In a retrospective analysis, correlations were sought between pretreatment electroencephalographic and clinical data obtained in 18 therapeutic studies conducted in elderly patients according to almost identical protocols. Power spectrum analysis was applied to the EEG tracings, while clinical status was observer-rated using the Sandoz clinical assessment - geriatric (SCAG) scale. The study population comprised a total of 286 patients between the ages of 51 and 97 years (median age 70 years), 162 of whom were male (median age 68 years) and 124 female (median age 73 years). A 2-week washout period and several 'adaptation recordings' preceded the pretreatment EEGs used in this analysis. These tracings were recorded under resting conditions between 8 and 10 a.m. and were followed by an assessment of clinical status. Spearman rank correlations with 4 EEG parameters--total slow waves, alpha and beta waves, and dominant alpha frequency--were computed for all 18 SCAG items and for 'overall impression of patient' as well as for 5 SCAG factors. All 96 - (19 + 5) X 4 - correlation coefficients were formally tested for statistical significance at the nominal level of alpha = 0.05. In this analysis, 9 of the 18 items, 'overall impression' and two of the factors ('apathy' and 'somatic dysfunction') showed nominally significant correlations with at least 1 of the 4 EEG variables. As expected, a positive correlation was found between percentage slow-wave activity and degree of clinical impairment. In addition correlations were identified between clinical data and alpha and beta activity.(ABSTRACT TRUNCATED AT 250 WORDS)