Objective: To compare the incidence of cough with the angiotensin II antagonist losartan, the angiotensin converting enzyme inhibitor lisinopril, and hydrochlorothiazide in hypertensive patients with previous angiotensin converting enzyme inhibitor cough. Design: Double-blind random stratified parallel-group comparison of losartan 50 mg, lisinopril 20 mg and hydrochlorothiazide 25 mg, each given once daily for a maximum of 8 weeks. Cough detected by self-administered questionnaire (the primary end-point) and cough frequency by visual analogue scale (a secondary end-point). Setting: Hypertension clinics in 20 centres in 11 countries. Patients: 135 hypertensive patients, all non-smokers, with angiotensin converting enzyme inhibitor cough confirmed by lisinopril rechallenge then placebo dechallenge. Results: Incidence of cough with losartan (29%) lower than that for lisinopril (72%; P < 0.01) and similar to that for hydrochlorothiazide (34%). Cough frequency by visual analogue scale lower for losartan than lisinopril (P < 0.01) and similar to that for hydrochlorothiazide. Conclusions: The specific selective angiotensin II receptor (AT1) antagonist losartan does not cause cough in patients with previous angiotensin converting enzyme inhibitor cough. Angiotensin converting enzyme inhibitor cough is likely to be related to kininase II inhibiting action.