The prevalence of phonological and surface dyslexia subtypes among Swedish university students with dyslexia (n = 40) was examined using both the regression method, developed by Castles and Coltheart, and latent profile analysis. When an academic-level control group was used as a reference group in a regression, eight students with phonological dyslexia and 15 students with surface dyslexia were identified. In contrast, 17 students with phonological dyslexia and two students with surface dyslexia were identified when the subtypes were defined by reference to a reading-level control group, indicating a deviant profile among the students with phonological dyslexia and delayed development among those with surface dyslexia. The latent profile analysis was based on five phonological and four orthographic tasks. Seven profiles were obtained, of which none exhibited deficits in orthographic but not in phonological skills or "vice versa." Thus, the analysis further supported the phonological deficit hypothesis of dyslexia. (Contains 4 tables and 5 figures.)