A method is presented for partitioning the variance associated with human smoking behavior into additive genetic, nonadditive genetic, prenatal environmental, postnatal familial environmental, and postnatal extrafamilial environmental components. Estimations can also be made of additive genetic and residual correlations between spouses and of the correlation between parental additive genetic effect and progeny nonadditive genetic and environmental effect. The variance estimates are free of the biases that might result from these correlations. The statistical genetic analysis is being applied to a large group of MZ and DZ twins, their spouses, and their adult children who live in southern Sweden. Blood samples from each subject will be used to identify their genetic constitution for a number of biochemical polymorphisms, some of which may be considered a priori to have possible relationships to smoking. Associations and genetic linkages between biochemical marker loci and quantitative behavioral traits will be sought. Traits of interest include a wide array of tobacco-use variables, motives for smoking, personality and cognitive variables, and other variables associated with drug use and health. Zygosity determinations based on biochemical polymorphisms have indicated MZ to DZ and DZ to MZ misclassification rates of 0% and 6.15%, respectively, when based solely on external morphology and questionnaire data. The nonpaternity ratio of the fathers with respect to their supposedly biological children is estimated to be 0.28%. Gene frequency estimates for 21 marker loci show that the sample of twins and their relatives is quite representative of the Swedish population at large. All loci were in Hardy-Weinberg-Castle equilibrium, with no evidence of assortative mating for biochemical traits. The MZ twins are significantly more concordant than the DZ twins with respect to whether they have ever had a smoking habit.