IELTS Reading: Matching features

Almost everybody has a view about birth order. The eldest is conscientious and anxious to please, the youngest is charming and takes risks, the middle child is overlooked and becomes a negotiator. The belief is held with remarkable confidence by people who would not claim to know anything else about personality, and it has a respectable scientific version. On that account, siblings compete for the attention of the same parents, and since the eldest occupies the obvious position of the responsible high achiever, later children must differentiate themselves by finding an unoccupied role, which tends towards the unconventional. It is an elegant theory and it fits everybody's family.

It also drew support from a substantial research literature, and the way that literature went wrong is worth understanding, because the error is subtle and was invisible for decades. Most of the early studies compared people from different families: a sample of adults was assembled, each was asked about their position among their siblings, and the groups were compared. The difficulty is that birth order and family size are not independent. If you select people at random, a first-born is more likely to come from a small family, because every family has exactly one, while fourth-borns can only come from families of four or more. Family size in turn correlates with income, parental education and a great deal else. Any comparison across families is therefore comparing social background as much as sibling position.

The correction, when it came, required data nobody had previously assembled: very large panel surveys containing several siblings from the same household, so that brothers and sisters could be compared with one another rather than with strangers. Within a family, the confound disappears. Studies of this kind, using tens of thousands of individuals across several countries, agree closely with one another and produce two findings. On measured intelligence there is a small but consistent advantage to being born earlier, on the order of one and a half points between a first and a second child. On personality, once the appropriate comparison is made, the effects are indistinguishable from zero. Not small. Zero.

The intelligence finding is real and requires an explanation, and there are three candidates. One is dilution: parental time and money divided among more children leave less for each, and the eldest has a period with none of that division. A second is tutoring: an older child who explains something to a younger one understands it better afterwards, and only the eldest has a full complement of younger siblings to teach. A third is biological, having to do with the changing uterine environment across pregnancies. The third can be tested against the other two, and the test is unusually clean. In a very large study of military conscripts, men who were biologically second-born but had been raised as the eldest because an older sibling died in infancy scored like first-borns rather than like second-borns. Whatever is producing the effect follows social position in the family, not the order of births.

Why, then, does everybody believe the personality version? Partly because families supply endless confirming instances and no disconfirming ones, since a family with an unconventional eldest is filed under exceptions. Partly because the theory explains things one already knows about one's own relatives, and an explanation that arrives after the fact is very hard to reject. And partly because the effects that do exist are far too small to be visible in a single case: a difference of one and a half points on a test with a standard deviation of fifteen is real across a hundred thousand people and completely invisible across a dinner table.

Not everybody accepts that the matter is settled. Tomas Wreford, who works on family dynamics, makes two objections. The first is statistical: a within-family design removes everything that families differ by, and some of what it removes may be exactly what birth order operates through, since a family's whole way of proceeding may be shaped by how many children arrived and in what order. The second is about measurement. The questionnaires used to assess personality in these studies produce five broad scores, and the folk theory is not really about broad traits at all; it is about roles, about who is expected to organise the holiday and who is forgiven for not turning up, and an instrument that measures conscientiousness on a nine-point scale will not detect that.

The second objection has some force and the first does not survive inspection. A within-family design does discard between-family variation, which is precisely what it is for, since that variation was the source of the original error; the proposal to put it back is a proposal to reinstate the confound. On measurement, Wreford is right that the instruments are coarse, and it is entirely possible that something real is going on which five broad scores cannot capture. But the burden sits with the claim. A theory that predicts effects too subtle for any available instrument to register has stopped predicting anything, and the correct response is to build the instrument rather than to treat the absence of evidence as a technicality.

What makes the episode worth studying is not the conclusion but the route to it. For half a century a large, active and perfectly competent research field produced consistent support for a hypothesis, and the support was an artefact of a design choice that everybody had inherited and nobody had examined. The studies were not fraudulent and the researchers were not careless; they were all making the same comparison, and the comparison was wrong. That is a more unsettling story than any individual failure, and it is the reason methodological arguments, which are tedious, deserve more attention than they get.