Core topic 06 of 13
Attraction and relationships, according to the studies
Attraction is predicted better by how often you run into someone and how similar they are to you than by anything about chemistry. Whether a relationship lasts is predicted better by how the two people behave during disagreements than by how compatible they looked at the start.
Relationship science has a reputation problem because its most quoted findings come from small studies with big headlines. A 2020 project led by Samantha Joel used machine learning on data from 11,000 couples and found that self-reported relationship quality predicted satisfaction well, while individual partner traits predicted almost nothing. Who you pick matters less than what the two of you do.
What reliably increases attraction
Proximity and repeated exposure Replicated
Festinger's 1950 study of a student housing complex found that friendships tracked floor plans almost exactly, with people on the same landing forming most of the ties. Zajonc's mere exposure work then showed that seeing a face repeatedly increases liking with no interaction at all. Moreland and Beach's 1992 lecture hall study found the effect for women who simply attended more classes without speaking to anyone.
Similarity Replicated
Donn Byrne's attitude similarity work in the 1960s produced a near linear relationship between proportion of shared attitudes and liking. The 2023 analysis of 133 traits across millions of couples by Horwitz and colleagues found positive matching on nearly everything measured, including education, political views and substance use. Opposites attracting is mostly a story about temporary novelty.
Self-disclosure builds closeness Replicated
Arthur Aron's 36 questions procedure escalates the personal weight of what two strangers share over 45 minutes. Compared with small talk controls, it raises closeness ratings substantially, and the design has been repeated across cultures and in classroom settings. Reciprocity is what does the work. One person opening up while the other stays closed reduces liking.
The halo effect on attractiveness Replicated
People rated as physically attractive are also rated as kinder, more competent and more trustworthy on no evidence. Dion, Berscheid and Walster documented this in 1972 and it replicates across hiring decisions, court sentencing and teacher evaluations. It is a judgement error, not a fact about attractive people.
Contested favourites
Misattribution of arousal Contested
Dutton and Aron's 1974 suspension bridge study is the source of the idea that fear can be mistaken for attraction. Men who crossed a high swaying bridge were more likely to call the female researcher afterwards. The sample was small, alternative explanations exist including who chooses to cross a scary bridge, and later attempts produced mixed results. The general principle that arousal transfers between sources has better support than this specific demonstration.
Gottman's four horsemen Contested
Criticism, contempt, defensiveness and stonewalling during conflict do predict poorer relationship outcomes, and contempt is the strongest single predictor in several independent samples. The 90 percent divorce prediction figures came from models fitted to known outcomes, so treat the behaviours as useful and the accuracy claims as marketing.
Attachment styles as fixed types Contested
Adult attachment measures predict relationship behaviour with modest reliability. The popular version, sorting people into four permanent categories that explain everything, goes past the evidence. Attachment scores move over time and vary by relationship, and they work better as continuous dimensions than as labels for a personality quiz.
What actually predicts a relationship lasting
Trait matching plays a smaller role than most people expect, including on the extraversion dimension covered in the introverts and extraverts, where quiet and talkative partners do about as well as matched pairs. What predicts outcomes is behavioural: responsiveness during ordinary conversation, how the couple behaves in the first three minutes of a disagreement, and whether either partner keeps score. Shelly Gable's research found that how you respond to your partner's good news predicts satisfaction better than how you respond to their bad news. Active, engaged interest in a small win beats sympathy during a crisis, which most people find counterintuitive.
Watch this on the channelRelationship science videos covering what the large studies found, including where Gottman's numbers came from.
Open Psychology Cue on YouTubeQuestions people ask about this topic
Do opposites attract?
The evidence points the other way. Similarity in attitudes and values predicts attraction and satisfaction, and a 2023 analysis of 133 traits across millions of couples found positive matching on almost all of them.
Does the 36 questions study really make strangers fall in love?
It makes strangers feel closer through escalating mutual self-disclosure. It was never designed to produce love, and the couple who married after trying it were a journalist's anecdote.
Can you predict divorce with 90 percent accuracy?
No. Those figures came from models fitted to couples whose outcomes were already known. The behaviours identified, especially contempt during conflict, do predict poorer outcomes far less precisely.
Why do people become more attractive the more you see them?
Repeated exposure without anything bad happening increases liking. Zajonc demonstrated it in 1968 and it has been confirmed across more than 200 experiments.