We have characterized the erythrocytes, granulocytes, and platelets of 54 patients with paroxysmal nocturnal hemoglobinuria (PNH) with antibodies to glycosylphosphatidylinositol-anchored proteins (anti-CD55, anti-CD59, and anti-CD16) and flow cytometry to establish the usefulness of this technique in the diagnosis of this disorder. All patients demonstrated either completely (PNH III) or partially (PNH II) deficient red cells and granulocytes. Anti-CD59 best demonstrated PNH II red cells, which were present in 50% of the patients. The proportion of abnormal granulocytes was usually greater than the proportion of abnormal red cells; 37% of the patients had >80% abnormal granulocytes. Anti-CD55 did not delineate the erythrocyte populations as well as did anti-CD59. Either anti-CD55 or anti-CD59 could be used equally well to analyze granulocytes; anti-CD16 did not demonstrate cells of partial deficiency. Platelets could not be used for detailed analysis as the normal and abnormal populations were not well distinguished. Flow cytometry of erythrocytes using anti-CD59 or of granulocytes using either anti-CD55 or anti-CD59 provides the most accurate technique for the diagnosis of paroxysmal nocturnal hemoglobinuria; it is clearly more specific, more quantitative, and more sensitive than the tests for PNH that depend upon hemolysis by complement (the acidified serum lysis [Ham] test, the sucrose lysis test, and the complement lysis sensitivity [CLS] test).