One of the problems in psychiatric epidemiology is the use of hospital admissions for the study of true morbidity. Clearly this depends upon whether hospitalized cases can be regarded as a representative sample, in other words if admission to hospital varies from one social group to another, from one region to another, etc. Various authors have shown that the duration of hospital stay varies with social status, with marital condition, etc., and such variations in the chances of being discharged suggest similar variations in the probability of being admitted.

As Piercy (1959) has pointed out, it is generally agreed that currently available psychological tests for detecting brain damage (Yates, 1954; Meyer, 1957) or early dementia (Shapiro, 1952; Inglis, 1958) are by and large poorly standardized and poorly validated. If the clinical psychologist attempts such detection he can make very few positive statements if all he does is administer those standardized tests which even approximately meet the commonly accepted standards of validity.

In conducting drug trials nowadays one is aware of at least some of the difficulties that have to be faced and the possible traps lying ahead. It is probably true to say that perfection is unattainable in trials of this nature because of the unpredictability of the human material concerned—both the tested and the observers sharing this common fallibility. It is necessary to attempt so to arrange the experiment that the magnitude of the errors is kept to a minimum and the results remain unobscured.