In this study we examine the hypothesis that monozygotic (MZ) twins in historical databases are less discordant for birth weight due to negative selection of severely discordant MZ twins. Furthermore, we test the hypothesis that MZ twins are less discordant for birth weight when comparing a volunteer based twin registry with a population based twin registry, due to selective registration. Data were available on 3927 twin pairs from the volunteer Australian Twin Registry born before 1964, 3059 volunteer twin pairs from the Netherlands Twin Register born 1987–1989 and 454 Belgian twin pairs from The East Flanders Prospective Twin Survey born 1987–1989. Intrapair relative birth weight differences (RBWD) were computed for MZ and dizygotic (DZ) twins from each twin registry. Comparing birth weight differences between MZ and DZ twins provides support for the hypothesis that MZ twins are subject to a negative selection in historical databases. Furthermore, Australian MZ twins have a lower RBWD compared to Dutch MZ twins when corrected for the RBWD of Australian and Dutch DZ twins, indicating circumstances which only affect MZ twins. Our hypothesis that MZ twins are less discordant for birth weight in a volunteer based twin registry compared to a population based twin registry had to be rejected. We suggest that investigators using historical databases to test the fetal origins hypothesis should be aware of this increased likelihood of selective exclusion of individuals with extreme morphometric parameters at time of birth.