To optimize the information obtained in '*fluorodeoxyglucose (FDG) studies of human cerebral metabolism, sources of metabolic variation should be identified and their relevance to clinical pathology and disease course determined. Specifically, although a definitive a priori hypothesis may be postulated in a given study, the sensitivity of a statistical procedure for detecting differences may be compromised if the sources of variability in the dependent variable are not fully understood. The purpose of this paper is to illustrate these statements using data derived from normal subjects, patients with symptomatic Huntington's disease, and patients at risk for Huntington's disease. With respect to positron emission tomography (PET) and FDG, a number of centers have published their findings. Although there is a general consensus that reductions in caudate metabolism are present, there is considerable debate regarding the time in the disease course such reductions occur, the best procedure for measuring such reductions, and the sensitivity of the PET/FDG methodology as a means of monitoring disease progression. The resolution of these issues is hampered by the large variation among individuals in time of onset and severity of disease course. Because of these properties, statistical examination of regional cerebral metabolism data in Huntington's disease may provide insight into the most effective procedures for maximizing data yield.