A case-control study of 155 women with pelvic inflammatory disease (PID) and 305 controls with other conditions was carried out. Compared with other methods of contraception, the age-adjusted relative risk estimate for use of any intrauterine device (IUD) within one month of admission was 8.6 (95% confidence limits, 5.3 and 13.8); for past use, the estimate was 1.6 (confidence limits, 0.6 and 4.0). The estimated risk was highest for users of the Dalkon Shield (11 cases and two controls), lowest for users of copper-containing devices (40 cases and 34 controls), and intermediate for users of other IUDs, principally the Lippes Loop (32 cases and 23 controls). Compared with copper-containing IUDs, the relative risk estimate for the Dalkon Shield was 6.5 (confidence limits, 1.5 and 27.5). When the Dalkon Shield and copper-containing IUDs were compared, in turn, with the Lippes Loop, the estimated risk was increased for the former and reduced for the latter, but the differences were not statistically significant. Nine of the 11 cases using the Dalkon Shield (82%) had severe PID (diagnosed at surgery or because of purulent cervical discharge), compared with 12 of 40 copper device users (30%) and nine of 32 Lippes Loop users (28%). The risk of PID did not seem to be related to duration of use of any of the three IUDs. This study confirms that IUD use increases the risk of PID. It also suggests that the Dalkon Shield is more harmful than other devices.