BACKGROUND The risk for panic disorder (PD) is substantially increased in relatives of probands with PD. Prior literature provides only limited information about the degree to which this increase is due to genetic factors or family environment. METHODS In personal interviews with both members of 3194 twin pairs, we assessed the lifetime history of lifetime panic attacks and PD. Twin resemblance was assessed by tetrachoric correlation and single and multiple threshold biometrical model fitting. RESULTS As fully syndromal PD, by DSM-III-R criteria, was too rare to analyse usefully we examined four other dichotomous definitions of increasing stringency: panic probe and very broad, broad and intermediate PD. For all four definitions and for the multiple threshold analyses, the best-fit model indicated that twin resemblance was due solely to genetic factors with a moderate heritability (33-43%). For the broad and intermediate dichotomous definitions of PD, however, a model with twin resemblance due to familial-environmental factors fit nearly as well. No gender effects were seen on the genetic risk factors for these PD-like syndromes. CONCLUSIONS Even with large epidemiological samples of twins, studying disorders as uncommon as PD is problematical. Despite these difficulties, our results suggest that: (i) narrowly and broadly defined PD are probably on the same continuum of liability; (ii) twin resemblance for these PD-like syndromes is likely due largely to genetic factors with a moderate level of heritability although a contribution of familial-environmental factors cannot be excluded, and, (iii) the same familial risk factors impact. to a similar degree, on the liability to PD in males and females.