OBJECTIVE The authors evaluate the accuracy of magnetic resonance (MR) imaging with body coil and endorectal surface coil techniques in the staging of cervical carcinomas and compare these results with those obtained with clinical staging (International Federation of Gynecology and Obstetrics [FIGO] classification) and postsurgical histopathology. METHODS Fifteen patients (average age, 48.6 years) with biopsy-proved cervical cancer were included in the study. After clinical staging (FIGO classification), MR imaging with body coil (BCMR) and subsequently with endorectal surface coil (ECMR) was performed. Using a 1.5-Tesla unit, axial and sagittal proton density weighted and spin echo T2-weighted and fast spin echo T2-weighted sequences were obtained with body and endorectal surface coil. During imaging analysis, special attention was paid to the uterine zones, the vaginal fornix, and the parametrial tissue. Ten patients were treated surgically (postsurgical histopathology was considered the gold standard), five patients, all clinically staged IIIb, underwent primary radiation therapy. RESULTS Clinical staging with FIGO classification was accurate in 12 of 15 patients, and understaged in 2 and overstaged in 1 patient. Body coil MR showed accurate staging in 13 of 15 patients. Using this technique, understaging of 2 patients, both with only minimal tumor infiltration depth (< 8 mm), was performed. Endorectal surface coil MR was accurate in 14 of 15 patients missing minimal parametrial tumor infiltration in 1 patient. Histopathology (n = 10) revealed 3 patients with tumor stage T1b, 2 with stage T2a, and 5 patients with stage T2b. CONCLUSIONS Although suggested by only a small number of patients, ECMR appears to be the most accurate modality for staging cervical carcinomas but seems to be unreliable in the detection of minimal tumor infiltration.