BACKGROUND Increased familial risks in multiple sclerosis (MS) range from 300-fold for monozygotic twins to 20-40-fold for biological first-degree relatives, suggesting a genetic influence. Yet if one identical twin has MS the other usually will not. One way of sorting out the contributions of genes and environment is to study half-sibs. METHODS In a Canadian population-based sample of 16 000 MS cases seen at 14 regional MS clinics one half-sib (or more) was reported by 939 index cases. By interview we elicited information on family structure and an illness in half-sibs and any full brothers or sisters. FINDINGS The age-adjusted MS rate in the 1839 half-sibs of these index cases was 1.32 percent compared with 3.46 percent for the 1395 full sibs of the same cases (p<0.001; likelihood ratio test). There were no significant differences in risk for maternal versus paternal half-sibs (1.42 percent vs 1.19 percent) or for those raised together versus those raised apart from the index case (1.17 percent vs 1.47 percent). INTERPRETATION Besides demonstrating the power and the feasibility of using half-sib studies to throw light on the aetiology of complex disorders, our findings show that a shared environment does not account for familial risk in MS and that maternal effects (such as intrauterine and perinatal factors, breastfeeding, and genomic imprinting) have no demonstrable effect on familial risk. Halving the number of potentially contributory genes (by comparing full-sib and half-sib rates) lowers the risk of MS by a factor of 2.62, an observation consistent with a polygenic hypothesis.