Neuropathological investigations have demonstrated brain-behavior relationships in senile dementia of the Alzheimer's type (SDAT), but CT studies have not produced consistent findings. We hypothesized that these discouraging results were in part due to limitations in the methods of CT scan evaluations, and to non-homogeneity of patient populations. The present study examined 43 out-patients with the presumptive diagnosis of SDAT using 37 cognitive test measures and 3 independent CT evaluation strategies. The CT methods included a new rank ordering procedure and two previously used techniques, physical measurement and 4-point rating. Highly significant (p less than or equal to 0.01) brain-behavior correlations were attained using the ranking and rating procedures for evaluation of ventricular and cortical pathology. It was found that rank ordering has high interrater reliability and is superior to the other methods for the evaluation of the ventricular system. The physical measurement of the third ventricle is the single most powerful linear correlate of cognitive impairment. Measurement of cortical sulci are of no correlational significance. Multiple regression analyses indicated that global assessments are the best cognitive predictors of both ventricular and cortical pathology. Thus the present study has demonstrated brain-behavior relationships in vivo in SDAT.