Despite historical emphasis on "specific" learning disabilities (SLDs), academic skills are strongly correlated across the curriculum. Thus, one can ask how specific SLDs truly are. To answer this question, we used bifactor models to identify variance shared across academic domains (academic "g"), as well as variance unique to reading, mathematics, and writing. Participants were 686 children ages 8 to 16. Although the sample was overselected for learning disabilities, we intentionally included children across the full range of individual differences in this study in response to growing recognition that a dimensional, quantitative view of SLD is more accurate than a categorical view. Confirmatory factor analysis identified five academic domains (basic reading, reading comprehension, basic math, math problem-solving, and written expression); spelling clustered with basic reading and not writing. In the bifactor model, all measures loaded significantly on academic "g." Basic reading and mathematics maintained variance distinct from academic "g," consistent with the notion of SLDs in these domains. Writing did not maintain specific variance apart from academic "g," and evidence for reading comprehension-specific variance was mixed. Academic "g" was strongly correlated with cognitive "g" (r = 0.72) but not identical to it. Implications for SLD diagnosis are discussed.