This study examined the relationship between knowledge of academic vocabulary and reading comprehension in data contributed by 5855 middle school students. Each student completed an academic vocabulary assessment, a standardized reading comprehension test, and one of four types of novel vocabulary-depth measures. "Multiword expressions" examined students' abilities to complete formulaic phrases. "Topical associates" items required students to identify a target word that was topically related to three others. The "hypernyms task" required students to identify the superordinate for each target word. The "definitions task" asked students to choose the definition of the target word. We modeled the relationship between performance on the reading comprehension task and each of the four types of assessments using a residual factors approach (Bentler & Satorra, 2000) with latent variables. Even though each depth measure tested exactly the same sets of words, we found that these measures had a differential impact on reading comprehension, with the definitions task explaining the largest portion of variance in reading comprehension beyond overall academic vocabulary. The knowledge of multiword expressions and topical associates--but not of hypernyms--also explained unique variance in reading comprehension even when controlling for academic vocabulary knowledge.