Purpose: This was an investigation of the dimensionality of oral and written language to test the hypothesis that a two-factor model with sound/word and sentence/discourse language levels would best fit language and literacy data for a population-based sample in the school-age years. Method: A stratified secondary data set of 1,500 participants was drawn randomly from a larger nationally representative U.S. data set (N = 1,853) gathered during standardization of the Test of Integrated Language and Literacy Skills. A sample of 254 students with prior diagnoses of language and literacy disorders (LLD) was drawn from the full data set. Confirmatory factor analysis was used to compare the hypothesized two-factor model with other theoretically possible models. Results: The hypothesized two-factor language-levels model had an acceptable-to-good fit to the full data set, as did the three-factor model, with verbal memory added. High interfactor correlation between verbal memory and sentence/ discourse constructs, as well as a preference for parsimony, led to the acceptance of the two-factor model as best. This language-levels model had a good fit to the data at ages 8-11 years, and an excellent fit at ages 12-18 years, but only a poor fit for ages 6-7 years (yet still better than other two-factor or unitary models). It had a reasonable fit for students with LLD, although the three-factor model fit their data slightly better. Conclusions: Oral and written language abilities during the school-age years are best explained by a two-factor model with sound/word and sentence/ discourse language levels and memory as a contributing factor. Implications for identifying and treating language and literacy disorders as multidimensional rather than categorical are discussed.