Background and purpose: To check the dose delivered to patients during intraoperative electron beam radiation therapy (IOERT) for early breast cancer and also to define appropriate action levels. Patients and methods: Between December 2000 and June 2001, 54 patients affected by early-stage breast cancer underwent exclusive IOERT to the tumour bed using a Novac7 mobile linac, after quadrantectomy. Electron beams (5, 7, 9 MeV) at high dose per pulse values (0.02–0.09 Gy/pulse) were used. The prescribed single dose was 21 Gy at the depth of 90% isodose (14–22 mm). In 35 cases, in vivo dosimetry was performed. The entrance dose was derived from the surface dose measured with thin and calibrated MD-55-2 radiochromic films, wrapped in sterile envelopes. Films were analysed 24–72 h after the irradiation using a charge-coupled-device imaging system. Field disturbance caused by the film envelope was negligible. Results: The mean deviation between measured and expected doses was 1.8%, with one SD equal to 4.7%. Deviations larger than 7% were found in 23% of cases, never consecutively, not correlated with beam energy or field size and with no evidence of linac daily output variation or serious malfunctioning or human mistake. The estimated overall uncertainty of dose measurement was about 4%. In vivo dosimetry appeared both reliable and feasible. Two action levels, for unexplained observed deviations larger than 7 and 10%, were preliminary defined. Conclusions: Satisfactory agreement between measured and expected doses was found. The implementation of in vivo dosimetry in IOERT is suggested, particularly for patients enrolled in a clinical trial.