This paper identifies typical sampling problems, including improper application of the Central Limit Theorem, that are associated with the probability-based sampling procedures currently used in clinical psychology research. It then presents two alternative research designs, the theory validation model and the extended case study model, which avoid some or all of these sampling problems. Both these models are based on purposive sampling, a type of non-probability sampling, which selects subjects according to a psychological theory and relates the results back to that theory. The procedural steps for each model are listed and the key advantages in assuring validity are outlined. These advantages are: (1) replication is built into each design through multiple experiments or cases; (2) data are interpreted in terms of a theory and generalized to a theory, rather than to a population; (3) data are gathered in naturalistic, clinical settings; and (4) small sample research is likely to have smaller or less bias using purposive sampling than using probability-based sampling. (JGL)