An effort was made to achieve different levels of plasma levonorgestrel (13-ethyl-17 alpha-ethynyl-17-hydroxy-gon-4-en-3-one) (Ng) concentrations by the intra-uterine admininistration of levonorgestrel-releasing intra-uterine devices (IUD). Three different models were designed to release different doses of Ng daily. Sixteen women volunteered for the study and had a Ng IUD inserted post-menstrually. The effect of the IUDs on the ovulatory pattern was studied by measuring plasma concentration of oestradiol and progesterone by radioimmunoassay. Ng plasma concentrations were also measured by radioimmunoassay. In vitro experiments to measure the release of Ng from each type of device were carried out. Although some overlapping of mean plasma levels of Ng occurred during use of the three models of the IUDs, suppression of ovulation was clearly dose-dependent. Wide variations in plasma levels of Ng were seen between individuals but quite uniform plasma levels were seen in each subject for up to one year. IUDs where in vitro release rates were equivalent to each other also resulted in wide individual variations in plasma concentrations of Ng.