We analyzed baseline and 12-month follow-up interview data from 98 women who had volunteered to use an experimental intracervical device (ICD) and from 155 women who had been randomly assigned to two control groups, the levonorgestrel-releasing intrauterine device (LNG-IUD; N = 86) and the copper Nova-T IUD (N = 69). All participants were clinic patients in Helsinki, Finland. Initial analysis of 12-month discontinuation data indicated that a significantly higher percent of ICD users (22.4) discontinued their device than did either LNG-IUD (7.0) or Nova-T (8.7) users. However, we were no longer able to detect significant differences in discontinuation after controlling for baseline variables that assessed predisposition to be dissatisfied with contraception in general. Adjusted probabilities of discontinuing the ICD, LNG-IUD, and Nova-T were 11.8%, 6.2% and 7.9%, respectively. These data indicate that the ICD is likely to be acceptable to Helsinki clinic patients; moreover, they suggest a definite place for hormonal intrauterine devices in the contraceptive armamentarium. Most importantly, the methodology used here can be generalized to acceptability studies of other contraceptive devices and drugs undergoing Phase I and early Phase II clinical trials (in situations where randomization may not be feasible) in order to identify and control for the bias introduced by nonrandom assignment procedures.