This paper reviews some problems in the methodology of clinical psychosomatic research. Concentration is on symptomatology rather than on aetiology of disease; and on diagnosis and treatment of known psychopathology, particularly mood disturbance, rather than on underlying speculative mechanisms. Reliability, sensitivity and validity of measurements are considered to be the keys to advance knowledge in this field. Samples of patients studied must be representative of a disease, and not of other factors which contribute bias to the observations. Account must be taken of the wide variability found in most measures required in psychosomatic patients, and hence of their lack of specificity to an individual patient.