Run Game The game

How attraction works: what's known and what isn't

6 min read

Researchers have studied attraction for decades and found several factors that reliably nudge it. They have also found that the one thing most people want to know, whether two specific people will click, is hard to predict in advance. Both halves are worth knowing.

Almost all of this work is about groups of people, mostly students, mostly in Western countries. It gives averages and tendencies. It doesn't give a rule for any one person.

Proximity: who you're near

People tend to like and befriend those who are physically close. In a 2008 study by Back and colleagues at Leipzig, 54 first-year university students were assigned random seat numbers at a single introductory session. A year later, students who had sat next to each other rated their friendship as more intense, and so did students who had only shared a row. The sample was small and from one university, and the study measured friendship, not romance.

The older classic is the Westgate housing research at MIT by Festinger, Schachter and Back (1950), cited as the first account of the propinquity effect. A common example from it is that residents living near a stairway had more friends on other floors. The sources cited here give no sample details for it, so treat it as background, not a measured effect size.

Evidence strength: consistent in many settings, though much of it is about friendship.

Familiarity: seeing someone again

Repeated exposure tends to raise liking, which psychologists call the mere exposure effect. A 2017 meta-analysis by Montoya and colleagues combined 268 curves from 81 articles. The relationship between exposure and liking looked like an inverted U: liking rises, then can flatten or fall with a lot of exposure. It held for visual stimuli but not auditory ones, and the authors concluded that existing explanations don't fully account for the pattern.

A 2020 paper by Kellen Mrkva and Leaf Van Boven argued that part of the effect comes from salience, meaning how much something stands out compared with other things, rather than raw repetition. So the claim "the more you see someone, the more you like them" is a simplification. Most of these studies used images or objects, not people at a bar.

Evidence strength: solid for the basic effect, contested on why it happens and when it reverses.

Similarity: people like people like them

A 2008 meta-analysis by Montoya, Robert Horton and Jeffrey Kirchner pooled 460 effect sizes from 313 studies. Both actual similarity (r = .47) and perceived similarity (r = .39) were linked with attraction. But actual similarity predicted attraction in studies with no interaction or short interaction, and its effect was not significant in existing relationships. Perceived similarity predicted attraction at all stages.

A speed-dating study that tested both together, with about 12 four-minute dates per person, found that perceived similarity, not actual similarity, predicted romantic liking. Perceived similarity was a much weaker predictor when it was measured on specific traits rather than a general impression.

In plain terms, feeling similar to someone seems to matter more than a checklist of shared traits.

Evidence strength: strong overall, with the "actual versus perceived" question still being argued.

Reciprocity: feeling liked

People tend to like those who seem to like them. Speed-dating work by Paul Eastwick, Eli Finkel and colleagues separated two kinds. Dyadic reciprocity is the interest two people share specifically with each other. Generalized reciprocity is the tendency for people who like everyone to be liked back. In a 2007 study, dyadic reciprocity was positive. If one person liked a particular partner more than they liked others, that partner tended to like them more than they liked others. Generalized reciprocity was negative: people who showed more romantic desire toward everyone they met were desired less in return.

One caution: this is speed dating. A review by Finkel and Eastwick notes that in three studies of strangers in non-romantic settings, liking everyone was returned, so the cost of unselective interest may be specific to romantic contexts. Treat it as one setting's pattern.

Evidence strength: moderate. The selective-versus-unselective split comes from speed-dating data, with few published replications cited here.

Appearance: it matters, and people agree on more than they'd guess

A 2000 meta-analysis by Judith Langlois and colleagues ran 11 meta-analyses. Raters agreed about who is and isn't attractive, both within and across cultures. More attractive children and adults were judged more positively and were treated more positively. The study covered many kinds of judgments, not only dating.

For romantic interest in people met in person, a 2014 meta-analysis summarized by UT Austin (95 studies from 15 countries) found attractiveness predicted positive evaluation for both men and women to a similar degree. Earning prospects had a smaller effect, also similar across sexes. The same summary says sex differences did appear for online dating profiles, where men weighted attractiveness more.

Evidence strength: strong that it matters, but it's one factor among several, and individual taste varies.

What can't be predicted

In 2017, Samantha Joel, Eastwick and Finkel tested whether more than 100 self-reported traits and preferences could predict who would be attracted to whom on four-minute speed dates. The models predicted some of a person's general tendency to like others (4 to 18 percent of that variance) and to be liked (7 to 27 percent). They could not predict attraction between two particular people beyond those general tendencies. The authors say compatibility is hard to forecast before people meet, and that a better method could do better.

Put simply, you can raise the odds with the factors above. You can't read the outcome off a profile.

What to do next

  1. Go to the same places on a schedule. Pick one venue or class and go three times in two weeks. The proximity and familiarity findings suggest repeat exposure can help, though they mostly come from friendship and lab studies, not dating.
  2. Look for one real similarity early. In a first conversation, ask until you find one shared thing (a place, a show, a habit) and say so out loud.
  3. Show interest in the person in front of you. Spend the conversation on them, not on how many people you're meeting that night.
  4. Don't treat one flat conversation as a verdict. Chemistry isn't predictable from outside, so keep your sample size up.
  5. Keep a line of notes. After each conversation, write what clicked and what didn't, so you're learning from your own data.

Run Game lets you add a note to each logged approach, which gives you that record. For a related summary of preferences, see what people look for in a partner, the profile of Helen Fisher, the evolutionary view in evolution and attraction, and first impression biases. For what you can work on, see building chemistry honestly and how to be more attractive.

Sources