What speed-dating research found about attraction

Speed-dating studies, mainly from Eli Finkel, Paul Eastwick and colleagues, found that what people say they want in a partner told researchers little about whom they actually liked after a few minutes of talking. In several studies, interest in a real person depended on how the conversation went, not on a checklist given beforehand. A 2017 study also found that which two particular people would click was close to unpredictable from the traits they reported in advance.

These findings are real, but narrow. They measure first-impression interest in short, structured dates, mostly among students and volunteers. They don't show who ends up in a good relationship.

Why researchers use speed dating

In a normal dating setting, researchers can't control who meets whom. Speed dating can: dozens of people meet in a set order for a few minutes each, then say who they'd like to see again, all with the same format. That makes it possible to compare what people said they wanted beforehand with what they did when faced with real people.

The main study: 163 students

The study most often cited is Eastwick and Finkel (2008), published in the Journal of Personality and Social Psychology. This summary rests on the authors' manuscript (the 2007 pre-publication version). What they did:

• Who: 163 undergraduates (81 women) at Northwestern University, average age 19.6, recruited through class email lists.
• The event: Each person had between 9 and 13 four-minute dates with everyone of the opposite sex who attended. After each date they filled out a short record. Afterward they said online whether they'd like to meet each person again. That produced 206 matching pairs.
• Follow-up: 92% of participants joined a 30-day follow-up. Every 72 hours they answered questionnaires about each match and anyone else they'd met.

Before the event, participants rated how much they valued things like physical attractiveness and earning prospects in an ideal partner. As in earlier research, stated preferences differed on average by sex: men on average rated attractiveness as more important, and women on average rated earning prospects as more important.

But once people met real partners, the pattern shifted. The authors reported no sex differences in how a partner's attractiveness and earning prospects were related to participants' romantic interest in that partner. And participants' stated preferences, measured before the event, failed to predict what inspired their interest at the event. The authors connect this to a classic idea in psychology that people often can't accurately report what drives their own judgments.

One caution from the authors: on one measure, the choice to exchange contact details, men were more likely than women to say yes to physically attractive partners, while on others, such as feeling chemistry and enjoying the date, the difference reversed. They said exchanging contact details is only one step in how a relationship starts, so they wouldn't lean on that single result.

What other studies found

• Replication (2015). Selterman, Chagnon and Mackinnon ran a direct replication of Eastwick and Finkel in SAGE Open with 307 participants and a 30-day follow-up. Participants were more romantically interested in partners seen as attractive and as good potential earners, and gender didn't change those links. Only the abstract page was read.
• Stated preferences versus choices. Eastwick and Finkel summarize an earlier study of speed-daters by Iyengar and colleagues (2005) in which stated preferences correlated only between .00 and .17 with actual yes or no decisions. That's their description of another team's work, not checked at the source.
• Other speed-dating data. The same paper describes Kurzban and Weeden (2005), data from professional speed-dating events (average age 32) in which income didn't make either sex significantly more desirable, and Fisman and colleagues (2006), a study of graduate students, in which men were more likely than women to say yes to partners they found physically attractive, while men and women didn't differ on ambition or wealthy hometowns. Again, this is how the Eastwick and Finkel paper describes them.
• A smaller 2007 event in Munich. Todd, Penke, Fasolo and Lenton studied 46 adults (aged 26 to 44) at a commercial speed-dating event in Germany. Stated preferences didn't predict who people chose. The sample was very small, many results were marginal or non-significant, and it covered one event and yes/no choices.

Different studies find different gaps between groups, which is one reason to treat any average difference as a tendency in a sample and not a rule about a person.

Can you predict who will click?

Joel, Eastwick and Finkel (2017, Psychological Science) tested whether more than 100 traits and preferences, reported before four-minute speed dates, could predict who'd be attracted to whom. According to the Association for Psychological Science's release, the models could predict some of each person's general tendencies, meaning who tends to like others and who tends to be liked. They could not predict the unique pull between two particular people. The lead author said she expected to predict "at least some portion of the variance" but "didn't expect we would find zero" for that part.

The release adds that the finding applies to the speed-dating context, and that the lead author qualified it with "at least with the tools we currently have available." Eastwick suggested romantic desire may be "more like an earthquake" than a chemical reaction, which is his interpretation.

What it can't tell you

• It's initial attraction. Four-minute dates measure a first impression. The studies didn't follow couples long enough to say whether stated preferences matter for how relationships last.
• The samples are specific. Mostly undergraduates or volunteers who chose to go speed dating. The Eastwick and Finkel authors note that this may not represent everyone and say firm conclusions need replication across diverse samples.
• They're group results. Averages across hundreds of people don't describe you or the person across the table.
• They tested specific traits. Looks and earning prospects, not deal-breakers such as values, location or wanting children, which may work differently.

What it means in practice

This is advice, not a finding. If stated preferences predict little about first-time chemistry, a few habits make sense:

• Meet someone before ruling them out for a trait on a profile.
• Keep your must-haves short and about values and logistics, and treat the rest as flexible.
• Judge the conversation you actually had.

What people look for in a partner covers the stated side, how attraction works covers the factors that do show up in research, and matches but no dates is about moving from an app profile to meeting. For a head-to-head look at apps and in-person meeting, see online dating vs cold approach.

What to do next

1. Write your must-haves in three lines or fewer: values, lifestyle and logistics, such as "wants kids," "lives within an hour," "kind to strangers."
2. List what's flexible, such as height, job, hobbies, and look at it again after your next three dates.
3. Meet before you decide. If a profile or first message is a maybe, suggest a short coffee or walk.
4. After each date, score the conversation, not the checklist: did time pass quickly, did you laugh, did you want to keep talking?
5. Give a good conversation a second meeting even if the person doesn't match your usual type.
6. Notice patterns, such as whether your best dates matched your stated list. That's your own small data set.

Sources

• Eastwick and Finkel, "Sex Differences in Mate Preferences Revisited: Do People Know What They Initially Desire in a Romantic Partner?" (JPSP, 2008), author manuscript: the full manuscript was read. Source for sample, procedure, findings, limits, and the descriptions of Iyengar (2005), Kurzban and Weeden (2005), and Fisman (2006). It is the 2007 pre-publication version.
• Selterman, Chagnon and Mackinnon (2015), SAGE Open, replication of Eastwick and Finkel (2008): only the abstract page was read.
• Todd, Penke, Fasolo and Lenton (2007), PNAS, "Different cognitive processes underlie human mate choices and mate preferences": read on PMC.
• Association for Psychological Science, "Romantic Matches Are Hard to Predict Before People Meet": the press release about Joel, Eastwick and Finkel (2017) was read. It gives no sample sizes. The paper itself wasn't read.