We utilized pedigree discriminant and factor analytic approaches to combine multivariate phenotypic information into a single liability phenotype in the isolate and general populations. We applied two-stage relative-pair quantitative trait linkage analysis to detect genetic contributions to variation in the resulting liability phenotypes. Linkage analysis revealed several regions of suggestive linkage in both the general and isolate populations, the majority of which appear in retrospect to be false positives. A likely explanation is an overall lack of power given that we tested hypotheses in data from only one replicate. However, it may be possible that a construct that ignores affection status when using liability-associated characteristics as indicators of this construct is not the most effective approach in modeling the liability underlying a complex phenotype.