The Bodies Themselves

1How To Read A Difference

“Men are more X than women” is not a claim. It is the beginning of one. Until you know how big the difference is, the sentence carries no information — and almost everything written about this subject stops before that point.

1.1 — The hall

Put a hundred men on one side of a hall and a hundred women on the other.

Somebody walks in, picks one person at random from each side, and measures whatever is being argued about — height, memory, patience, grip, anything.

The only question worth asking is how often the man comes out higher. That number, and nothing else, tells you what a difference between men and women is worth.

Fifty in a hundred means no difference at all. Fifty-six means a difference that is real and almost useless for predicting anything about a person you have just met. Seventy-one is what researchers call a large difference. Ninety-two is enormous — visible across a room, obvious to a child.

Researchers do not usually report the number that way. They report something called an effect size, and this chapter exists to let you convert one into the other.

1.2 — The conversion table

Here is the table. It is worth returning to, because every number in the next nine chapters will be read through it.

Look at the row for 0.8 and stay there a moment, because it is the one that does the most damage in public argument.

A “large” difference means that if you pick one person from each group, the woman comes out higher three times in ten. That is not a small residue. In a company of a thousand, it is three hundred people.

So a large sex difference in something does not license a rule about who may do it. Even at the top of this table, using sex to predict an individual means being wrong at a rate no employer, teacher or family would accept in any other context.

1.3 — What the headlines do

Three specific moves account for nearly all bad coverage of this subject, and once you can name them you are largely immune.

Reporting a difference without its size. “Study finds men and women differ in X.” True of almost everything, at almost any sample size, because with enough people even a d of 0.03 becomes detectable. The finding is not that a difference exists. The finding is how big it is, and that number is usually in the paper and not in the article.

Converting an average into a category. “Men are the ones who X.” An average difference becomes a statement about kinds, and the enormous overlap disappears. This is the error in the box above, dressed as a headline.

Reporting the exception as the rule, or the rule as the exception. Both directions happen. A single unreplicated study showing a difference gets a headline; so does a single study showing none. Part One’s rule applies to both: one study is not a fact.

2The Body

This is where the differences are largest, and there is no honest way to soften them. Physical sex differences are among the biggest measured differences between any two human groups — and the one with the greatest consequence of all runs the other way.

2.1 — Size and composition

Start with the easy one. Men are taller, by something in the region of thirteen or fourteen centimetres on average in most populations. In the language of Chapter One, that is an effect size of roughly two — the man is taller about ninety-two times in a hundred pairs.

Underneath height sit differences that matter more for what a body can do.

Men carry substantially more lean muscle mass and less body fat as a share of weight. The muscle difference is not evenly distributed: it is considerably larger in the upper body than the lower. Men also have larger hearts relative to body size, more circulating haemoglobin — meaning more oxygen delivered per litre of blood — and denser, heavier bones.

Women have, on average, greater joint flexibility, better tolerance of some kinds of sustained low-intensity effort, and different injury patterns rather than uniformly more or fewer injuries.

2.2 — Strength

Strength is where the numbers become extreme, and I am going to report them plainly because rounding them off would be dishonest.

Upper-body strength shows the largest gap. Across studies the male advantage in upper-body strength is commonly in the range of forty to sixty per cent; in the lower body it is smaller, often around twenty-five to thirty per cent, partly because everybody’s legs are trained by walking.

2.3 — Sport as an accidental experiment

Competitive sport is the best available natural experiment on physical capacity, because it is the one domain where enormous numbers of people are measured on identical tasks under identical rules with maximum motivation.

The results are consistent. In running, the gap between the best men and the best women is roughly ten to twelve per cent. In swimming it is smaller, around six to ten per cent, because water reduces the advantage of mass. In jumping and throwing it is larger. In events dominated by upper-body power it is larger still.

Those percentages sound modest. At the elite level they are not, because the top of any distribution is crowded.

2.4 — Where the gap narrows, and where it closes

Now the part usually left out, and it is not a consolation prize — it is evidence about mechanism.

The gap is not constant across activities. It shrinks as events get longer and as raw power matters less. In ultra-endurance events the difference is markedly smaller than in the marathon, and in some very long swimming events women have won outright against mixed fields. In events requiring extreme cold tolerance, women’s higher body fat is an advantage rather than a disadvantage.

That variation tells you the gap is not a general fact about capability. It is specific to the demands of a task, and it can reverse when the demands change.

Which sets up the correct way to read this chapter. The male physical advantage is real, very large, and narrow. It concerns power, speed and the ability to apply force. That is not a summary of what a human body is for. It is one axis, and it is the axis on which almost all competitive sport happens to be built, which is why sport makes the difference look like the whole story.

2.5 — The difference that kills

There is one more physical difference, it is the largest in consequence of anything in this chapter, and it runs the other way.

Men die younger. In essentially every country on earth, at every income level, under every political system and every religion, female life expectancy exceeds male. The global gap is in the region of five years, and it has been consistently in one direction for as long as reliable records have existed.

It is not a single cause. Male infants die at higher rates than female infants from the first days of life, before any behavioural explanation can apply. Men are more vulnerable to a range of infections. They die more from most major causes of death, at most ages. And they die vastly more from the causes that involve doing something dangerous — which is behaviour, and which Chapter Six takes up.

This belongs in a chapter about physical differences and it is almost never in one, because it does not fit the shape of the argument either camp is having. Part Fifteen counts what it costs in full.

2.6 — And what it does not establish

Two things need saying before this chapter can be used for anything.

Almost no modern work requires the capacities where the gap is largest. The tasks that separate the sexes most sharply are the ones machines took over first. Upper-body power is not what a surgeon, a lawyer, an engineer, a farmer with a tractor or a soldier with modern equipment is primarily selling. Where the requirement is genuine — some emergency and military roles — the honest response is a job-relevant physical standard applied to every applicant, which is a test of the person rather than of the category, and which some women pass and many men fail.

And nothing in this chapter reaches a single rule in Parts Two and Three. Grip strength does not explain why a widow may not remarry, why a woman must marry inside her caste, why she may not go out after dark, or why a photograph of her at a protest was answered with a rumour about her body. The largest measured difference between men and women is irrelevant to every one of them.

That mismatch is worth holding on to for the rest of the part. Whenever somebody reaches for biology to justify one of these rules, ask which measured difference is being invoked and what it has to do with the rule. The answer is usually that the rule was there first and the biology was recruited afterwards.

3The Mind, Measured

Take every mental ability that has been measured and put each one in the hall from Chapter One. Nearly all of them come out close to a coin toss. The exceptions are real, and one of them runs in a direction almost nobody discusses.

3.1 — The general finding is similarity

In 2005 a researcher reviewed forty-six meta-analyses of psychological sex differences — a review of reviews, covering thousands of studies. She found that around four fifths of the measured differences were small or close to zero.

That is the headline finding of this field and it is almost never reported as one, because a finding of “not much” does not travel.

It should be held firmly, because everything that follows is a list of exceptions, and a list of exceptions read on its own gives a false impression of the whole.

3.2 — The list

Here is what has actually been measured, converted into the hall. The final column is how often a randomly picked man scores higher than a randomly picked woman. Fifty is no difference; below fifty means women score higher.

Read the middle of that table and the picture is unmistakable. Most of the abilities people argue about most furiously sit within a few points of a coin toss.

3.3 — Mathematics

The maths gap deserves its own section because it is the most confidently asserted difference in this field and the mean difference is, to a very good approximation, nothing.

Large national assessment data covering millions of schoolchildren produces an effect size close to zero. Meta-analyses of the research literature agree. In some countries girls score higher. The average boy and the average girl are, on this measure, the same.

3.4 — The gap nobody mentions

Now look at the two rows near the bottom of the table, and notice that they run the other way.

In reading, girls outperform boys. This is not a local finding. In international assessments covering dozens of countries, girls have scored higher in reading in every participating country, in every cycle. There is no country in the dataset where boys read better.

In writing, the gap is larger still, and on some large national datasets it is the biggest sex difference in academic achievement that has been measured — around half an effect size, which puts it in the hall at roughly thirty-six in a hundred.

Put that beside the maths gap of approximately zero and consider how the two are treated. One is a matter of national concern, funded programmes, and decades of public argument. The other is roughly ten times larger, entirely consistent across countries, and most readers of this document are learning about it now.

I want to be careful about what I am claiming. I am not saying the attention to girls in mathematics was wrong; there were real barriers and the gap in participation was never only about scores. I am saying that the pattern of what gets measured, funded and discussed does not track the size of the difference — which is a finding about us rather than about children, and Part One, Chapter Six predicted exactly this.

3.5 — The one large difference

Mental rotation is the exception at the top of the table, and it is worth being precise about what it is.

The task is to look at a three-dimensional shape, then look at another shape, and decide whether the second is the same object rotated or a mirror image. It is a narrow, specific ability, tested in a laboratory, and the male advantage on it is the largest cognitive sex difference reliably found — depending on the version of the test, somewhere between 0.5 and 0.9, which is sixty-four to seventy-four in the hall.

Two honest qualifications sit next to it.

It trains. Performance on mental rotation improves substantially with practice, including with practice on action video games, and the sex gap narrows when both groups are trained. It does not usually vanish, but a difference that moves with a few hours of training is not a fixed property of anything.

And its real-world reach is smaller than it sounds. Mental rotation correlates with some engineering and technical outcomes, modestly. It is not a general spatial faculty, and other spatial abilities in the table show much smaller differences. A single narrow laboratory task is doing a great deal of rhetorical work in public argument that the underlying finding does not support.

4The Tails

Two groups can have exactly the same average and produce very different numbers of people at the extremes. This is the most consequential idea in the part, it is used badly by everybody, and it cuts in a direction almost nobody expects.

4.1 — Same middle, different edges

Everything so far has been about averages. Averages are the wrong place to look for most of the outcomes people actually argue about.

4.2 — The arithmetic

The best way to see this is to do it, so here it is with numbers.

Take two groups with exactly the same average. Give one group a spread seven per cent wider than the other — a difference far too small to notice in any ordinary setting.

Now count how many people from each group appear above various thresholds.

Read that again with the first line of §4.2 in mind. The averages are identical. Nothing in this table comes from one group being better at anything. All of it comes from a seven per cent difference in spread.

And if the spread difference is fifteen per cent rather than seven, the ratio at four spreads above average rises to something in the region of eight to one.

This is the single most important statistical fact in the part, and it explains why arguments about averages so consistently fail to explain what people see. You can have complete equality in the middle and a large imbalance at the edge, produced by nothing except a difference in spread.

4.3 — Does it actually happen?

4.4 — The half that is never mentioned

Here is where this chapter turns, and it turns hard.

Greater variability means more people at both extremes. Not one. Both. The same arithmetic that puts more men at the top of a distribution puts more men at the bottom of it, in exactly the same proportion.

And that is what the data shows. Males are substantially over-represented in intellectual disability, in severe learning difficulties, in diagnoses of attention and developmental disorders, in school exclusion, in failure to complete education, and — as Part Fifteen will count in detail — in almost every measure of catastrophic life outcome.

So anybody invoking male variability to explain why the top of a profession is male has, in the same breath and by the same mathematics, explained why the bottom of society is male. The argument does not have a top-only setting.

4.5 — What it explains, and what it cannot

One practical note before leaving this chapter. Whenever somebody invokes the tails at you, ask which end they are invoking. Almost everybody who reaches for this argument reaches for one end only, and the arithmetic does not have a one-ended setting.

5What People Want To Do

The largest psychological difference between men and women that has ever been measured is not an ability. It is a preference. It is bigger than every gap in Chapter Three combined, it has been measured on half a million people, and it is bad news for both camps.

5.1 — The finding

In 2009 a group of researchers pooled every study they could find on vocational interests — what work people say they would like to do — and analysed them together. The combined sample was over half a million people.

They found one dimension along which men and women differ more than on anything else in psychology. They called it Things versus People.

5.2 — The numbers

Two things in that table deserve attention beyond the top line.

Interest in mathematics differs by 0.34 while ability differs by roughly zero. That is the whole argument about mathematics and women, relocated. The gap was never in what girls could do. It was, at least substantially, in what they wanted to do with it — which is a completely different problem requiring a completely different response.

And business interest is dead level. An effect size of 0.04 is nothing at all. Whatever explains the composition of company boards, it is not a difference in wanting to be there.

5.3 — Where does the preference come from?

The obvious reply is that interests are shaped by society, and this is clearly partly true. Children are handed different toys, praised for different things, and shown different futures. Nobody serious disputes that this affects what they later want.

Two findings complicate the simple version.

The first is that the Things–People difference appears in every country where it has been measured, across more than fifty nations with very different economies, religions and gender arrangements.

The second is the one that does real damage to the simple version, and it is the subject of the whole of Part Five: the difference tends to be larger in richer and more gender-equal countries, not smaller. Whatever produces it, the pattern does not shrink as barriers come down.

I am not going to resolve that here. Part Five is entirely about it, including the serious criticisms of the finding. What matters for this chapter is that the straightforward socialisation account predicts the opposite of what is observed, and an account that predicts the opposite of the observation has a problem.

I want to be clear that naming the dilemma is not the same as resolving it, and I am not resolving it here. Chapter Nine takes it up properly, once the rest of the measurements are on the table.

6Personality, And A Statistical Trap

Measured one trait at a time, personality differences are modest. Measured all at once by one particular method, they look enormous. And one disposition — aggression — shows more clearly than anything else in this part what a moderate average does at the far edge of a distribution.

6.1 — One trait at a time

Personality research mostly organises itself around five broad traits, and the sex differences on them have been measured many times in many countries.

Taken separately, the findings are consistent and moderate. Women score higher on agreeableness by around 0.4 to 0.5 — roughly sixty-two to sixty-four in the hall, running the other way. Women score higher on neuroticism by a similar amount. Differences on conscientiousness and openness at the broad level are small. Extraversion is close to level.

There is one wrinkle worth knowing, because it explains part of the dispute that follows. Each broad trait contains narrower components, and the sexes sometimes differ in opposite directions within the same trait. Within extraversion, for instance, women tend to score higher on warmth and enthusiasm while men score higher on assertiveness. Average those together and they cancel, producing a broad difference near zero that conceals two real ones.

6.2 — The move

Now the trap.

Suppose you have sixteen personality measurements rather than five. On each one, men and women differ modestly. What happens if you ask not “how different are they on this trait” but “how well could you tell a man from a woman using all sixteen at once?”

That is a different question, and it has a different answer. Because the small differences do not all point the same way, combining them can separate the groups far better than any single measure. A team applying this method to a large sample reported a combined distance so large that it implied only around one tenth of the two groups overlapped — a figure vastly beyond anything in Chapter Three.

That claim has been quoted enormously widely, usually without the argument that followed it.

6.3 — Aggression, and the clearest tail in the book

One disposition deserves separate treatment, because it is among the largest behavioural differences and because it demonstrates Chapter Four’s arithmetic more clearly than any other example available.

Men are more physically aggressive. Meta-analyses put the difference somewhere around 0.5 to 0.6 — in the hall, roughly sixty-four to sixty-six in a hundred. That is a moderate difference, comparable to writing ability running the other way, and if you stopped there you would conclude it was interesting and not dramatic.

Two details make it more than that.

It appears very early. The sex difference in physical aggression is detectable in children by around the age of two or three, before most proposed socialisation could plausibly have done its work, and it is found across cultures.

And indirect aggression runs the other way. Aggression conducted through exclusion, reputation and social manoeuvre — damaging somebody without touching them — shows either no sex difference or one favouring girls and women, depending on age and measure. So the honest statement is not that one sex is more aggressive. It is that the sexes differ in method, and that the male method is the one that leaves marks and enters crime statistics.

The same structure applies to risk. Men take more physical and financial risk, by a modest margin on most measures — the effect varies a great deal by the kind of risk, and is smaller than most people assume. And men die from accidents, drowning, falls and violence at rates far beyond that modest margin, for the same reason.

Hold this beside §2.5. Men die around five years earlier, everywhere. A meaningful share of that gap is not a fact about male bodies at all. It is the tail of a moderate difference in behaviour, arriving as a mortality statistic.

6.4 — What survives

Strip the dispute away and three things are solid.

Women score higher on agreeableness and on neuroticism, by moderate amounts, consistently, across many countries. These are among the more replicable findings in personality research.

Broad traits hide opposite-running components, so a reported difference of zero at the broad level is not evidence of no difference underneath.

And personality differences, like the interest differences in Chapter Five, tend to be larger in richer and more gender-equal countries. The same pattern, in a different literature, measured by different people. Part Five deals with it properly, and the fact that it appears independently in interests and in personality is the reason it cannot be waved away.

7What Is Actually In The Brain

More has been written about male and female brains than about any other topic in this part, and a large share of it is unreliable. This chapter separates what survives scrutiny from what does not, and the second pile is much bigger.

7.1 — The one solid number, and why it means less than it looks

Men’s brains are larger, by roughly ten to twelve per cent on average. This is not in dispute.

Nearly all of it is body size. Larger bodies come with larger brains across the animal kingdom, and within each sex the same relationship holds — taller people have bigger brains. Once you account for body size, most of the gap goes.

And brain size is a poor predictor of anything cognitive within the human range. The correlation between brain volume and measured intelligence is real and modest, and it does not do the work that people who cite brain size want it to do — which is why Chapter Three found essentially no difference in general intelligence despite this ten per cent gap.

7.2 — What is reliably found

The best current evidence comes from very large population imaging studies rather than the small laboratory samples that dominated earlier decades.

Those studies do find average differences: in the raw volumes of many regions, in the proportions of grey and white matter, and in cortical thickness, where women tend to score higher. The differences are statistically robust at these sample sizes.

They are also, without exception, differences with large overlap. Not one of them separates the sexes. In the language of Chapter One, they sit in the small-to-moderate range, and most shrink further once total brain size is accounted for.

So the honest summary of the reliable brain literature is: measurable average differences in many structures, none of them large enough to characterise an individual, and none of them yet connected to a specific behavioural difference by a demonstrated causal chain.

That last clause is the important one, and it is where most popular writing goes wrong. Finding that a region differs on average is not the same as showing that the difference causes anything.

7.3 — A famous finding that was never there

That case is not an isolated embarrassment. It is representative of a structural problem in this literature, and the scale of the problem is worth stating with a number attached.

7.4 — Two brains or a mosaic?

So the practical instruction from this chapter is unglamorous. Treat any striking claim about male and female brains as probably wrong until you know the sample size, and treat any claim that connects a brain difference to a behaviour as almost certainly unestablished, because that chain has very rarely been demonstrated.

8Where The Differences Come From

Having measured the differences, the obvious next question is whether they are built in or taught. That question has been dead in the research for about fifty years, and understanding why it died is more useful than any answer to it.

8.1 — The strongest evidence available

You cannot run the experiment. Assigning babies to hormone levels or to upbringings is not something any ethics committee will approve, and the one notorious attempt to raise a child as the opposite sex was a single case, ended in tragedy, and is cited with equal confidence by both camps — which is what a single case always permits and never justifies.

So researchers use situations where nature has arranged something close to an experiment.

The comparison with sisters is what makes this worth the space. Two girls in one house, one exposed to high androgens before birth and one not, raised by the same parents with the same expectations. Whatever differs between them cannot easily be explained by upbringing.

8.2 — Monkeys and toys

A second line of evidence removes human socialisation entirely.

Researchers offered toys to monkeys — animals with no exposure to advertising, parental expectation, school, or any idea of what a toy is for. Male monkeys spent more time with wheeled objects and female monkeys more with plush ones. The result has been found in more than one species.

It is striking and it should not be over-read. Critics have pointed out that the objects differed in colour, shape and how they moved, any of which could drive a preference without anything sex-typed being involved. The effects are moderate. And a monkey has no concept of a doll, so what the animals were responding to is genuinely unclear.

What the studies do establish is narrower and still useful: a sex difference in object preference can arise without human culture. That does not tell you how much of the human difference is cultural. It tells you the answer is not all of it.

8.3 — The most misused number in the subject

Before the framing itself, one term has to be dealt with, because it appears in almost every argument about this and almost nobody using it means what it means.

8.4 — Why the question is broken

8.4 — The honest scoring

Given all that, here is what can actually be said about causes, scored plainly.

Physical differences (Chapter Two): overwhelmingly biological. Driven by hormones at puberty, visible in every population, and unchanged by any social arrangement anybody has tried. Training moves individuals substantially and does not close the population gap.

Interest differences (Chapter Five): mixed, with a real biological component and a real social one. The prenatal hormone evidence and the animal work establish that some of it is not taught. The size of the effect in humans, and the share of the total it accounts for, are not established. Anybody giving you a percentage is guessing.

Cognitive differences (Chapter Three): mostly too small to have an interesting cause. Where a difference exists and is large enough to study, such as mental rotation, it responds to training — which does not make it purely learned, but does mean it is not fixed.

Personality differences (Chapter Six): unresolved, and complicated by the cross-country pattern. Differences being larger in freer societies is not what either a simple biological or a simple social account predicts. Part Five is about that.

And the tail differences (Chapter Four): unknown. The mathematics ratio moved by two thirds in twenty years, which rules out anything wholly fixed, and it then stopped moving, which rules out nothing else.

9Same, Equal, And Equal Outcomes

Three completely different ideas share one word, and almost every argument about equality is two people using it to mean different things. Separating them dissolves most of the argument — and leaves one genuine dilemma that nobody has an answer to.

9.1 — Three ideas

Take the sentence “men and women are equal.” It can mean three unrelated things.

Same. That men and women have the same distribution of traits. This is an empirical claim. It can be measured, and eight chapters have just measured it. The answer is: identical or nearly so on most cognitive measures, substantially different on some physical ones, and substantially different on interests. So as a general claim it is false, and as a claim about any specific trait it needs checking one trait at a time.

Equal in worth. That men and women are owed the same consideration, the same rights, the same standing. This is a moral claim. No measurement bears on it. You cannot confirm it with a study and you cannot refute it with one.

Equal in outcome. That men and women should end up in the same proportions in every role, at every level, with the same pay and the same representation. This is a political claim about results, and whether it follows from anything depends entirely on arguments that have to be made separately.

Now the observation that matters. Nobody has ever derived the second from the first, and nobody needs to.

Consider how obvious this is elsewhere. A weak man and a strong man are not the same on any measure of strength — the difference is larger than most in this part. Nobody concludes that the weak man has fewer rights. A person with a measured intelligence of 90 and one with 140 differ enormously. Nobody thinks the first should be governed by the second.

Equality of worth was never a claim about measurements. It has never rested on sameness in any other context, and there is no reason it should be made to rest on sameness here.

Which means the entire eight chapters you have just read are irrelevant to it. Every number in this part could double and the moral claim would be untouched.

9.2 — What separate categories actually concede

There is a version of this worth taking seriously because it is what people usually mean when they say the sexes cannot really be equal. It goes: if a man lifts a hundred kilos and a woman lifts fifty, then calling them equal is a fiction — real equality would mean the same number, and since the numbers differ, the equality is invented.

Look at what competitive sport actually does with that problem, because sport has been running the experiment for a century.

It creates separate categories. And a separate category is an explicit, public admission that the sameness claim is false in this domain — nobody sets up a women’s hundred metres if the times are the same.

But notice what it is not conceding. It is not conceding that women’s competition is less worth having, or that female athletes are lesser people, or that the winner deserves less. Weight classes in boxing work identically, and nobody thinks a flyweight champion is a fraud.

So the argument proves less than it seems to. Separate categories are not a fudge covering an embarrassment. They are what you do when a difference is real and the activity is still worth having for both groups. That is a coherent position, held openly, in public, by millions of people every weekend.

There is a limit to it, and the limit is the useful part.

Adjusted standards make sense where the activity is the point. A race exists to find the fastest person in a field. You can define the field however you like, because the race has no purpose outside itself.

Absolute standards make sense where the outcome is the point. A person trapped in a burning building weighs what they weigh. There is no adjusted category of human being to be carried down a ladder. Where a job genuinely requires a physical capacity, the honest arrangement is a standard set by the task and applied to every applicant regardless of sex — which some women pass and many men fail, and which is a test of the person rather than of the category.

Almost every real dispute about physical standards is a dispute about which of those two situations applies, conducted by people who have not noticed there are two.

9.3 — The dilemma nobody wants

Now the genuine problem, which the separation above does not dissolve.

Chapter Five found that men and women differ substantially in what they want to do — around 0.93 on the Things–People dimension, larger than any ability difference in this part.

Follow that through. If a society removed every barrier — no discrimination, no harassment, no discouragement, nothing but people choosing freely — the outcomes would not be equal. Engineering would remain disproportionately male. Nursing and primary teaching would remain disproportionately female. Not because anybody was stopped, but because of what people chose.

Which produces the following, and it is the crux of the entire modern argument:

Equality of opportunity and equality of outcome are not two ways of saying the same thing. Given a real difference in preferences, they are in conflict, and you cannot have both.

Neither camp can accept this cleanly.

For the equality-of-outcome position, it means that any programme aiming at proportional representation must, past some point, be working against what people want — and the honest versions of that programme have to say either that the preferences are illegitimate, or that the outcome matters more than the preference. Both are sayable. Neither is usually said.

For the traditionalist position, it is worse than it looks, and the reason is the pattern in Chapters Five and Six. The interest and personality differences get larger in freer and richer societies. Which means the preferences being appealed to are most visible precisely where women have most choice — so they cannot easily be used to justify restricting choice. You do not need a rule to make people do what they were going to do anyway. Every rule in Parts Two and Three exists to stop somebody doing something. A preference argument cannot justify a prohibition, because a preference needs no enforcement.

And the whole thing sits on an unresolved question. Preferences form inside societies. If they were installed, the argument changes completely. The cross-country pattern is the strongest evidence against the simple installation story, and it is contested, and Part Five is about it.

9.4 — What this part actually established about the rules

Set every finding beside the thing this series is about.

The largest difference measured here is physical, in power and speed. It does not explain why a widow may not remarry.

The next largest is interests. It does not explain why a woman must marry inside her caste.

Cognitive differences are mostly close to zero. They do not explain seclusion, dowry, or a girl married before puberty.

Male variability is real at the extremes. It does not explain why a photograph of a woman at a protest was answered with a rumour about her body.

Not one measured difference in this part maps onto a single rule in Parts Two and Three. The rules are not what you would design if you were responding to these findings; they bear no relationship to them at all. They are what you would design if you were solving an inheritance problem and a boundary problem, which is what Parts Two and Three found.

What I can do is keep saying it. Every time a number in this part is quoted at you as though it settled something about what a woman may do, the missing step is the same one, and it has never been supplied by anybody.

10An Honest List Of What We Do Not Know

Two lists, no hedging in either. What is genuinely unknown about differences between men and women, with the reason. And what is solid enough to carry into the rest of the series.

10.1 — Genuinely unknown

This part rests on measurement, which makes its gaps sharper and easier to state than the historical parts.

How much of the interest difference is biological. This is the central unknown and it is the one everybody claims to know. The prenatal hormone work and the animal studies establish that some of it is not taught. Nothing establishes the proportion, and the cross-country pattern points in a direction that neither simple account predicts. Anybody quoting you a percentage is guessing.

Why the differences are larger in freer societies. The pattern appears independently in interests, in personality and in preferences, measured by different teams using different instruments — which is why it cannot be dismissed. Every proposed explanation has serious problems, and the finding itself has been challenged on methodological grounds. Part Five is about this and will not resolve it either.

What produces the difference at the extremes. The mathematics ratio at the very top fell by roughly two thirds in twenty years and then stopped. That the movement happened rules out anything wholly fixed. That it stopped rules out nothing else, and nobody can currently distinguish a biological floor from a social one that has ceased shifting.

Whether variance ratios are stable. They differ between countries and have changed in some. A biological constant does not vary by nation, so something social is involved — how much, nobody has established.

How much of the brain literature is real. Chapter Seven’s honest position is that a large share of published findings came from samples now understood to be far too small, compounded by a statistical problem discovered in 2016. Which specific results survive is not knowable in advance, and the field is only now producing studies at the required scale.

Whether the mental rotation gap means anything outside a laboratory. It is the largest cognitive difference and it is one narrow task. Its connection to real-world outcomes is modest and its relationship to broader spatial ability is unclear.

10.2 — Solid

That physical differences are very large, and narrow. Height near an effect size of two, strength larger still, grip so separated that the distributions barely meet. And confined to power, speed and force application — shrinking in ultra-endurance events and reversing where body composition helps.

That most cognitive differences are close to zero. A review of forty-six meta-analyses found around four fifths small or negligible. This is the headline finding of the field and the least reported.

That average mathematical ability does not differ. Millions of students, multiple countries, favouring girls in several. Settled.

That girls read and write better, everywhere measured. The reading advantage appears in every participating country in every cycle. The writing gap is around half an effect size — roughly ten times the maths gap, and largely absent from public discussion.

That the Things–People interest difference is the largest measured psychological sex difference. About 0.93, from over half a million people, present in every country studied.

That greater male variability is real and operates at both ends. The arithmetic is not a theory. And it puts more men at the bottom in exactly the proportion it puts more at the top.

That equal averages are compatible with very unequal extremes. A seven per cent wider spread produces nearly three to one at four spreads above the mean, with identical means. This is arithmetic and cannot be argued with.

That heritability within a group says nothing about differences between groups. Not contested by anybody who works on this, and misused constantly.

And that none of the above bears on what anybody is owed. Established in Part One, restated in Chapter Nine, and the only conclusion in this part that does not depend on a single measurement.

10.3 — Who this was measured on

One closing observation, following the pattern of the previous parts.

Look at where these findings come from. University students, mostly. School assessments in wealthy countries. Population databases in Britain and Scandinavia. Volunteers who answered online surveys in English.

Part One named this problem: the people most studied by psychology are among the least typical humans available, drawn from countries holding a small minority of the world’s population. Nearly every number in this part carries that limitation.

The specifically Indian version is worse. There is comparatively little of this research on Indian populations, and where cross-national work includes India it is usually as one data point among dozens. So a series about Indian women has just spent sixty pages on measurements taken largely elsewhere.

I have applied them anyway, because the alternative is applying nothing. But the reader is entitled to know that when this part says “men and women differ by 0.93 on interests”, the sentence is built substantially out of people who have never been to the country this series is about — and that the one finding most relevant to India, the pattern in Chapter Five and Six about freer societies, is precisely the one nobody can explain.

Every Difference, In One Table

This part had no chronology to offer, so its reference table is the measurements themselves. Every difference discussed, sorted from largest to smallest, with the hall figure attached. Positive means men score higher.

Three things are visible in that table that are not visible in any single chapter.

The top of it is entirely physical and interests. Not one ability appears above 0.5 except a single narrow laboratory task.

The middle is a coin toss. Everything from mental rotation down to writing sits between 40 and 62 in a hall of a hundred — which means that for any of these, sex is a poor guide to a person.

And the largest non-physical differences at both ends are about what people want, not what they can do. Engineering at one end, caring work at the other. The abilities are in the middle where nothing much happens.