The risk of breast cancer following radiotherapy for Hodgkin's lymphoma appears to be dose related. In this study we compared breast dose in an anthropomorphic phantom for conventional 'mantle'; upper mediastinal/bilateral neck (minimantle) and unilateral neck fields, and evaluated the accuracy of computer planned dose estimates for out-of-field doses. For each field, computer-planned breast dose (CPD) estimates were compared with thermoluminescence dosimetry measurements in five locations within 'breast tissue'. CPD were also compared with ion chamber measurements in a slab phantom. Measured dose and CPD were within 20% of each other up to approximately 10 cm from the field edge. Beyond 10 cm, the CPD underestimated dose by a factor of 2 or more. The mini-mantle reduced the breast dose by a factor of approximately 10 compared with the mantle treatment. Treating the neck field lowered the breast dose by a further 50% or more. Modern involved-field radiotherapy for lymphoma substantially reduces breast dose compared with mantle fields. Computer dosimetery underestimated dose at larger distances from the field. This needs to be considered if computer dosimetery is used to estimate breast dose and, by extrapolation, breast cancer risk.