Computer aided diagnosis is reviewed and an assessment made of the different models used for this purpose. These are discussed under three headings:1. Models based on physicians’ thought processes2. Models based on the physiological relationships manifested in the disease state,3. Statistical models.The models are assessed in terms of their accuracy by comparison with clinical methods. Typical results show that using the computer improves diagnostic accuracy by about 10%.It is concluded that Bayesian models are likely to give the best results since they are well suited to the probabilistic nature of medical data and can be adapted for use in many situations. Further research necessary for the improvement of Bayesian models is discussed.