Whether a relationship is good or not depends mainly on how you feel within it, not on your partner’s traits.
An analysis of data from 29 laboratories and 43 longitudinal studies on couples shows that the strongest predictor of relationship quality is how each partner personally feels about the relationship — not their objective traits. Factors such as perceived partner commitment, appreciation, sexual satisfaction, and frequency of conflict all play a role. Self-reported feelings are two to four times more predictive than what partners report about each other. However, no questionnaire can reliably predict whether a relationship will improve or deteriorate over time.
There is no monetary cost, nor does it require much time.…
Machine learning applied to data from 29 laboratories and 43 longitudinal studies tracking the same couples ov…
There is no monetary cost, nor does it require much time. The real challenge lies in shifting your perspective: put aside your checklist of ideal partner traits and instead focus on how you interact day to day.
Machine learning applied to data from 29 laboratories and 43 longitudinal studies tracking the same couples over time reveals that the most important predictors of relationship quality are each partner’s subjective experiences: perceived partner commitment, appreciation, sexual satisfaction, and feelings of mutual contentment, alongside conflict levels. Among individual-level factors, life satisfaction, negative emotions, depression, and attachment styles (avoidant and anxious) also contribute. At the start of these studies, self-reported relationship factors explained up to 45% of differences in relationship quality between couples; by the end, this figure dropped to 18%. Crucially, self-reported feelings remain two to four times more predictive than partner-reported ones. Adding individual differences or partner reports to these self-reported factors yields no additional predictive power.
Joel S, Eastwick PW, Allison CJ, et al. (2020). Machine learning uncovers the most robust self-report predictors of relationship quality across 43 longitudinal couples studies. Proceedings of the National Academy of Sciences, 117(32), 19061-19071. https://doi.org/10.1073/pnas.1917036117
Open source linkThis same research also found that no set of self-reported measures can reliably predict whether a relationship will improve or worsen over time. Therefore, the practical takeaway is this: to assess whether a relationship is currently healthy, rely on your own feelings rather than checking off a list of partner traits. Just don’t expect this to forecast its future trajectory. This finding aligns with the earlier point in this section that emotional value cannot be judged solely by presence or absence; here, we identify exactly which factors matter most.