# What predicts attraction? Less than anyone selling it admits

_Over a hundred questionnaires could not tell researchers which speed daters would fall for whom. What does hold up, including the parts that make us look bad._ (https://musicrush.ai/blog/what-predicts-attraction, published 2026-11-03)

Everyone has a list. Tall, funny, reads, texts back, owns a plant that is still alive. Then you meet someone who fails four items and you lose a whole weekend thinking about them. The research on this is humbling for anyone who sells certainty about love. That includes dating sites. That includes us.

## What predicts attraction?

The best-supported predictors of attraction are how physically attractive someone seems to you, how similar you believe the two of you are, and how you respond to each other once you actually interact. Traits and preferences measured in advance do a poor job of predicting whether two specific people will click. A person's general desirability can be forecast to a modest degree. The spark between one particular pair, so far, cannot.

A 2014 meta-analysis led by Paul Eastwick found that physical attractiveness predicted romantic evaluations with a correlation of about .40, for men and women alike, while earning prospects came in near .10. On similarity, Matthew Montoya and colleagues pooled 313 studies and found that real similarity mattered in brief encounters but faded after that, and was not a significant predictor in existing relationships. Perceived similarity kept predicting attraction at every stage. Feeling alike seems to travel further than being alike.

## The study every matchmaker should read

In 2017 Samantha Joel, Paul Eastwick and Eli Finkel ran the test the dating industry never volunteered for. In two speed-dating studies, single people filled out more than 100 measures of their traits and preferences, the kind research had linked to choosing a mate. Then each person spent four minutes with each opposite-sex attendee at the event. A flexible statistical method called random forests then tried to forecast the results.

It could predict something about individuals: between 4 and 18 percent of the variation in how much a person tended to desire others, and between 7 and 27 percent of how much a person tended to be desired. But the thing a matchmaker actually promises is the unique pull of this person toward that person, above and beyond those general tendencies. For that, using any combination of the measures taken before the dates, it came up empty.

> A hundred questions could say who tends to be wanted. They could not say who would want whom.

## Your own list is not much better

But surely you know what you like? Eastwick and Finkel tested that in 2008. Speed daters stated their ideals ahead of time, and those ideals failed to predict what actually inspired their desire at the event. Men said they cared more about looks and women about earning prospects, yet those differences did not appear with people they had actually met. The authors' polite conclusion is that we may lack insight into what moves us.

## What happens in the room

The more encouraging finding is that attraction responds to contact. When Harry Reis and colleagues had unacquainted people talk for varying amounts of time, more interaction led to more attraction, helped along by feeling listened to, feeling comfortable, and feeling you know the other person. Those pairs were same-sex strangers, so read it as liking in general rather than romance.

The theme continues later in love. In 2020 Joel and dozens of colleagues pooled 43 longitudinal couples datasets from 29 labs, looking for what predicts relationship quality. The strongest relationship-specific predictors were believing your partner is committed, appreciation, sexual satisfaction, believing your partner is satisfied, and conflict. Once those were counted, individual differences added nothing further. None of that can be read off a bio.

## Where this leaves us

Musicrush reads your listening, places you across five dimensions of taste, and blends that with genre overlap and shared artists into a resonance reading. That is information gathered before you meet someone, which is exactly the category of information these studies should make you wary of. Joel's team tested questionnaires, so their result does not settle whether listening data fares any better. Nobody has shown that it does, and that includes us. Resonance is a hypothesis.

What we can defend is smaller. Rentfrow and Gosling found that music was the most common topic when strangers were asked to get acquainted, and that people could form fairly accurate impressions from taste alone. So shared music is a decent reason to start talking, and talking is where the evidence says attraction gets built. A high reading is an invitation to find out. It is not a forecast.

**Takeaway:** Treat every compatibility claim, ours included, as a reason to meet rather than a verdict. The part of attraction nobody can predict is the part that happens between the two of you, so go and give it the chance.

**Citations:**
- Joel, S., Eastwick, P. W., & Finkel, E. J. (2017). Is romantic desire predictable? Machine learning applied to initial romantic attraction. Psychological Science, 28(10), 1478-1489. https://doi.org/10.1177/0956797617714580
- Eastwick, P. W., Luchies, L. B., Finkel, E. J., & Hunt, L. L. (2014). The predictive validity of ideal partner preferences: A review and meta-analysis. Psychological Bulletin, 140(3), 623-665. https://doi.org/10.1037/a0032432
- Montoya, R. M., Horton, R. S., & Kirchner, J. (2008). Is actual similarity necessary for attraction? A meta-analysis of actual and perceived similarity. Journal of Social and Personal Relationships, 25(6), 889-922. https://doi.org/10.1177/0265407508096700
- Eastwick, P. W., & Finkel, E. J. (2008). Sex differences in mate preferences revisited: Do people know what they initially desire in a romantic partner? Journal of Personality and Social Psychology, 94(2), 245-264. https://doi.org/10.1037/0022-3514.94.2.245
- Reis, H. T., Maniaci, M. R., Caprariello, P. A., Eastwick, P. W., & Finkel, E. J. (2011). Familiarity does indeed promote attraction in live interaction. Journal of Personality and Social Psychology, 101(3), 557-570. https://doi.org/10.1037/a0022885
- Joel, S., Eastwick, P. W., Allison, C. J., Arriaga, X. B., Baker, Z. G., et al. (2020). Machine learning uncovers the most robust self-report predictors of relationship quality across 43 longitudinal couples studies. Proceedings of the National Academy of Sciences, 117(32), 19061-19071. https://doi.org/10.1073/pnas.1917036117
- Rentfrow, P. J., & Gosling, S. D. (2006). Message in a ballad: The role of music preferences in interpersonal perception. Psychological Science, 17(3), 236-242. https://doi.org/10.1111/j.1467-9280.2006.01691.x