sipe, they identified the P129T mutation in human populations. As you mentioned, there have been some human studies that indicate there is some tendency towards drug abuse, except in the case of marijuana use where the opposite has been reported to occur. The initial paper showing an association of this polymorph with obesity did not seem to pan out in larger studies. I agree that the N.Y. Times article was not balanced. It should have mentioned the studies that show an association with drug abuse. New York Times editors should apply their generally high journalistic standards to their science articles even when they are written by a "CONTRIBUTING OP-ED WRITER." Someone should have at least searched PubMed. Most of the history of FAAH is now on Wikipedia:

Comment from Joseph LeDoux of the Center for Neural Science, New York University, a leading investigator of the biology of fear (and recent speaker at my school):

In my book Anxious, due out on July 14, I am very critical of these kinds of extrapolations from behavioral findings to conscious feelings. The study shows that people with the gene show certain behavioral and brain responses that correlate with the gene variant but do not show that the gene is a cause of conscious feelings. The authors do illustrate a very small reduction in self-reported anxiety in a graph, but do not mention this finding once in the text. Moreover, even if the gene is correlated with less self reported anxiety, less anxiety does not mean what a "feel good" gene implies. This similar to the rampant, and inappropriate, equation of dopamine with pleasure.

Comment from David Goldman, Chief, Lab of Neurogenetics, Acting Clinical Director, National Institute on Alcohol Abuse and Alcoholism:

I respectfully disagree with the overall tone of Horgan's piece. There are many well-validated gene/behavior findings, particularly at the level of neuroscience-based measures. Including the disclaimer that was suggested above would be bizarre. One might as well suggest that all experimental physics papers be accompanied by a disclaimer because of obvious, egregious errors such as mismeasurement of gravity, "discovery" that the speed of light was not as expected, measurement of gravitational waves, cold fusion… the list goes on and on, and some of these mistakes were obvious the instant they were published. The validity of any particular finding stands on its own, which is why I occasionally am allowed to report such findings in journals such as Nature. Behavior is heritable. It would be very strange if no one had found a functional locus altering behavior. However, that's also an alternate universe from the one we live in. Thanks to Seegmiller and others, we have known about the relationship between HPRT and self-mutilation for half a century. Via genetic association, which is what genome wide association ultimately is, many such loci have been found, and verified (which is a standard higher than replication). Examples of common variants securely linked to behavior, and whose mechanisms we understand, include the COMT Val158Met polymorphism, with its roles in executive cognitive function (Egan et al) and anxiety/emotionality (Zubieta et al); FKBP5, whose functional locus leads to impairment in the ultrashort feedback loop damping cellular cortisol response and increasing risk of PTSD (Binder et al); and the serotonin transporter (SLC6A4) polymorphism, which alters expression of this protein and leads to anxiety and stronger effects on brain structure and connectivity (Hariri et al). Mentioning the numerous other validated examples of which geneticists should be aware, and debating the ones for which the evidence is borderline, would turn this retort into a review, and is unnecessary to deflate an overly exaggerated critique of candidate gene studies. It is critical to winnow false positives, but detecting these often requires more insight than identifying the study as a candidate gene study. Also, we can learn equally, or more, from what has worked as by amassing examples of what failed. I feel sorry for some people who have been in the field their whole career and never, like Seegmiller and his student Lesch, discovered a gene, but they should try to maintain their objectivity. I feel less sorry for them when they insist that only one or two frameworks for hypothesis testing--namely genome wide association or meta-analysis of associations to crudely measured behavior, and generation of a p value--is sufficient for validation. They should broaden their minds to other types of science. Even when the meta-analysis is positive, and for example as happened for the serotonin transporter when negative meta-analysis was followed by positive meta-analysis (and which does not invalidate meta-analysis as a scientific activity), some of these naysayers cling to blanket negativism. This is close-minded.