The gap between men and women is smaller than the gap between two men

Two books did most of the work. John Gray's 1992 Men Are From Mars, Women Are From Venus argued for enormous psychological differences between the sexes and has sold over 30 million copies in 40 languages. Deborah Tannen's You Just Don't Understand advanced what is called the different cultures hypothesis: that men's and women's ways of speaking are so fundamentally unalike that they amount to separate linguistic communities. It sat on the New York Times bestseller list for nearly four years[1].

Neither book invented the intuition. They gave it a vocabulary, and the vocabulary spread further than either book did. You can hear it now from people who have read neither.

The framing is appealing for an obvious reason. If the problem is that you and your partner are running incompatible operating systems, then the fights are not about anything either of you did. That is a considerable relief, and it is the main thing the premise sells.

In 2005 the psychologist Janet Shibley Hyde gathered the major meta-analyses that had been run on psychological differences between men and women and did something simple with them: she sorted the results by size[1].

Across 46 meta-analyses she extracted 128 effect sizes, of which 124 could be classified. Thirty percent were close to zero. A further 48 percent were small. That is 78 percent of the accumulated literature landing in the range where the difference is real but slight. Fifteen percent were moderate, six percent large, and two percent very large. She named the result the gender similarities hypothesis: men and women are similar on most, though not all, psychological variables.

This was contested, so a decade later Ethan Zell and colleagues redid it on the much larger body of work that had accumulated since. They synthesised 106 meta-analyses containing 386 meta-analytic effects. The average absolute difference between men and women across domains came out at d = 0.21, and the size held steady across age, across culture, and across generations[2].

Two independent passes, a decade apart, over largely different literatures, arriving at the same answer. As evidence goes in this field, that is about as solid as it gets.

This is the part worth slowing down for, because everything else depends on it.

Effect sizes here are reported as Cohen's d. The d is the distance between two group averages, measured in units of how spread out the groups are internally. That second half is what makes it useful and what the popular framing throws away.

Men are not all alike. Some men talk constantly and some barely speak. Women are not all alike either, in exactly the same way and to roughly the same extent. Each group is spread across a wide range. A d of 0.21 says the two group averages sit about a fifth of that range apart. The two spreads are not sitting side by side. They are stacked almost on top of each other, nudged very slightly out of line.

Hyde puts a number on how much they share, using Cohen's statistic for the non-overlap of two distributions: at d = 0.20, only 15 percent of the two distributions fails to overlap[1]. The other 85 percent is common ground.

Here is another way to picture the same fact. Pick one man and one woman at random and compare them on a trait where d = 0.21. The man will score higher about 56 times in 100. A coin gives you 50. That figure is arithmetic that follows from the effect size rather than a separate research finding, but it is the honest translation of what a small effect means when applied to two actual people.

So the range among men is far wider than the distance between men and women. That is not a technicality. It is the whole thing. Knowing a person's sex moves your estimate of how they communicate by a small amount, and leaves almost all of the variation unexplained.

Bobbi Carothers and Harry Reis tested this from another angle, asking whether psychological gender differences form genuine categories or lie on continuous dimensions. Physical measures like strength behaved categorically, as expected. The psychological measures, including empathy and interpersonal orientation, with few exceptions did not. Their conclusion was that average differences between men and women are not in dispute, but that the dimensional structure makes those differences inappropriate for diagnosing an individual's psychological traits from their sex[3].

The general finding holds when you narrow to communication, which is where the Mars and Venus premise makes its strongest claim.

From the meta-analyses Hyde collected: self-disclosure, women slightly higher at d = 0.18, rising to 0.28 with a friend and falling to 0.07 with a stranger. Interruptions in conversation among adults, men slightly higher at 0.15, rising to 0.33 for the intrusive kind. Smiling among adolescents and adults, women higher at 0.40, which is the largest of these and comfortably in the moderate range[1].

That smiling figure repays a closer look. When people knew they were being observed, the difference was 0.46. When they did not know, it dropped to 0.19. More than half of the gap disappeared once nobody was watching. That is the signature of a social expectation being performed, not a fixed disposition being expressed.

On talkativeness, the stereotype is simply inverted. Campbell Leaper and Melanie Ayres meta-analysed adult language use and found men marginally more talkative, at d = 0.14, with men slightly higher on assertive speech (0.09) and women slightly higher on affiliative speech (0.12). All three are negligible, and the direction of some reversed entirely depending on context[4].

And the specific version of the claim that circulates as a hard number, that women speak roughly three times as many words a day as men, has been measured directly. Matthias Mehl and colleagues fitted 396 people with recorders that sampled ambient sound across several days and extrapolated daily word counts. Women and men both spoke about 16,000 words a day[5]. The variation between individuals was enormous. The variation between the sexes was not there.

None of this means couples do not get stuck in a recognisable pattern. They do, and it has been studied for decades under the name demand-withdraw. One partner presses: raises the issue, criticises, complains, will not let it drop. The other retreats: goes quiet, deflects, waits for it to be over. The more one presses, the more the other retreats, and the more the other retreats, the harder the first presses.

Most people who have seen this pattern have seen it with the woman demanding and the man withdrawing, and the research agrees that this is the more common arrangement. But Andrew Christensen and Christopher Heavey built a study specifically to find out what the pattern was actually tracking. They had 31 couples hold two separate discussions: one about a change the husband wanted from the wife, and one about a change the wife wanted from the husband.

The wife-demand and husband-withdraw configuration was more likely than its mirror image only when the couple were discussing the change the wife wanted. Underneath that, both spouses were more demanding when the topic was their own request and more withdrawing when the topic was their partner's. The study also found a residual sex effect that honesty requires reporting: men were more withdrawn overall, while women were not more demanding overall[6].

A replication with 29 couples three years later sharpened the finding. When the couple discussed the husband's issue, husbands and wives did not differ in demanding or withdrawing at all. When they discussed the wife's issue, the familiar asymmetry reappeared[7].

Both studies are small, both are lab discussions rather than kitchen arguments, and both sampled American couples of that era. Neither could carry a general claim alone. The weight comes from the meta-analysis: Paul Schrodt and colleagues pooled 74 studies covering 14,255 people. Wife-demand with husband-withdraw was associated with poorer outcomes at r = .380. Husband-demand with wife-withdraw was associated with poorer outcomes at r = .392. The two are effectively the same size[8]. Whatever damage the pattern does, it does not much care which spouse is standing in which position.

The pattern also travels. Christensen and colleagues later surveyed 363 people in Brazil, Italy, Taiwan and the United States and found demand-withdraw negatively associated with relationship satisfaction in all four[9].

Put the two findings side by side and the practical difference is stark.

The gender explanation tells you that your partner withdraws because he is a man, which is not a fact you can do anything with. It is a diagnosis with no treatment attached. Worse, it is a licence: if this is simply what men are, then nothing is owed and nothing is expected to change.

The structural explanation tells you that in this conversation, one of you is asking for a change and the other is being asked to make one, and that those two positions produce those two behaviours in almost anyone. The person who wants something different presses, because letting it drop means it stays the same. The person being asked to change wants the conversation to end, because ending it also means it stays the same. Both responses are rational from where each person is sitting. Neither is a personality.

And positions move. You can test this yourself. Raise something you want him to change, and watch what he does. If he presses and you find yourself wanting the conversation to be over, you were never looking at a trait. You were looking at a seat at a table, and you have just swapped seats.

That reframe does not make the conversation easy. It does tell you where to aim: at the request itself, at whether it is being heard, at what the person being asked is afraid the change will cost them. None of those are answerable by consulting anybody's sex.

Hyde made the cost of the alternative explicit at the end of her paper. Resolving conflict in a relationship depends on communication. If couples believe what they have been told, that men and women are so different as to be nearly unable to understand each other, they may simply stop trying to resolve conflict by talking[1]. The belief is not inert. It removes the tool.