The idea that “opposites attract” may seem to hold true often enough, but it is definitely not the whole story: When it comes to choosing a partner, people actually tend to pair up with those similar to them—in qualities ranging from height and weight to education, income and personality. In a new study published this week in JAMA Psychiatry, a team of researchers in Sweden reported that partners also tend to be more alike in their psychiatric status.

Using national anonymized databases containing medical and social information on all Swedish residents, the researchers identified everyone—more than 700,000 individuals from 1973 to 2009—with an inpatient or outpatient diagnosis that included one of 11 different psychiatric conditions: schizophrenia, bipolar disorder, autism spectrum disorder, anorexia nervosa, substance abuse, attention-deficit hyperactivity disorder, obsessive-compulsive disorder, major depressive disorder, social phobia, agoraphobia, and generalized anxiety disorder. The researchers then searched marriage and reproduction records to identify these individuals’ partners and determine whether they, too, had been diagnosed with any of these disorders. Only heterosexual couples were considered. For baseline comparison, the team also performed these steps for people who had not been diagnosed with any of the 11 conditions.

They found correlations in the psychiatric diagnostic status of partners—where correlations within the same condition were higher than correlations across conditions. For instance, someone diagnosed with schizophrenia is more likely to choose a partner with schizophrenia than one diagnosed with, say, depression. In fact, further analysis showed, on average, the partner of someone diagnosed with a psychiatric condition is two to three times more likely to also have a diagnosis than the partner of an undiagnosed individual. “Taken together, these results suggest that individuals with psychiatric diagnoses are mating—to a degree greater than would be expected by chance—with other diagnosed individuals,” says lead study author Ashley Nordsletten, a psychologist at Karolinska Institute in Stockholm. The finding did not apply or was far less significant for nonpsychiatric conditions such as Crohn’s disease, type 1 and type 2 diabetes, multiple sclerosis and rheumatoid arthritis.

Nordsletten’s results corroborate evidence found in previous studies that were more limited in size and scope. “The unique quality and quantity of data available in the Swedish registers has offered a means of addressing these limitations,” Nordsletten says, adding that the new study has not only greatly expanded the total number of subjects but has for the first time allowed the examination of nonrandom mating patterns across an entire range of conditions.

The underlying reasons for these patterns are complex, and not entirely known. Nordsletten believes that the most likely explanation is assortative mating—the selection of a partner based on a shared trait or behavior. “At a certain level, individuals feeling that their way of looking at the world is shared are drawn together,” adds Victor Reus, a psychiatrist at the University of California, San Francisco, who was not involved with the study. Another possibility, however, is that this correlation is due to marital interaction—the process by which partners become more similar over time. “Some of the association that’s found could also be produced by one spouse’s reaction to their partner’s condition and the stress associated with it,” Reus suggests. “It may be that a partner who has chronic bipolar disorder or schizophrenia, which perhaps wasn’t evident at the onset [of the marriage], could trigger in a spouse a major depression—something that would make it look like an association after the fact but which doesn’t speak to the temporal sequence of when these disorders occurred in individuals.”

“Our findings of resemblance for neurodevelopmental disorders, which are present through the life span, would seem to favor assortment over interaction, at least in these populations,” Nordsletten says. “Ultimately, further studies are really needed to offer insight into the mechanisms at work here.” She also wants to expand this study once more registry data is available to include homosexual partnerships as well as to better analyze populations—particularly individuals with autism spectrum disorder—who by nature of their conditions tend to have lower rates of mating.

Even so, the implications of her team’s findings are wide in scope, especially for psychiatric genetics research. Because a great deal of genetic analyses are predicated on the assumption of random mating, the presence of nonrandom mating patterns affects the use and significance of certain statistical models and tests. Nordsletten refers to twin studies as an example, pointing out that neglecting correlations between spouses could result in underestimating the heritability of diseases. “It’s interesting that here autism spectrum disorder seems to have a high relative risk,” Reus says. “I don’t know that most studies of autism factor in assortative mating in their genetic analyses.” In other words, the widely touted findings that autism is approximately 50 percent hereditary might be underestimating the role of genetics.

Reus also notes that this research may provide a greater understanding from an evolutionary standpoint. “People have struggled to understand why schizophrenia, a condition that is so much associated with early mortality and decreased mating success, hasn’t just died out,” he says. Whereas some scientists theorize that the occurrence of genetic mutations may play a role in keeping the syndrome alive, “in this case mating seems to be another contributing factor.”

Nordsletten hopes to eventually determine whether the correlation found in her study reflects a significant genetic correlation between partners. As to whether that may one day mean being able to identify at-risk individuals or groups—“That is certainly the hope,” she says. “Our results only raise questions to be answered, and answering them will represent the first real steps towards identifying and targeting any risk.”