This study analyzed the word-, sentence-, and passage-level demands of high-interest, low-level books in a manner consistent with an interactive model of reading comprehension. Cases consisted of three randomly selected passages from sixty different books. Cases were analyzed across five variables: high-frequency words, decodable words, sentences, T-units, and coherence. A cluster analysis was performed on the 180 cases. Resulting clusters were compared to hypothetical cluster profiles that were created a priori. The principal finding was that these books vary in their profiles of characteristics across the five variables in relation to different reading instructional approaches. Two clusters were found to be written in accordance with readability formulas, and three other clusters were found to support differing models of reading instruction. Implications for book selection, materials development, and instruction are discussed.