The potential of magnetic resonance (MR) imaging in the detection of vaginal tumors and the assessment of their extent was evaluated retrospectively in 87 patients in whom primary, metastatic, or recurrent vaginal cancer was clinically suspected. MR findings were compared with the results of surgery and/or biopsy. Results of histopathologic study verified a normal vagina in 51 patients, benign cysts in two, primary tumor in four, metastatic cancer in 22, and recurrent tumor in eight. MR imaging demonstrated the two benign cysts, correctly depicting their size and location, and demonstrated and enabled correct staging of the four primary tumors. MR accuracy for demonstrating metastatic cancer was 92%, sensitivity was 95%, and specificity was 90%. Tumor extension to the vagina was reliably excluded (negative predictive value = 97%), but false-positive interpretations occurred (positive predictive value = 84%) primarily because inflammation (n = 2) and congestion (n = 2) could not be distinguished from tumor invasion. An additional role of MR imaging was the ability to differentiate tumor from fibrotic tissue in patients with suspected recurrent vaginal carcinoma. The overall accuracy of MR imaging for recurrent vaginal cancer was 82% (14 of 17).