In consecutive 56 patients with cervical carcinoma, we retrospectively analyzed the accuracy of magnetic resonance imaging (MRI) in regard of determining the depth of stromal invasion and reviewed the pitfalls in MRI interpretation. All patients underwent hysterectomy and were categorized into 3 groups : (1) no stromal invasion, (2) invasion to inner 2/3 of stromal thickness, (3) invasion to outer 1/3 of stromal thickness. In each group, MR image was analyzed for the size of the tumor mass and the ratio of spared stromal thickness to normal stromal full thickness (invasion ratio). Most tumor masses were not detected in the group of no stromal invasion. The degree of the depth of stromal invasion increased with the tumor size, and invasion ratio decreased as stromal invasion deepened. Overall accuracy for stromal invasion on MRI was 69.6% and positive predictive value (PPV) for no stromal Overall accuracy for stromal invsion on MRI was 69.9% and positive predictive value (PPV) for no stromal invasion was relatively lower than those of the other two groups. MRI isa promising modality to evaluate stromal invasion of cervical carcinoma by measuring the tumor size and invasion ratio, if MRI is performed prior to biopsy in axial scan perpendicular to the endocervical long axis.