The domain-referenced approach to assessment results in a set of tests that are closely linked to theoretical propositions about aspects of reading comprehension. In this paper, the definition, validation, and use of domains are described, and one strategy for developing domain-referenced measures is proposed. The strategy involves identifying a plausible candidate domain, identifying prototypic items from that domain, specifying relevant variables thought to affect the difficulty of responding correctly to those items, creating sample items, and verifying the domain empirically. An attempt was made to implement this strategy in measuring "ability to find the main point of a passage." In general, the strategy seems to be a useful approach to test development and theory refinement, in that it effectively samples and screens variables affecting performance in the domain. On the basis of this study, suggestions are made for further research on finding the main point, and appropriate next steps in developing domain-referenced measures are proposed. (Author)