Educational performance in twins

In this week's BMJ, Christensen and colleagues1 investigate two questions that are of popular and medical-scientific interest. Firstly, do twins have lower intellectual skills and educational achievements than singletons and, secondly, is birth weight associated with intellectual and educational performance? The authors look at the second question in both twins and singletons. They also open up the question of the link between intelligence and education, because the main comparator studies used IQ-type outcomes rather than educational performance.2 3

The study uses the Danish registration system linked with the Danish demographic database, the national hospital discharge register, the register of compulsory school completion assessments and test scores, and the Danish twin registry. They therefore had data on standard national educational outcomes, birth weights, and other demographic and parental variables for the entire relevant population of twins, and for a large representative sample of the comparable population of singletons.

Firstly, do twins have lower intellectual skills and educational achievements? Three recent studies of large samples of Scottish children born between 1921 and the 1950s strongly suggest they do. The Scottish mental surveys of 1932 and 1947 tested the IQ of most 11 year old Scottish children born in 1921 and 1936, respectively. In both surveys, twins scored about 5 IQ points (one third of a standard deviation) lower than singletons.2 Father's social class, overcrowding in the home, height during childhood, school attendance, and number of people in the family did not account for the twin-singleton difference. The third study, in children from Aberdeen in the 1950s, found a similar twin-singleton difference in IQ test scores at ages 7 and 9.3 It had the added benefit of comparing twins and singletons in the same families. The authors of that study found partial attenuation of the effect after adjusting for birth weight and gestational age.

However, Christensen and colleagues found similar test scores for twins and singletons. One possible reason for their findings, apart from possible differences between countries and populations studied, is the age at testing (at least five years later in the present study). This is supported by a Dutch study of adult twins, which found no significant difference in IQ between singletons and twins from the same families.4 It is therefore possible that differences in ability or educational performance (or both) exist between twins and singletons as late as 11 years, but that they disappear by 16.

The comparison of Christensen and colleagues' findings with other recent large twin studies relies on there being a strong association between intelligence and educational performance. A large longitudinal representative study of more than 70 000 English schoolchildren supports this link.5 General intelligence scores at age 11 years, derived from a battery of 10 separate cognitive tests, were highly correlated (r >0.8) with general performance in the GCSE examinations at age 16. Interestingly, despite no differences in general intelligence being seen between boys and girls at age 11, girls performed considerably better in GCSEs at 16.

Secondly, is birth weight related to intellectual skills and achievements? Certainly, infants born well below the normal range of birth weights have some disadvantage.6 Few studies have focused on the normal range of birth weights and term births, and the variability in the design of such studies does not allow a meta-analysis to be carried out. Overall, though, a narrative systematic review of such studies found a small positive association between birth weight and childhood (up to age 17) IQ that was not accounted for by parental social class.7

The study by Christensen and colleagues also finds a small effect of birth weight on educational performance. Even that small effect could be due to confounding, however. A recent study based on the US national longitudinal survey of youth 1979 found that the small significant association between birth weight and language and mathematical achievements in childhood and adolescence was largely accounted for by the mother's IQ score. This indicates that brighter mothers have brighter and heavier children.8 If we put this potential confounding factor aside though, educational attainments are almost identical for each centile of birth weight (not absolute weight) within the singleton and twin groups. Christensen and colleagues conclude that the relative position of twins within their own group with respect to birth weight is most important. Such an effect is not unique to twins versus singletons or to birth weight and educational performance. Consider a comparable situation. A meta-analysis found a moderate sized correlation between brain volume and IQ9; men have, on average, bigger brains than women, yet men and women differ little in mean general intelligence.10 Therefore, it is possible for birth weight and IQ to be related within both singletons and twins, and for twins to be on average lighter than singletons, and for twins and singletons not to differ in intellectual skills and achievements.

There remain unresolved issues about the possible effects of differences in geography, year of birth, and age. But the mechanisms by which these factors could eliminate twin-singleton differences is unclear. Consider year of birth, for example. Christensen and colleagues suggest that better, more recent obstetric care might be responsible. But it is tenable that such differences in care might also assist the survival of more at-risk babies generally—twins as well as singletons. Therefore, it is not clear whether better obstetric care would reduce, increase, or not affect any prior twin-singleton difference in mental ability. Despite these issues, Christensen and colleagues' study is comprehensive and well executed enough to reverse a trend in our thinking—that twins perform less well than singletons.