To test the assumption that questions measuring literal comprehension and those measuring inferential comprehension are equally valid indices for both oral and silent reading tests at all skill levels, questions from the Analytic Reading Inventory were classified as either literal or inferential. Subjects, 94 children in grades two to five, read grade-appropriate test passages either orally or silently and answered the comprehension questions. Although showing no direct effects from test format (oral or silent reading) or kinds of questions asked, results showed several interaction effects at different levels of competence. These effects included the following: (1) poor readers comprehended better during oral reading than during silent reading; (2) poor readers tested better on inferential questions than on literal ones when inferential questions included items measuring main idea, cause-effect relationships, and use of implied meanings; (3) the average reader comprehended better during silent reading than during oral reading and comprehended all questions equally well, (4) the good reader was generally strong in both oral and silent reading on various measures of comprehension, and (5) the single best indicator of competency was literal comprehension. (MM)