In this paper practical problems of sampling in the given situation of the psychiatric hospital are discuseed. T heory and method underlying patient selection have far reaching consequences for the outcome of an investigation. Even if we leave out of consideration the statistical problems, and the interrelated problems of theoretical concepts and goals of an investigation, there remain a lot of technical pitfalls of sampling itself which means: so many risks for the success of the investigation.
On the basis of the difficulties met by sampling schizophrenic patients it is demonstrated that in the situation of the psychiatric ward the principle of "probability sampling" is not applicable, when there is no welt defined population. Clinical diagnosis in psychiatry is based on implicit evidence, for the greater part. Reliability of these diagnoses, as the basis for sampling and further investigation in order to get reliable information, is insufficient.
"Non probability sampling" is a useful tool in trying to find scientific rules. But conclusions are only valid for the patient who are in the sample and they can not be generalized to other patients unless replication with other samples is done.
This means that the criteria chosen as a basis for sampling must be objective, otherwise no replication is possible.
It is stated then, that further stratification on relevant criteria is not only inevitable, bat also preferable in order to get conclusions that are useful.
Ir is stressed that for replication circumstances ander which the investigation is done must also be stated explicitly. Only in this way, by being modest in planning and expectations, further reliable research in the psychiatric hospital is possible by means of sampling techniques. Sampling will always be an inevitable and necessary tool.
These problems are illustrated by examples taken from the research going on in the University Hospital for Psychiatry of the Willem Arntsz Stichting in Utrecht.