Lithium fluoride (TLD-700) dosimeters were used to measure exit surface absorbed doses in external beam radiotherapy using an automated TLD reader. Delivered tumour absorbed doses were derived from these measurements for head and neck, pelvis and breast treatments. For the head and neck treatments (first fraction only), the mean percentage difference between prescribed and delivered tumour absorbed doses was -0.15 +/- 3.0% (+/- 1 SD), for the pelvic treatments -0.83 +/- 2.8% and for the breast treatments +0.26 +/- 2.9%. The spread of results is approximately +/- 3% (+/- 1 SD). This is comparable with the estimated uncertainty in a single TLD absorbed dose measurement in phantom (+/- 2%; +/- 1 SD). Thus, ICRU recommended tolerances for absorbed dose delivery of +/- 5% may not be unequivocally detectable using this method. An action level of +/- 10% is suggested, allowing investigation of possible gross errors in treatment delivery at an early stage, before the course of treatment has progressed to a point at which absorbed dose compensation is impossible.