This study investigated the prevalence and stability of latent classes at risk for reading disabilities (RD) in elementary-aged children whose first language is Spanish. To this end, children (N = 489) in Grades 1, 2, and 3 at Wave 1 were administered a battery of reading, vocabulary, and cognitive measures (short-term memory [STM], working memory [WM], rapid naming, random letter and number generation, and teacher ratings of inattention) in both Spanish and English. These same measures were administered 1 and 2 years later. Four important findings occurred. First, 4 latent classes emerged (nonbalanced bilinguals-good readers, balanced bilinguals-good readers, reading disabled, and classroom inattentive children) across all 3 testing waves. Second, probability estimates indicated that 23% of the total sample was at risk, and approximately 10% of the inattentive class transitioned into the RD risk class by Wave 3. Third, phonological process were primary predictive of latent classes at Wave 1, whereas working memory measures emerged as predictive at Waves 2 and 3. Finally, the best model for correctly predicting the odds of being correctly classified as RD at the final testing wave included measures of Spanish phonological processing, Spanish naming speed, English and Spanish oral language, English verbal WM, and visual-spatial WM. The results support the notion that statistically distinct and stable latent classes emerge under the umbrella of reading proficiency, and that children at risk for RD can be separated among a heterogeneous sample of children who are English language learners.