What Buss found.

In 1989, David Buss published data from 10,047 respondents across 37 cultures on what men and women prefer in long-term partners. The study has been cited in thousands of subsequent papers and misrepresented in millions of social media posts. The distance between what the paper says and what people claim it says is substantial. This is a piece about that distance.

What the study actually measured

The Buss (1989) methodology is straightforward enough that its limitations are visible on a first reading. Participants rated the importance of eighteen characteristics in a potential mate — financial resources, ambition, good looks, intelligence, dependability, emotional stability, kindness, health, and others — on a scale from 0 (irrelevant) to 3 (indispensable). They also ranked those characteristics relative to each other. The data collection involved collaborators in each of the 37 cultures and was conducted in the late 1980s.

What the study measured: self-reported preferences for characteristics in hypothetical long-term partners. It did not measure actual partner selection. It did not track who subjects subsequently chose as partners. It did not measure attraction in real interactions. It measured what people said they would want, on a standardized questionnaire, about an imagined partner.

This distinction — between stated preferences and revealed preferences — is not a minor methodological note. It is the primary interpretive constraint on everything the study can be used to argue.

The 37-culture design is the paper’s genuine methodological achievement. Cross-cultural replication at that scale is difficult, expensive, and rare. When a pattern replicates across samples drawn from sub-Saharan Africa, East Asia, South America, Western Europe, and the Middle East, it has more claim to universality than a pattern that emerges only from undergraduate psychology subjects at a single American university — the sample that underlies the majority of psychological research from the same era. The Buss design was, by the standards of 1989, genuinely ambitious.

What replicated across cultures

The finding that replicated most consistently: women rated the financial resources and ambition of a potential partner higher than men did, and men rated the physical attractiveness and youth of a potential partner higher than women did. These patterns held across all 37 cultures with enough regularity that Buss described them as universal sex differences in mate preferences.

Kindness and intelligence replicated as highly valued by both sexes across cultures, which receives less attention in popular accounts of the paper than the sex-differentiated findings do.

Trivers’ (1972) parental investment theory provides the evolutionary frame Buss invokes to explain the sex-differentiated preferences. The logic: the sex that invests more in offspring — in humans, women, through gestation, lactation, and the extended dependence of human children — should be more selective in mate choice. Males, with lower obligatory parental investment, should place greater weight on cues of reproductive value (youth, physical indicators of health) while females should place greater weight on resource acquisition capacity (financial resources, ambition).

The framework is coherent as evolutionary logic. It correctly predicts the direction of the sex differences Buss found. Where it becomes problematic is in the inferential leap from “this pattern exists in self-report data” to “this pattern is immutable, genetically determined, and predictive of individual behavior.” None of those claims follows from the Buss design.

What people get wrong

The most common misreading: the replication of a preference rating across cultures is treated as evidence that the preference is the primary driver of partner selection, that it is fixed rather than flexible, and that it explains individual behavior. None of these follow from the data.

The replication of a preference rating means that, on average, when people in 37 different cultures are asked to rate the importance of financial resources in a potential long-term partner, women rate it higher than men do. That is what was found. It does not mean that every woman prioritizes financial resources above other characteristics. It does not mean that preferences don’t vary with context, relationship type, or life stage. It does not mean that the preference patterns identified in 1989 will be identical in surveys conducted in 2026. And it does not mean that what people say they want in a survey predicts what they choose when they are actually selecting a partner.

Eastwick and Finkel (2008) tested this last point directly. In a speed-dating study, they found that participants’ self-reported mate preferences — including the sex-differentiated preferences Buss identified — predicted their actual attraction responses poorly. The correlation between stated importance of physical attractiveness and attraction to a specific person during the speed-date was near zero for both men and women. Stated preferences, in other words, do not reliably predict who we find attractive in real encounters with real people.

This is not a finding that demolishes Buss’s work. It is a finding that specifies its interpretive limits more precisely. The preferences Buss documented are real patterns in self-report data with meaningful cross-cultural replication. They are not reliable predictors of individual behavior in actual social situations.

Feingold’s meta-analysis

Feingold’s 1992 meta-analysis examined the evidence on physical attractiveness across a large body of studies. Several findings are relevant to how Buss’s paper is used.

First, the gender difference in stated preferences for physical attractiveness — men rating it higher than women — is substantially larger in hypothetical preference measures than in behavioral measures. When you ask people whether looks matter, men say yes more emphatically than women do. When you look at actual partner selection behavior, the difference shrinks considerably.

Second, physically attractive people are more likely to date physically attractive partners — the matching hypothesis, originally articulated by Walster et al. (1966) and replicated extensively. Both men and women tend to select partners who match their own level of physical attractiveness, a pattern that emerges from revealed behavior rather than from stated preference.

Third, the halo effect — the cognitive tendency to attribute positive characteristics to physically attractive people — operates similarly across genders in behavioral measures, despite the asymmetry in stated preferences. Both men and women overestimate the desirable characteristics of physically attractive individuals. Both are affected similarly by the correlation between physical appearance and attributed traits like intelligence, warmth, and competence.

Feingold’s meta-analysis does not contradict Buss. It contextualizes the Buss findings within a broader literature on physical attractiveness, and it documents the consistent pattern across that literature: stated preferences and behavioral patterns diverge in predictable ways.

The evolutionary frame and its limits

Trivers’ parental investment theory is the most influential framework in evolutionary psychology of mating. It is a genuinely powerful explanatory structure — it correctly predicted the sex differences Buss found, and it makes specific predictions that can be tested across species and across human populations.

Where it over-predicts: the theory is cleaner than the data. The assumption that males have universally lower obligatory parental investment than females is not uniformly true even within mammals, and the human case is complicated by the extended period of offspring dependency that makes paternal investment advantageous for offspring survival. Trivers himself acknowledged this. The theory generates the directional prediction — female choosiness, male competition — without generating precise effect sizes, and the effect sizes matter.

Where cultural and individual variation matters: the cross-cultural replication Buss found does not imply uniformity across cultures. The magnitude of sex differences in mate preferences varies substantially across societies, in patterns that correlate with the degree of economic and political gender equality. Gangestad et al. (2006) and subsequent researchers have documented this variation. The universal pattern in the Buss data is in the direction of the difference, not in its magnitude. The magnitude is shaped by cultural conditions.

The evolutionary frame is not wrong. It provides the correct directional predictions for the patterns in the data. What it cannot do is serve as the complete explanation for individual behavior, ignore cultural modulation, or be used to argue that the patterns it predicts are immutable or optimal. The distance between “this pattern exists” and “this pattern should determine how you think about attraction and partner selection” is not a distance the evolutionary literature closes.

That distance is where most of the misrepresentation lives.