Issues in the development of client-treatment matching hypotheses.

This article discusses observations and insights that were gained in the course of critiquing the a priori client-treatment matching hypothesis that were developed for Project MATCH. A matching taxonomy is offered that differentiates eight types of ordinal and disordinal interactions and discusses their clinical and substantive implications. Previously reported alcohol treatment studies are used to illustrate many of these. Various issues in matching studies, such as power versus complexity, prediction versus explanation, matches versus mismatches and capitalization versus compensation, are discussed. Selective issues in choosing treatment and client-matching variables as they affect matching hypotheses are also addressed. The importance of developing a putative "causal chain" for testing the theory underlying matching processes is explicated and a procedure for testing this causal chain is outlined.