In patients with sick-sinus syndrome, single-chamber atrial pacing has been reported, in retrospective studies, to be associated with lower frequencies of atrial fibrillation, thromboembolism, heart failure, and mortality than ventricular pacing; although single-chamber ventricular pacing is most commonly used. We did a prospective randomised trial in 225 consecutive patients (142 women, 83 men; mean age 76 years) with the sick-sinus syndrome, randomised to atrial (n = 110) or ventricular (n = 115) pacing and followed for up to 5 years (mean 40 [SD 18] months). During follow-up, the frequency of atrial fibrillation was higher in the ventricular group, except at the first follow-up at 3 months. Thromboembolic events (stroke or peripheral arterial embolus) occurred in 20 patients in the ventricular group and in 6 patients in the atrial group (p = 0.0083). 25 patients died in the ventricular group compared with 21 in the atrial group (p = 0.74). Heart failure estimated by the New York Heart Association classification and by the daily doses of diuretics did not differ between the two groups. Atrioventricular block occurred in 2 patients in the atrial group. Patients with sick-sinus syndrome should be treated with atrial pacing rather than ventricular pacing because atrial pacing is associated with lower frequencies of atrial fibrillation, thromboembolic complications, and a low risk of atrioventricular block.