The Wisdom Audit, part 5 of 9. This series tests famous sayings against published research and ends each test with a verdict.
Most questions in this series end in “it depends.” This one doesn’t. On the question of who pairs with whom, the data has delivered one of the most lopsided verdicts in social science, and it’s not the one romantic comedies bet on.
This post tests three sayings: birds of a feather flock together, opposites attract, and two heads are better than one. Who should you pick, in love and in work?
Birds of a feather vs. Opposites attract
Origins: The birds are ancient. Homer’s Odyssey has a goatherd sneering that “the god is bringing like and like together,” which tells you the idea was already proverbial 2,700 years ago. William Turner’s 1545 pamphlet has birds “of on kynde and color” flocking and flying together, and the modern English wording lands in John Minsheu’s 1599 dictionary. “Opposites attract” is the upstart; its closest thing to a scientific formulation was Robert Winch’s complementary needs theory of mate selection in the 1950s, built on a sample of 25 married couples.
What it actually claims: people pair either by similarity or by complementarity, and this is directly measurable. You take couples, measure traits, and compute correlations. Positive means similar pairs; negative means opposites.
The verdict arrived in 2023, and it was a landslide. Horwitz, Balbona, Paulich, and Keller published an analysis in Nature Human Behaviour that pooled 480 partner correlations from 199 prior studies, then added a fresh analysis of 133 traits in up to 79,074 couples from the UK Biobank health database. Of those 133 traits, 86 percent showed significant positive partner correlations, and just over 2 percent were negative. The average correlation was .19, running as high as .87 for birth year. Out of 133 measured dimensions of human variation, the number where opposites meaningfully attract was three, and the largest was chronotype, meaning whether you are a morning person or a night person: night owls do pair with morning larks slightly more than chance. That is the entire empirical kingdom of “opposites attract.” It consists of bedtimes.
Winch’s theory, meanwhile, never replicated. Reviewers note that measured husband-wife correlations came back positive, not complementary, in the decades of work after his 25 couples. The pattern extends past romance too: McPherson, Smith-Lovin, and Cook’s classic review of homophily, the tendency of similar people to connect, found “similarity breeds connection” across every type of tie they examined, including friendship, work, and marriage, with race, age, religion, and education drawing the strongest lines.
Verdict: Birds of a feather Holds, and Opposites attract Fails, by about as wide a margin as social science produces.
The rewrite: Opposites attract in magnets and movies. People marry their own kind, with a bedtime exception.
Two heads are better than one
Origin: Ecclesiastes 4:9 reads “two are better than one; because they have a good reward for their labour.” The English proverb form is in John Heywood’s 1546 collection.
What it actually claims: groups outperform the same individuals working separately.
For generating ideas, the answer is a measured no. This is one of the best-replicated embarrassments in organizational psychology. Diehl and Stroebe showed in 1987 that brainstorming groups produce fewer and less creative ideas than “nominal groups,” meaning the same number of people working alone with their output pooled. The main culprit is production blocking: only one person can talk at a time, and while you wait your turn, your idea dies. Mullen’s 1991 pooled analysis confirmed it and added a kicker: the bigger the group, the bigger the loss. Every whiteboard session that opens with “there are no bad ideas” is running a protocol the literature falsified before most of the people in the room were born.
For making decisions, the answer is yes, under two strict conditions. Bahrami and colleagues put pairs on a perceptual task in Science in 2010. When two people had roughly equal competence and could freely discuss their confidence, the pair beat its better member. When their competence differed substantially, the pair did worse than the better member alone. Two heads are better than one only when the heads are well-matched and say how sure they are. Otherwise the weaker head is a tax.
If you’re building a team rather than a pair, Woolley’s 2010 collective intelligence work found a group’s performance across tasks was only modestly related to members’ average or maximum IQ. What predicted it was social sensitivity, equality of conversational turn-taking, and the proportion of women in the group. Groups that let everyone talk outperform groups with the smartest person in the room, because the smartest person in the room is usually blocking the aggregation.
Verdict: Conditional. Ecclesiastes gets partial credit; note that the verse promises a good reward for labour, not for brainstorming.
The rewrite: Two matched heads that share their doubts are better than one. One head is better than a meeting.
What this means for picking your people
The two findings point in different directions, and the tension is the useful part.
| Dimension | What the data says to seek | Source of the rule |
|---|---|---|
| Values, goals, rhythms | Seek similarity; it’s what pairs stably, at scale, across 86 percent of traits | Horwitz 2023, McPherson 2001 |
| Competence, in a working pair | Seek parity; mismatched pairs underperform their better member | Bahrami 2010 |
| Information, in a team | Seek diversity of input plus equality of airtime | Woolley 2010 |
| Idea generation | Seek separation; generate alone, pool, then discuss | Diehl & Stroebe 1987 |
Seek similarity in what you want, parity in what you can do, and structure that ensures everyone actually talks. None of the three sayings contains the full rule, but between them, they were circling it for 2,700 years.
Next in the series: Practice Makes Twelve Percent.
Sources: Horwitz, Balbona, Paulich & Keller 2023 (Nature Human Behaviour 7); Winch 1958, Mate-Selection; McPherson, Smith-Lovin & Cook 2001 (Annual Review of Sociology 27); Diehl & Stroebe 1987 (JPSP 53); Mullen, Johnson & Salas 1991 (Basic and Applied Social Psychology 12); Bahrami et al. 2010 (Science 329); Woolley et al. 2010 (Science 330).
Laurie Stach 