Cervical injury is one of the most frequent complications of suction curettage abortion, yet little is known about its risk factors or prevention. In 15 438 suction curettage abortions carried out at less than or equal to 12 weeks' gestation in hospitals in the USA from 1975 to 1978 the incidence of cervical injury requiring suturing was 1.03 per 100 abortions. Among factors potentially within the physician's control, use of laminaria rather than rigid dilators for dilatation had a strong protective effect (relative risk 0.19), whereas performance of the abortion by a resident rather than an attending physician (relative risk 2.0) and use of general rather than local anaesthesia (relative risk 2.6) had detrimental effects on rates of cervical injury. Among other factors, a previous abortion had a protective effect (relative risk 0.46), whereas patient age less than or equal to 17 years had a detrimental effect (relative risk 1.9). Use of laminaria, performance of the abortion by an attending physician, and local anaesthesia together yield a 27-fold protective effect.