Heritability of interests: A twin study.

924 twin pairs reared together plus 92 pairs separated in infancy and reared apart; 291 interest and talent items reduced to 39 factors and 11 superfactors

Minnesota Twin Registry and Minnesota Study of Twins Reared Apart; adults reporting vocational and recreational interests and self-rated talents.

Classical twin design augmented by the reared-apart sample, which is what makes it worth citing here: twins separated in infancy still resemble each other in what kind of work they want to do, so the resemblance cannot be attributed to a shared home, a shared school, or a shared parental occupation. Standard equal-environments and representativeness caveats apply, and the reared-apart cell is small (92 pairs across all interest domains). The authors note evidence of NONADDITIVE genetic effects, which is why family-study heritabilities of interests come out much lower — a real methodological finding and also a reason to treat the 50% as an upper-ish bound on the additive component.

About 50% of the variance in vocational and recreational interests — roughly two thirds of the STABLE variance, once measurement error is set aside — is associated with genetic variation. Heritability rises with the level of aggregation, from items to factors to superfactors, exactly as it does for cognitive measures. The authors propose that heritable precursor traits of aptitude and personality guide the development of interests through gene-environment correlation and interaction — i.e. children select themselves toward the work they are already suited to.

Not applicable — this study IS the genetic design. It is included as the confound anchor for every observational comparison of vocational versus academic students: track choice is an expression of a substantially heritable trait, so "students who chose the vocational route earn less at 45" is not evidence that the route caused it.

The finding is one of the better-replicated results in behaviour genetics — Moloney, Bouchard & Segal (1991) report the same pattern in the reared-apart sample, and Schermer & Vernon (2008) reproduce it on the Jackson Vocational Interest Survey in an independent Canadian twin sample. Estimates vary from roughly 30% to 50% depending on aggregation level and on whether nonadditive variance is modelled.