Using genotypes of 280 marker loci on the 22 autosomes of 105 alcohol-dependent probands, their affected and unaffected sibs, as well as their parents, we iteratively constructed a genetic similarity function that enabled us to quantify the interindividual genetic distances d(x(i), xj) between feature vectors x(i), xj made up by the allelic patterns of individuals i, j with respect to loci l1, l2,...,ln. Based on this similarity function, we investigated the sib-sib similarities that are expected to deviate from "0.5" in affected sib pairs if the region of interest contains markers close to disease-causing genes. The reference value "0.5" was derived from the parents-offspring similarities, because these are independent of the affection status. The question of population admixture was addressed by means of multivariate structural analyses. These analyses led to four "natural" groups whose validity was tested through the father-mother similarities. Additionally, we determined the eigenvectors that optimally represented the genetic variation and found several marker configurations on chromosomes 1, 3, 7, 15, and 17 that reproducibly discriminated (p < or = 0.01) affected probands/sibs from unaffected sibs, while no such differences were found between affected probands and affected sibs.