In interpreting diagnostic information the differences between morbidity patterns in general practice and those in hospitals must be taken into consideration. This article demonstrates the importance of prevalence for predictive values of complaints, symptoms and test results. When the general practitioner refers patients to the specialist, these values may change, and it is also possible that associations between symptoms and diagnoses are distorted by selection bias. Moreover, attention must be paid to the differences in clinical stages encountered in general practice and specialist practice. It is concluded that a large part of the ;diagnostic field' of general practice has still to be discovered and developed.