What predicts romantic interest and relationship formation?

Summary

Study Purpose and Design

The video reviews a preregistered, longitudinal study of what predicts romantic interest and relationship formation. Earlier research often measured attractiveness or choices in photo-rating and speed-dating settings, but rarely followed participants to see which relationships actually formed.

• The study followed about 210 college students, most of them 18 or 19 years old, for seven months.
• Every three weeks, participants named up to two people they were romantically interested in and rated their interest on a seven-point scale.
• Researchers measured a large set of personal characteristics and perceptions of potential partners, including roughly 164 individual-difference constructs and numerous target-specific measures.
• The study used machine-learning methods to handle the many variables. It was preregistered, meaning its hypotheses and methods were specified in advance.

Main Findings

1. Romantic Interest Tended to Be Strongest at First

• Interest in potential partners generally declined after the initial period, particularly over the first few weeks.
• Interest tended to remain higher for people who continued to be potential partners or became partners than for other acquaintances.
• The study did not show the later rebound in interest sometimes suggested by other models. The presenter notes that the seven-month follow-up may be too short to draw broader conclusions about longer relationships.

2. Perceptions of the Other Person Predicted Interest Better Than Personal Traits

• Individual-difference measures, such as personality and attachment-related traits, accounted for comparatively little variation in romantic interest.
• Ratings of the potential partner and the interaction were more informative about initial and mid-period interest.
• The strongest target-specific predictor was perceived physical attractiveness, with a reported effect size of about 0.57, falling to about 0.42 in another analysis.
• Other predictors, mostly with smaller effects, included:
  • Proximity seeking: The person makes an effort to seek the participant out or talk to them.
  • Perceived interest: The participant believes the other person is interested.
  • Separation distress: The participant misses or wants to be with the person when apart.
  • Mixed signals, which were associated with greater interest in this study.
  • Seeing the person as exciting, as a secure base, or as someone who encourages responsibility.
  • Feeling comfortable with self-disclosure.
  • Perceived relative power: Interest was associated with the partner being seen as less powerful.
  • Perceiving that the person has desirable alternatives.
• Confidence and prior physical or romantic contact had relatively small effects, according to the presenter.

3. Few Personal Traits Were Meaningfully Predictive

• Of the many individual-difference measures, sociosexuality—the willingness to have sex outside a committed relationship—was the strongest reported predictor, but its effect was still small.
• Other statistically significant measures had very small effects. The presenter emphasizes that statistical significance alone does not mean an effect is large or practically important.

4. The Measured Variables Did Not Predict Who Was Together After Seven Months

• When individual differences and target-specific measures were considered together, they explained only about 6% of the variation in relationship status at the end of the study.
• Neither attractiveness nor the other measured variables reliably identified who would still be together.
• The presenter frames this as a reminder that relationship formation is complex and that confident, simple claims about what guarantees a relationship should be approached cautiously.

5. Stated Partner Preferences Did Not Predict Who Participants Chose

• Participants’ stated preferences for traits did not meaningfully predict their romantic interest or eventual partner selection, even when they said a trait was important.
• This contrasts with retrospective studies of existing couples, where people’s stated ideals often correlate with the traits they attribute to their current partners.
• The video raises one possible explanation: people may adjust their preferences or perceptions after entering a relationship. The study cannot establish why retrospective correlations occur, but its prospective design suggests that people may not initially select partners because they possess the traits they say they want.

Descriptive Statistics and Broader Interpretation

• About 67% of participants reported some form of romantic physical contact, while about 38% reported dating during the study. The contact measure included a broad range of behavior, from kissing to sex, so the study could not distinguish among types.
• The presenter reports that most participants had between zero and three such contacts over the seven months, with a mean of about 1.5.
• Serious relationships were relatively uncommon; the presenter reports that about 11% became serious by the end of the study.
• The video’s practical interpretation is that attractiveness appears important for initial interest, but the study does not show that it—or any other measured factor—predicts who remains together.
• The presenter cautions against treating the results as proof that common dating advice is useless. Instead, the findings call for humility: the study may not have captured the factors that determine longer-term outcomes, and its results do not establish causal effects.

Limitations Noted or Implied

• Participants were overwhelmingly very young college students, so the findings may not generalize to older people or other populations.
• Follow-up lasted only seven months.
• The sample and number of relationships were limited, and relatively few participants formed serious relationships.
• The study identifies associations and predictive performance; it does not explain the causes of relationship formation or establish why particular couples stayed together.

Speakers and Sources Featured

• Speaker: Alexander, the video’s narrator and presenter.
• Research source discussed: A preregistered study on predicting romantic interest during early relationship development by Paul Eastwick, Samantha Joel, and coauthors. The narrator particularly identifies Eastwick and Joel as experienced researchers in this area.