We propose that an investigation of the serum digitalis concentration as a test for digitalis toxicity should (1) study patients with similar toxic manifestations, (2) obtain control concentrations from nontoxic patients with symptoms suggesting toxicity, (3) define criteria for toxicity and nontoxicity, (4) select representative patients, (5) describe the study population and (6) analyze how much diagnostic information the serum digitalis concentration provides that cannot be inferred from other observations. To determine if available evidence validates the digitalis concentration as a test for toxicity 27 reports were reviewed. No investigation employed symptomatic controls. Of five studies most consistent with points 1-5 only three demonstrated higher mean serum digitalis concentrations in toxic patients. Whether knowledge of the digitalis concentrations was diagnostically more useful than knowledge of the digitalis dosage, renal function, serum potassium concentration and cardiac status was not determined in any study. The usefulness of the serum digitalis concentration as a test for digitalis toxicity is therefore not established.