A single 10 mg dose of mifepristone, and two 0.75 mg doses of levonorgestrel 12 h apart, are effective for emergency contraception. Because no studies had compared the efficacies of both compounds, or investigated a single dose of 1.5 mg levonorgestrel, we undertook this three-arm trial. We did a randomised, double-blind trial in 15 family-planning clinics in 10 countries. We randomly assigned 4136 healthy women with regular menstrual cycles, who requested emergency contraception within 120 h of one unprotected coitus, to one of three regimens: 10 mg single-dose mifepristone; 1.5 mg single-dose levonorgestrel; or two doses of 0.75 mg levonorgestrel given 12 h apart. The primary outcome was unintended pregnancy; other outcomes were side-effects and timing of next menstruation. Analysis was by intention to treat, but we did exclude some patients from the final analyses. Of 4071 women with known outcome, pregnancy rates were 1.5% (21/1359) in those given mifepristone, 1.5% (20/1356) in those assigned single-dose levonorgestrel, and 1.8% (24/1356) in women assigned two-dose levonorgestrel. These proportions did not differ significantly (p=0.83). The relative risk of pregnancy for single-dose levonorgestrel compared with two-dose levonorgestrel was 0.83 (95% CI 0.46-1.50), and that for levonorgestrel (the two regimens combined) compared with mifepristone, 1.05 (0.63-1.76). Side-effects were mild and did not differ greatly between groups, and most women menstruated within 2 days of the expected date. Women who took levonorgestrel had earlier menses than did those who took mifepristone. The three regimens studied are very efficacious for emergency contraception and prevent a high proportion of pregnancies if taken within 5 days of unprotected coitus. Mifepristone and levonorgestrel do not differ in efficacy. A 1.5 mg single levonorgestrel dose can substitute two 0.75 mg doses 12 h apart.