This paper reviews past work on the development of techniques for the validation of computer programs for predicting the thermal performance of buildings. The major achievements of national and international projects are described and linked to ongoing work that is further developing and applying such techniques. The major conceptual issues are described, together with outstanding problems, both pragmatic and philosophical. Significant progress has been achieved since the early pioneering work, with suites of testing tools currently being considered for inclusion in both USA and European standards. However, validation can never be exhaustive and absolute. The consequences of this for the practical use of computerized design tools are discussed.