A 2003 paper in this journal reported results from a large sample of twins assessed at 2, 3 and 4 years of age on parent-administered tests and reports of their verbal and nonverbal ability. We found clear evidence for phenotypic general cognitive ability (g) that accounted for about 50% of the variance, for modest genetic influence on g (about 25% of the variance), and for substantial shared environmental influence (about 65%). As part of the same longitudinal study, we assessed g at 7 years using telephone-administered tests (4530 pairs), at 9 years using parent-supervised booklets mailed to the homes (2886 pairs), and at 10 years using tests administered online (2348 pairs). Despite very different measurement techniques, consistent evidence was found for phenotypic g accounting for about 50% of the variance. Compared to early childhood, g scores in middle childhood showed stronger genetic influence (about 40%) and more modest shared environmental influence (about 30%). Longitudinal analysis revealed strong genetic correlations (r = 0.71-0.81) across age and method of administration, with genes (about 65%) and shared environment (about 35%) accounting for stability, while genes, shared environment and non-shared environment all contributed to change.