We retrospectively studied the staging accuracy of magnetic resonance (MR) imaging after neoadjuvant hormonal therapy (NAH) for 21 localized prostate cancers. MR imaging was performed using a 1.5-Tesla magnetic resonance system with a pelvic phased array coil. T2-weighted MR images were obtained on axial and coronal planes, and T1-weighted MR images using the dynamic technique with Gd-DTPA bolus enhancement were obtained in axial planes for each patient. On T2-weighted imaging, the signal intensity of the normal tissue in the peripheral zone became lower after NAH. Therefore, it was more difficult to detect residual malignant lesions in many cases than before NAH. The accuracy of T staging for prostate cancer after NAH in MRI was 71%. The accuracy, sensitivity, and specificity of the extracapsular invasion was 76%, 0% and 94%, respectively, and those of the seminal vesicular invasion 85%, 0% and 100%, respectively. While 2 of the 4 patients judged as downstaged cases in MRI showed corresponding pathological findings, 5 of the 21 cases (23.8%) were underdiagnosed. Local staging with only MRI for prostate cancer after NAH seems to have limits in applicability.