Linking the Standard and Advanced Raven Progressive Matrices tests to model intelligence covariance in twin families

An abundance of research shows significant resemblance in standardized IQ scores in children and their biological parents. Twin and family studies based on such standardized scores suggest that a large proportion of the resemblance is due to genetic transmission, rather than cultural transmission. However, most studies used standardized intelligence scores that were based on different tests for different age groups, which makes it hard to say if the exact same construct is measured. Here we re-analyze intelligence data on two different versions of the Raven Progressive Matrices test, collected in Dutch twin children (Standard test version) and their biological parents (Advanced test version). First, the data from parents and their offspring were harmonized using test linking through an item response theory measurement model. This required collecting data from extra participants who were assessed with items from both test versions. Next, the raw item data were analyzed to study transmission of intelligence, correcting for the differences in difficulty of the items in the parental and child test versions and differences in measurement reliability. Results showed a significant difference in the phenotypic variance in intelligence in the two generations. Model fitting showed that the surplus variance in the parental generation is likely due to surplus environmental variance that is not transmitted to the offspring. This could reflect that there was extra measurement error under the parental testing conditions. Genetic modelling showed that intelligence covariance in parents and their children is most likely based on genetic transmission without cultural transmission.