Since the emergence of twin studies in the 1920s, time and again the question of possible twin-singleton differences in intelligence has been posed. This study addressed the issue through a meta-analysis of published studies on this theme. Twins on the average seem to have lower IQs than singletons. The best estimate for this group difference is 4.2 IQ points (less than one-third of a standard deviation), with a great divide between study outcomes of less vs more recent birth cohorts (5.1 vs 0.5 IQ points, respectively). The evidence is based on studies from six countries (including population-based ones of entire birth cohorts), a massive database (comparisons of more than 30,000 twins with nearly 1.6 million singletons), a variety of intelligence tests, and birth cohorts spanning most of the 20th century, but, for the most part, on an age range limited to children and adolescents. The effect shows considerable between-study heterogeneity but appears robust (fail-safe N calculations), not due to influential individual studies (sensitivity analysis) or publication bias, was present since the very first published studies (cumulative meta-analysis), and appears generalizable across sex, zygosity, and other intelligence domains beyond the verbal. There are insufficient data as to whether the effect persists over the lifespan, exists as well within families, or has ceased in recent birth cohorts of highly developed countries (Denmark and The Netherlands). Likely causes of the effect appear to be prenatal and perinatal factors (reduced fetal growth and shorter gestation for twins).