H567/03 Section B Option 1 - Van Leeuwen et al. (2008) and assessing intelligence

Why include an extended family?

Relatives may resemble one another because of inherited differences, shared experiences, parents' influence on learning, or combinations of these. Monozygotic twins share almost all their DNA; dizygotic twins share, on average, half their segregating genes. Greater MZ than DZ resemblance supports genetic influence, but a simple comparison can be distorted if assumptions about parents and environments are wrong.

The researchers extended the twin design to include siblings and biological parents. This allowed models to consider assortative mating and cultural transmission alongside inherited influences. Phenotypic assortment means partners resemble each other because selection relates to observable characteristics such as intelligence or a related trait. Social homogamy means resemblance arises because partners meet within similar social environments. These produce different expectations about genetic and environmental resemblance.

Cultural transmission here means an environmental influence from parents to offspring, modelled through the relationship between parental IQ and the offspring environment. It is narrower than every possible meaning of culture. Gene–environment interaction concerns whether environmental effects differ with genetic characteristics. Gene–environment correlation concerns genes and environments being associated.

The study aimed to assess assortative mating, gene–environment interaction and the heritability of childhood intelligence. Additive genetic effects were expected to explain a large part of individual differences. Competing family models tested whether phenotypic assortment or social homogamy better accounted for the data. This was a quantitative correlational twin-family investigation with naturally occurring relationships, not an experiment that allocated genes or home environments.

The families and testing procedure

Families were recruited from the Netherlands Twin Registry. Two twin birth cohorts, 1995–1996, were selected with an additional sibling in the intended 9–14-year range. Of 214 families invited by letter and subsequently contacted by telephone, 112 agreed to participate, approximately 52%. They came from across the Netherlands. There was no significant difference in maternal education between participating and non-participating families, although this does not rule out every participation bias.

The twins' mean age at assessment was 9.1 years, with a range of 8.9–9.5. The pairs comprised 23 MZ male, 25 MZ female, 23 DZ male, 21 DZ female and 20 opposite-sex DZ pairs: 48 MZ and 64 DZ pairs altogether. There were 103 participating full siblings, 59 female, with a mean age of 11.9 and observed range of 9.9–14.9. The biological parents included 94 fathers, mean age 43.7, and 95 mothers, mean age 41.9. The study described 516 participants overall.

Selection also reflected a wider MRI project. Families were excluded for relevant medical, psychiatric, special-education, physical or sensory conditions; MRI-related exclusions included a pacemaker or metal in the head. Thus the sample does not represent every child, particularly children whose assessment needs may differ.

The project collected cognitive, behavioural, hormonal, pubertal and structural MRI data over two days. Cheek swabs for DNA were collected at home. For cognitive testing, families arrived between 9 and 11 am. Children were tested individually in separate rooms with a test battery including Raven's Standard Progressive Matrices; parents completed the Advanced version. The whole protocol lasted about five hours with two short breaks and a lunch break. The IQ analysis should not be mistaken for a finding that an MRI measurement directly determined intelligence.

Zygosity was determined using eight genetic markers in 110 pairs and questionnaire information in the remaining two. The research had committee approval. Parents signed consent for themselves and their children, children also signed their own forms, travel expenses were reimbursed and children received a present.

How Raven's matrices measured intelligence

A Raven matrix presents a visual pattern with a missing element. The participant selects the option that completes the relationship. It provides an index of general reasoning, especially fluid or non-verbal reasoning, rather than a complete inventory of knowledge, creativity or social skill.

Children completed the Standard Progressive Matrices at their own pace after verbal instructions. It contains 60 problems in five sets of 12, becoming more difficult within each set. Parents used the Advanced Progressive Matrices because the Standard version could be too easy. They received written instructions, 12 familiarisation items and 36 progressively complex test items. The paper cited retest reliability of .88 for children and .91 for adults; these are reported properties of the instruments, not reliability newly established by repeatedly testing this sample.

Different versions have different difficulty and maximum raw scores. A parent's raw mark out of 36 therefore cannot simply be compared with a child's mark out of 60. The researchers used Rasch modelling to estimate ability while considering item difficulty. They also modelled item responses directly to account for unreliability and differing precision across the scales. This supported more defensible family comparisons than comparing unadjusted totals.

A model remains an assumption-based representation. The Rasch approach assumes the items adequately measure a common dimension, and the authors discussed imperfect fit and possible bias. Visual problems reduce dependence on vocabulary, but instructions, schooling, puzzle experience and vision can still influence performance. A non-verbal measure is therefore not automatically culture-free.