The number of polymorphonuclear leukocytes on gram-stained urethral specimens was compared with the number in the sediment of the first-voided urine to determine what constitutes an abnormal number of polymorphonuclear leukocytes in gram-stained specimens. Seventy-three men attending a clinic for sexually transmitted diseases were studied. An attempt was made to select primarily asymptomatic men and men who had minimal symptoms or signs of urethritis because the diagnosis of urethritis is most difficult in these groups. There was a distinct bimodal distribution of the numbers of polymorphonuclear leukocytes by both techniques. Means of four or fewer polymorphonuclear leukocytes per field in five fields (X 1,000) with gram stain and of < 15 polymorphonuclear leukocytes in all of five fields (X 400) in the first-voided urine sediment were considered normal. The results with the two techniques were highly correlated (P = 1.4 x 10(-10)). In the absence of cultures for Neisseria gonorrhoeae, Chlamydia trachomatis, and possibly Ureaplasma urealyticum to detect the presence of urethral infection, a mean of more than four polymorphonuclear leukocytes per field (X 1,000) in a Gram-stained urethral specimen indicates the presence of urethritis.