The Project Talent Twin and Sibling (PTTS) study includes 4481 multiples and their 522 nontwin siblings from 2233 families. The sample was drawn from Project Talent, a U.S. national longitudinal study of 377,000 individuals born 1942–1946, first assessed in 1960 and representative of U.S. students in secondary school (Grades 9–12). In addition to the twins and triplets, the 1960 dataset includes 84,000 siblings from 40,000 other families. This design is both genetically informative and unique in facilitating separation of the ‘common’ environment into three sources of variation: shared by all siblings within a family, specific to twin-pairs, and associated with school/community-level factors. We term this the GIFTS model for genetics, individual, family, twin, and school sources of variance. In our article published in a previous Twin Research and Human Genetics special issue, we described data collections conducted with the full Project Talent sample during 1960–1974, methods for the recent linking of siblings within families, identification of twins, and the design of a 54-year follow-up of the PTTS sample, when participants were 68–72 years old. In the current article, we summarize participation and data available from this 2014 collection, describe our method for assigning zygosity using survey responses and yearbook photographs, illustrate the GIFTS model applied to 1960 vocabulary scores from more than 80,000 adolescent twins, siblings and schoolmates and summarize the next wave of PTTS data collection being conducted as part of the larger Project Talent Aging Study.

We examined differences between school classes with respect to three aspects of psychosocial adjustment at school, namely the extent that children in the class liked to play with each other, the number of teacher-reported behaviour problems, and children's feelings of well-being at school. The sample consisted of 1282 4- to 5-year-olds from 94 school classes and 51 schools, but due to nonresponse actual sample sizes were somewhat smaller for most analyses. Multilevel analyses showed that on average 87% of the variance was at the child level, 11% at the class level, and 3% at the school level. This indicated that a non-negligible amount of variance could not be accounted for by factors at the child level. Furthermore, this variance was mainly associated with differences between classes instead of differences between schools. A set of variables that pertained to sociodemographic characteristics of schools, school facilities, organisational aspects of classrooms, and the teacher did not provide an adequate explanation for the differences in adjustment levels. In contrast to these traditional variables, social network indices yielded substantial correlations, showed consistent trends across the different adjustment measures, and fulfilled the necessary requirement that to explain differences between school classes the predictor variables themselves should differ for classes within the same school. These results suggested that aspects of the interpersonal relations of children in the classroom such as proximity, integration, and the amount of contact could be determinants of differences between school classes in psychosocial adjustment.