The 3-dimensional (3-D) dose distribution as calculated in clinical practice for tangential breast treatment was verified by means of in vivo dosimetry. Clinical practice in our institution implies the use of 8 MV X-ray beams, a 2-D treatment planning system, collimator rotation and a limited set of patient data for dose calculations. By positioning diodes at the central beam axes as well as in the periphery of the breast the magnitude of the dose values at the isocentre and in points situated in the high-dose regions behind the lung could be assessed. The position of the diodes was verified by means of an on-line portal imaging device. The reproducibility of these in vivo dose measurements was better than 2% (1 SD). Our study showed that on the average the dose delivery at the isocentre is 2% less and at the points behind the lung, 5.7% higher with respect to the calculated dose values. Detailed analysis of these in vivo dosimetry results, based on dose measurements performed with a breast shaped phantom, yielded the magnitudes of the errors in the predicted dose due to several limitations in the dose calculation algorithms and dose calculation procedure. These limitations are each introducing an error of several percent but are compensating each other for the dose calculation at the isocentre. We concluded that the dose distribution in a patient for our treatment technique and dose calculation procedure can be predicted with a 2-D treatment planning system in an acceptable way. A more accurate prediction of the dose distribution can be performed but requires an estimation of the lack of scatter due to missing tissue, the change in the dose distribution due to oblique incident beams and the incorporation of the actual output of the treatment machine in the assessment of the number of monitor units.