The specific performance characteristics of eight alternative cloze test formats were examined at the fourth and sixth grade levels. At each grade, 64 subjects were randomly assigned to one of four basic treatments (every-fifth/standard, every-fifth/cued, total random/standard, and total random/cued) and tested. Responses on each of the cloze passages were scored in both a verbatim and synonymic manner. Subjects were administered the Ekwall Reading Inventory to determine instructional reading levels. Analysis indicated significant first-order interaction effects for the deletion pattern by scoring mode and blank condition by scoring mode interactions at both grade levels. In the sixth grade analysis, the deletion pattern by blank condition interaction also attained significance. High positive relationships were observed between verbatim and synonymic scoring across all basic cloze forms. Two analyses tested the maximum item independence hypothesis associated with a context distance of four words between blanks. At fourth grade, the results were equivocal; however, at sixth grade, the evidence favored a rejection of this notion. None of the fourth grade cloze formats was significantly related to silent instructional reading levels. At sixth grade, appreciable increases in both the concurrent validity and reliability of the formats were noted. (Author/PN)