My twenty-five years of cloze testing research: So what?

This paper examines one strand of my research (a dozen or so studies on cloze testing) published between 1978 and 2002 in terms of: (a) the results of these studies, (b) the questions each raised that led to further studies, and (c) the answers I now have to the following questions: (1) How do score distributions generally affect cloze results? (2) How do score distributions affect cloze reliability and validity statistics? (3) Does tailoring cloze, using item analysis, lead to a more reliable and valid test? (4) Are cloze tests basically sentential or intersentential? (5) Why does K-R21 consistently underestimate cloze reliability? (6) Are there cloze items that are not contributing at all to test variance? (7) How do cloze tests administered to low and high proficiency groups differ overall? (8) How many items are not functioning well in cloze tests administered to low and high proficiency groups? (9) Are different items functioning well in cloze tests administered to low and high proficiency groups? The discussion summarizes answers to those nine questions, and the conclusion considers how (a) I learned from my mistakes, (b) I learned more than I set out to learn, and (c) the whole is greater than the sum of the parts.