The aim of this study was to evaluate whether regressive changes after neoadjuvant chemotherapy for breast cancer affect the accuracy of preoperative MRI measurements of tumor size. Thirty-one patients with breast cancer underwent MRI before and after neoadjuvant treatment. Besides pre- and post-contrast T1-weighted MRI, dynamic MRI with high temporal resolution (turbo-FLASH) was performed. Contrast enhancement in dynamic MRI was quantified using a pharmacokinetic two-compartment model, where two parameters, amplitude and k(ep), were calculated and color coded on transversal parameter maps. Considering the conventional MR images, tumor diameters were measured on the color maps and compared with histological tumor size. Histological regression was scored on a five-point scale regarding cytopathic effects, reactive changes, and tumor cell reduction. The correlation between tumor sizes measured by MRI and histopathology was 0.83 ( p<0.0007) in 12 tumors without regressive changes (score 0), and 0.48 ( p<0.051) in 17 tumors with regressive changes scored 1 or 2, without any tendency for systematic over- or underestimation. In two cases without residual tumor (score 4), MRI likewise showed no signs of persistent tumor. The decrease of the contrast enhancement parameters was significantly more marked in tumors with signs of histological regression than in those without. Whenever MRI is used to judge the response of breast cancer to chemotherapy, the reader must be aware that therapy-induced changes may cause significant over- or underestimation of tumor size. We saw a high precision only when there was either no response - according to histological criteria - or when the tumor had regressed completely.