In 2014, psychologists ran the largest review ever done of mate preferences: 97 studies, tens of thousands of people, asking whether what you say you want in a partner predicts who you actually end up attracted to. The correlation was r = 0.04 to 0.08 (Eastwick et al., 2014).
That’s effectively zero.
Your “type” is mostly fiction. This note is about what actually moves attraction, split by what you need to know depending on who you’re trying to attract.
The biggest empirical finding in 20 years of mating research
Stated ideals barely register against the things that actually predict attraction. Physical attractiveness of the actual person sits at r = 0.43. What you said you wanted hovers near zero.
This isn’t one study. Kurzban & Weeden confirmed the same in commercial speed dating with N = 10,526. Asendorpf et al. (2011) confirmed it across a one-year community sample. Fisman et al. (2006) confirmed it at Columbia. The pattern is one of the most robust findings in the entire field.
What you say you want and who you actually pick are different people.
One trait does almost all the work
A 2024 conjoint experiment had 445 daters make 5,340 swipe decisions with profiles that varied across 6 dimensions (Witmer, Rosenbusch & Meral, 2025):
A one standard-deviation boost in physical attractiveness had roughly 7 times the effect of intelligence on swipe-right odds. Height, occupation, bio quality: all close to zero.
And the twist: men and women had nearly identical priorities. The classic stereotype that “men care about looks, women care about status” mostly disappears in revealed behavior. Both sexes weight attractiveness more than anything else, and roughly equally.
Physical attractiveness isn’t a male bias. It’s the primary filter for both sexes, but only one sex is socially rewarded for admitting it.
Same app, two realities
Before getting into specifics, understand the market both sexes are operating in. Men and women use the same dating app and have completely different experiences.
From Tyson et al. (2016), a crawl of 480,000 profiles:
- Women match on roughly 10% of swipes. Men match on 0.6%. A 17x asymmetry.
- Men like ~45% of profiles they see. Women like ~10%.
This isn’t men being indiscriminate or women being mean. It’s a structural consequence of imbalanced selectivity compounding across millions of users.
Both sexes chase above their station
Bruch & Newman (2018) analyzed messaging patterns in four US cities. They built a recursive desirability score (the dating-app version of Google’s PageRank) and found:
- Both sexes message ~25% above their own desirability rank. Men reach up ~26%, women ~23%.
- Reply rates collapse with desirability gap. When you message someone meaningfully more desirable than you, your reply probability drops below 21%.
- The top of the distribution receives a wildly disproportionate share of all attention.
The aspiration gap is real. Most attempts fail. This is what the market actually looks like under the hood.
What men actually respond to (for women)
Age
Antfolk (2017), a revealed-behavior study with 2,655 Finnish adults: men’s preferred partner age stays centered on 24–25 across their entire adult life. A 50-year-old man’s ideal female age is still ~24.
Women’s preferred age tracks their own: same age or slightly older, across their lifespan.
Messaging behavior is slightly less extreme than stated ideals. 40-year-old men actually message women in their early 30s most. But the direction is identical and one of the few places where survey, behavior, and cross-cultural data all converge.
BMI, but the target varies by culture
The “universal female body ideal” isn’t universal. The International Body Project I (Swami et al. 2010, N = 7,434 across 26 countries) found preferred female BMI ranges from ~19 in wealthy Western cities to ~26 in rural food-insecure societies.
The signal is energy security. Where food is scarce, weight signals health reserves. Where food is abundant, leanness signals the discipline to stay thin in spite of calories.
There is no universal “right” body. There is a range, calibrated to local ecology.
The famous 0.7 waist-to-hip ratio is much weaker than pop-psych claims
Singh’s 1993 claim that men universally prefer a 0.7 waist-to-hip ratio has had 30 years to prove itself. It hasn’t held up.
Bovet’s 2019 systematic review of the WHR literature concluded:
- Preferred WHR varies from ~0.7 in Western samples to 0.8–0.9 in subsistence cultures
- BMI is a stronger predictor than WHR in studies that separate the two
- Singh’s original images confounded WHR with body size
BMI does most of the work. WHR adds a small amount. The “universal 0.7” claim is overclaimed.
Smile
Tracy & Beall (2011), N = 1,041: men rated smiling women as the most attractive expression. Proud and confident-looking women as the least attractive. Effect sizes d ≈ 0.30–0.50.
For women in photos, warmth signals work. For men, they backfire. We’ll come back to that.
What men don’t respond to
This catches people off guard:
- Female income: essentially no effect on male contact rates (Hitsch et al. 2010)
- Female education: no effect on Tinder swipes (Neyt et al. 2019)
- Female job prestige: null
Men barely differentiate on status cues in women. They sort almost entirely on photos.
What women actually respond to (for men)
Height. A lot.
Women filter on height 4 times more often than men do (Yancey & Emerson, 2016): roughly 55% of women versus 14% of men set explicit height minimums.
From Hitsch et al. (2010), Table 5.5: a 5’6” man needs $175,000 more in annual income to match the desirability of a 6’0” man. A 5’2” man needs roughly $277,000 more.
Height is the single biggest non-face, non-body lever for men in revealed behavior.
Income and status signals
Men earning over $250K received roughly 2x the first contacts of men earning under $50K (Hitsch et al. 2010). Female income, as we just saw, is a flat line.
Status cues generalize beyond money. Dunn & Searle (2010): the same man photographed with a Bentley was rated significantly more attractive by women than the same man with a Ford Fiesta. The reverse manipulation (women with high-status cars) had no effect on male ratings.
Women track status indicators on men. Men barely register them on women.
Education
Neyt, Vandenbulcke & Baert (2019), a field experiment with real Tinder users:
- Women strongly preferred men with master’s degrees over men with bachelor’s, significant at p < .01
- Men showed no significant preference for women’s education
There’s no sign men are “intimidated” by educated women. They just don’t sort on it. Women do.
Muscle
This is the finding that most undercovers.
Sell, Lukazsweski & Townsley (2017), published in Proceedings of the Royal Society B:
Upper-body strength alone explains ~70% of the variance in women’s ratings of male body attractiveness.
Add height and leanness and you hit ~80%. Almost nothing else matters for the body itself.
Frederick & Haselton (2007): moderately muscular men reported about 2.5x more lifetime sex partners than average or below-average men. There’s an inverted-U at the bodybuilder extreme (hypermuscularity is rated less attractive than moderately muscular), but for most men, more muscle is more attractive, full stop.
Voice
Women consistently prefer lower-pitched male voices, with correlations of r ≈ 0.20 to 0.40 with attractiveness ratings.
The interesting part: Pisanski et al. (2018) showed that during speed dating, men automatically lowered their voice pitch when talking to women they later chose. It isn’t conscious. The signal modulates on its own.
Uncomfortable truths
Market position is mostly inherited
Height is genetic. Face structure is genetic. Age is time. Born-income is geography and family.
For most men, being in the bottom 50% of desirability isn’t a moral failing. It’s a default condition that requires other levers (fitness, income, status, style, social skills) to escape. Same for women past their early 20s in the age-preference data.
The market is brutal in both directions, and neither side chose it.
Racial response asymmetries are real, large, and peer-reviewed
Lin & Lundquist (2013), N ≈ 7 million dating-site interactions, confirmed patterns first published by OkCupid:
- Black women receive the fewest replies overall, ~25% below non-Black women, even from Black men
- Asian men receive the fewest replies from non-Asian women, with a similar penalty
- White men show no group penalty
- Higher education moderates the penalty for minority men
These effects are larger in revealed behavior than in surveys. People deny racial preferences more than they act on them.
What you can actually do with this
The findings split cleanly into controllables and non-controllables.
Big levers for men:
- Build muscle. It explains 70% of the variance in body attractiveness ratings. Nothing else comes close.
- Photo posture: pride over smile. Easy to change. Measurable effect.
- Long-game: education, income, status. Women weight all three heavily.
- Voice, grooming, height-signaling clothing. Smaller margins, but real.
Big levers for women:
- Photos: smile. You are not rewarded for looking tough or proud. The opposite.
- Body weight calibrated to your local culture. Not “thinner is always better.” Not “heavier is always better.” It is contextual.
Largely non-controllable:
- Height (men), age, face structure, race, born-status
- Most of what the market actually sorts on
Honest framing: most of attraction is inherited. The controllables are real, but they operate within constraints you didn’t choose.
Why this matters
Not to optimize yourself for the swipe market. To recognize when you’re being played.
- Dating apps are engineered to amplify the one trait that already dominates revealed behavior. You’re not shallow for responding. The app funnels you there.
- Advertising runs the same exploits. Cars signal male status; cosmetics signal female fertility cues. All calibrated to this circuitry.
- Manipulation tactics (negging, false confidence, conspicuous wealth) hijack signals that evolved to read sincere ones. The dominance doesn’t need to be real to trigger the response.
The circuitry isn’t disabled by awareness. But awareness reduces the grip.
A 50-year-old man messaging 22-year-olds isn’t having an original thought. He’s running default software. So are you, whenever you feel that pull. The goal isn’t to suppress the response. It’s to recognize when the response is the thing choosing for you.