“Likes long walks in the woods on autumn days”

Valentine’s Day is for many just a cruel reminder that they have not yet found the love of their life, their soul mate, their life partner. And let’s face it, finding that special person can be tough in 21st century America. The village matchmakers are long gone, along with the villages themselves, and most of us are spread far and wide, without the traditional networks of family and old friends.

That’s why millions are turning to on-line dating services, which promise to use math and science to find people dates—and often more than dates, life partners. But how reliable are these popular services, and the matchmaking algorithms they use? A new and exhaustive study of these on-line matchmakers—and of romantic prediction in general—raises real doubts about these services’ methods and results. But this critique goes beyond eHarmony and Match.com and Chemistry.com. It questions the entire enterprise of predicting lasting love for any two people who have never met.

Five psychological scientists at five universities spent a year distilling and analyzing more than 400 scientific studies related to dating and romance and marriage, to determine what traits are measurable and valuable in successful matchmaking. The effort was headed up by Eli Finkel of Northwestern University, and the resulting analysis is discouraging for anyone who is gambling on these Internet dating services. But the bottom line of the study, published this week in the journal Psychological Science in the Public Interest, is that no third party—not your sister or best pal or the matchmaker of yore—would do much better in finding you your soul mate.

The scientists identify three broad categories of information that matchmakers might use to match people up for lifelong relationships: Quality of personal interaction, life circumstances, and individual traits and attitudes. All three are important in determining whether a romantic relationship thrives or fails, the scientists say, but in reality much of this vital information is inaccessible or ignored.

Take personal interactions, for instance. This is everything about how two people are with each other—the way they talk, or don’t; how critical or kind they tend to be; how distant or intimate; how good at resolving disputes. Clearly this is important stuff in any relationship—arguably the most important—but as Finkel and colleagues point out, it plays no part at all in on-line matchmaking. Think about it. These matchmaking formulas are designed to predict romantic outcomes for two people who have never met—complete strangers—so how could they possibly factor in such interactive qualities? The short answer is that they don’t, but neither do other, more traditional matchmakers. Your sister may have seen you and a potential partner in action, independently, so she can at least imagine the two of you together and make an educated guess about your dynamic—but it’s just a guess.

Traditional matchmakers also have a slight advantage over computers when it comes to weighing life circumstances. Some of the best predictors of romantic and marital success are things beyond our control—social and economic status, for example. Some of this could in theory be known ahead of time—before two people meet—and factored into a prediction. But the fact is, on-line matchmakers don’t pay much attention to economic and financial issues. Nor do they factor in crucially important life stresses—including unanticipated stress from losing a job, or chronic illness, infertility, a flood or cyclone. These things are unknowable in advance, and even things that are knowable—life alcohol abuse or family pathology—are hidden from on-line matchmakers. Traditional matchmakers have a better chance of knowing some of these circumstances in advance, but even your sister can’t predict a factory closing or the onset of cancer.

So that leaves individual traits, which is really all that these on-line matchmakers have to work with. These traits include not only personality—outgoing, shy, daring, gloomy—but also views and attitudes and values. Do Ron Paul’s politics resonate for you? How about Thai food? Long walks in the woods? On-line services are well equipped to gather a lot of this kind of information and to match up strangers who share such interests and values.

But how important are these things really? Does matching up on tastes and preferences predict long-term satisfaction as a couple? Probably not, the scientists conclude. Most of the on-line matchmaking services match people up based on the assumption that similarity is important to relationship success, but the existing studies of this theory are mixed in their findings and not easily interpretable. For one thing, it’s not at all clear which dimensions of similarity are important. You may both like those long walks in the woods, but have very different tastes in food or politics. What trumps what in the search for compatibility?

Electronic matchmaking’s preoccupation with compatibility may itself be a problem, these scientists conclude. More important than compatibility, they suggest, is something called relationship aptitude. Aptitude is the constellation of traits, preferences and personal history that makes a person more likely to have good relationships in general—not necessarily with a specific other person. One of the most robust findings from relationship science is that the capacity for intimate relationships is a relative stable quality in individuals—regardless of partner—as is the incapacity. That all-important trait may not show up in preferences for Thai food, libertarian politics or autumn strolls. That’s what used to be called good character, which no matchmaking algorithm can possibly capture.

Wray Herbert’s book, On Second Thought, is available in paperback. Excerpts from his two blogs—“We’re Only Human” and “Full Frontal Psychology”—appear regularly in The Huffington Post and in Scientific American Mind.