We compared dynamic contrast-enhanced MRI (DCE-MRI) and sonography (US) for monitoring tumour size in 21 patients with breast cancer undergoing primary chemotherapy (PCT) followed by surgery. The correlation between DCE-MRI and US measurements of tumour size, defined as the product of the two major diameters, was 0.555 (P=0.009), 0.782 (P<0.001), and 0.793 (P<0.001) at baseline, and after two and four cycles of PCT, respectively. The median tumour size was significantly larger when measured by DCE-MRI than by US at baseline (1472 vs 900 mm(2), P<0.001) and after two cycles of PCT (600 vs 400 mm(2), P=0.009). After PCT, the median tumour size measured by the two techniques was similar (256 vs 289 mm(2) for DCE-MRI and US, respectively, P=0.859). The correlation with the histopathological major tumour diameter was 0.824 (P<0.001) and 0.705 (P<0.001) for post-treatment DCE-MRI and US, respectively. Measurements of the final major tumour diameter by DCE-MRI tended to be more precise, including cases achieving a pathological complete response. Randomized trials are warranted to establish the clinical impact of the initial discrepancy in tumour size estimates between DCE-MRI and US, and the trend towards a better definition of the final tumour size provided by DCE-MRI in this clinical setting.