This investigation examined the structural relations among latent variables of language, decoding, and reading comprehension, invariance of the patterns of predictions, and unique and common variance over two years in grade cohorts 5 to 6, 7 to 8, and 9 to 10. Participants were 321 students in grade 5 in six elementary schools, 299 students in grade 7 in three middle schools, and 137 students in grades 9 in one high school in Florida. The dimensionality of language measures (vocabulary and syntax) and decoding measures (real word and nonword fluency) was examined using confirmatory factor analysis and related to a reading comprehension factor consisting of state and national tests. Structural relations among language, decoding, and reading comprehension were stable across grade cohorts. The interaction of decoding and language in predicting reading comprehension was significant only in the grade 5 cohort, thereby confirming a multiplicative functional form only in that cohort. Multiple group invariance testing using structural equation modeling allowed results across grade cohorts to be compared, revealing the increasing importance of language over decoding to predicting reading comprehension in the middle and high school cohorts. By high school, language and reading comprehension were essentially one dimension. The contribution of shared variance in language comprehension and decoding in jointly predicting reading comprehension was large in all cohorts. Implications for the Simple View of Reading and for education are discussed.