A review of research on "incomplete text," which refers to a variety of adaptations of cloze procedures, illustrates how construct validity has been considered and will need to be considered in future research on the subject. Much of the early research on incomplete text focused on product--providing quantitative assessments of comprehension--necessitating efforts to establish reliability and concurrent validity of the measures. As researchers became more involved in manipulating incomplete text, they began increasingly to investigate qualitative aspects of text processing and the issue of construct validity became more important than face validity, concurrent validity, and content validity. The trend has been toward considering text and to a lesser extent reader variables involved in processing information during reading. It appears that incomplete-text researchers need to look beyond simple differences in good and poor readers' cloze scores to determine why differences appear--and to understand how strategies are text or reader related. These insights will be easier to obtain when constructs underlying performance and a sampling of performance (through incomplete-text measures) are identified. (RL)