Pre-operative magnetic resonance imaging (MRI) was carried out in 50 women scheduled for operative treatment of invasive carcinoma of the cervix. The extent of the primary tumour (stage), its dimensions and the presence of lymph node enlargement were assessed and compared with findings at surgery and/or histopathological examination of the resected uterus. In 45 patients undergoing radical hysterectomy, accuracy of MRI staging of the primary tumour was 84.4%. In the group as a whole, including four patients with inoperable disease, staging accuracy was 84%. Most errors were due to difficulty in identifying early vaginal or parametrial invasion by tumour. There was close correlation between the volume of tumour measured from pre-operative MRI scans and measurements made on the hysterectomy specimen (r = 0.95). MRI had a sensitivity of 75% and a specificity of 88% in predicting metastatic lymphadenopathy, based solely on the criterion of enlargement of any pelvic or para-aortic nodes to 1.5 cm or greater. However, retrospective analysis of the presence or absence of metastases by site in 49 patients undergoing lymphadenectomy or lymph node sampling at laparotomy showed that true sensitivity to be 57.1% and the specificity 96.8%. Differentiation between malignant and reactive lymphadenopathy was not reliably achieved on MRI, and in several patients, metastases were present in normal-sized lymph nodes.