Objective: To compare endovaginal with pelvic phased array coil magnetic resonance imaging (MRI) in detection of Stage I cervical carcinoma by correlating the findings with histopathology. Patients and Methods: Forty consecutive patients with Stage I cervical carcinoma confirmed histologically were studied using an endovaginal coil alone immediately followed by a pelvic phased array coil. T1-W transverse and T2-W FSE sagittal images made with each coil were analyzed independently by two radiologists noting the presence and size of a mass within the cervix and any parametrial extension or involvement of adjacent organs. Tumor volumes were measured using the electronic calliper to compute tumor area on each slice and multiplying by the slice thickness. Thirty patients underwent radical hysterectomy, one a trachylectomy, one simple hysterectomy and four extended cone biopsies. Four patients had radiotherapy to the primary tumor. Following surgery, histopathologic findings were recorded and tumor volumes measured. Results: Tumor volumes ranged from 0–106 cm3(median 1.4 cm3, mean 9 ± 19.4 cm3). Thirty-six patients had correlation of the primary site with the surgical specimen. Agreement between observers was excellent for both endovaginal (k = 0.90) and pelvic phased array (k = 0.96) techniques. Combined sensitivity and specificity for both observers of endovaginal MR imaging for detection of tumor was 96% and 70%, respectively; for pelvic phased array imaging sensitivity was substantially less at 54%. Specificity was higher at 83.7%, probably because small abnormalities were seldom visible. In patients treated surgically, early parametrial involvement in four women on endovaginal MRI was confirmed histologically in two. Pelvic phased array imaging showed early parametrial involvement in four women and was confirmed in one. Conclusion: Endovaginal MRI adds substantially to information from pelvic phased array images in the preoperative assessment of patients with early cervical cancer. J. Magn. Reson. Imaging 2000;12:1020–1026. © 2000 Wiley-Liss, Inc.