The question of what draws two people together has occupied researchers for decades. The findings are more surprising, and more nuanced, than the popular account - which tends to reduce attraction to physical appearance and leaves out most of the mechanisms that actually operate. Understanding the research means understanding both what pulls people into initial contact and, separately, what sustains desire over time. These two things are not the same, and they are driven by different processes.

Physical attractiveness - what the research actually found

The most systematically studied dimension of initial attraction is physical appearance. Research in this area has produced several robust findings, some of which confirm popular intuitions and some of which complicate them significantly.

Judith Langlois and colleagues conducted a landmark series of studies examining what makes a face attractive. Their key finding was that facial averageness - faces that represent the mathematical average of many individual faces - is consistently rated as more attractive than distinctive or unusual faces. This counterintuitive result has been replicated across cultures. The proposed explanation is evolutionary: average faces signal genetic heterozygosity, which correlates with immune system robustness and developmental stability. An average face is, in this reading, a health signal rather than an aesthetic one.

Symmetry is the related finding. Asymmetries in facial and body structure can result from developmental stress - illness, poor nutrition, genetic mutations - during development. Symmetrical faces and bodies may therefore function as indicators of developmental health. Multiple studies have found preferences for symmetry across a range of features, though the effect size varies and the interpretation remains debated.

What is important to note is what physical attractiveness does not predict well: long-term relationship satisfaction. Initial attractiveness drives first impressions and the likelihood of pursuing contact, but once people enter relationships, its influence on satisfaction diminishes rapidly relative to other factors. Responsiveness, communication, shared values, and emotional availability become far stronger predictors of how the relationship goes.

Proximity and the mere exposure effect

One of the most replicated findings in attraction research is also one of the most counterintuitive to people who believe attraction is primarily about properties of the other person. Proximity - simply being physically near someone repeatedly - is one of the strongest predictors of whether two people will form a relationship.

The mechanism is the mere exposure effect, documented by Robert Zajonc in a 1968 paper that has since become one of the most cited in social psychology. Zajonc demonstrated that repeated exposure to a stimulus - a face, a symbol, a sound - reliably increases liking for it, even in the absence of meaningful interaction and even when the exposure is not consciously noticed. The effect is robust and has been replicated across a wide range of stimuli and populations.

In the context of attraction, this means that people we encounter regularly - classmates, colleagues, neighbours - have a structural advantage that has nothing to do with their qualities as people. Familiarity produces warmth. This is why residential proximity in university dormitories predicts friendship formation, why so many long-term partnerships begin in workplaces, and why the intuition that attraction comes from across a crowded room is, statistically, less accurate than attraction coming from the person sitting next to you every Tuesday.

Similarity and attraction

The "opposites attract" thesis is popular and largely unsupported by research. The similarity-attraction effect - the consistent finding that people are more attracted to those who resemble them in attitudes, values, and interests - is one of the most replicated results in relationship science. Donn Byrne's 1971 research on attitude similarity established the basic framework, and it has held up across decades of subsequent work.

The effect operates across multiple dimensions. Attitudinal similarity - sharing views on politics, ethics, and lifestyle - produces liking. Personality similarity matters, though the effect is somewhat smaller. Even similarity in relatively surface-level preferences - music, food, leisure activities - contributes to attraction, likely because it creates opportunities for shared positive experience and signals a deeper alignment of values.

The proposed mechanism is partly cognitive: people who agree with us validate our view of the world, which is reinforcing. There is also a social comfort component: interactions with similar people are typically lower in friction, easier to navigate, and more rewarding on average. The prediction error of encountering someone very different requires more processing effort and introduces uncertainty about how to behave.

This does not mean that complementarity plays no role - there is some evidence that certain trait pairings (higher dominance with higher agreeableness, for example) may work well together. But the dominant and more consistently supported finding is that similarity, not difference, drives initial attraction.

Scent and genetic compatibility

Among the more unexpected findings in attraction research is the role of scent in partner selection. Claus Wedekind and colleagues published a study in 1995 that has become known as the sweaty t-shirt study. Women were asked to smell t-shirts worn by men for two days and rate them for attractiveness. The key finding was that women consistently preferred the scent of men whose major histocompatibility complex (MHC) genes differed from their own.

MHC genes are a cluster of genes central to immune function. They govern, among other things, how the immune system recognises pathogens. The hypothesis is that attraction to dissimilar MHC is an evolved mechanism for producing offspring with broader and more diverse immune coverage - combining two different immune repertoires is more adaptive than combining two similar ones.

The effect has been replicated and extended. It operates below conscious awareness - people do not know they are processing MHC information when they find a scent appealing. There is also evidence that the effect is disrupted by hormonal contraception: women on the pill, whose hormonal state mimics pregnancy, tend to show the opposite preference, favouring MHC-similar scents. The practical implications of this finding are contested, but the biological mechanism appears real.

Scent operates alongside visual attractiveness, not instead of it. Initial visual contact drives approach; scent operates at closer range. Both contribute to the overall assessment that takes place rapidly and largely without conscious deliberation in the early moments of attraction.

Initial attraction versus sustained desire

The mechanisms that produce initial attraction are substantially different from those that sustain desire in an established relationship. This matters practically, because people sometimes evaluate long-term relationships by the standard of initial-encounter attraction, and find them wanting by a comparison that was never meaningful to begin with.

Initial attraction is driven by novelty, uncertainty, and the signals described above - physical appearance, proximity, scent, similarity. These are all mechanisms suited to the question of whether to pursue contact with a stranger. They are not particularly suited to the question of whether to stay with someone you know well.

Sustained desire in long-term relationships is better predicted by responsiveness - the degree to which a partner understands, validates, and cares for you - by ongoing novelty within the relationship, by sexual communication, and by the quality of the broader relationship. Research on long-term couples has consistently found that the most satisfied pairs are those who continue to engage in new experiences together, maintain genuine curiosity about each other, and communicate openly about what they want.

The implication is that long-term attraction is not a fixed property that two people either have or do not have. It is a quality that is actively built and maintained - or allowed to atrophy. The mechanisms that create it in the long run are different from those that create the first spark, and understanding this difference is practically useful for anyone trying to sustain a relationship over time.