The purpose of this study was to assess the diagnostic value of rapid automatized naming (RAN) in the identification of poor readers in two alphabetic orthographies: English and Greek. Ninety-seven English-speaking Canadian (mean age = 66.70 months) and 70 Greek children (mean age = 67.60 months) were followed from Kindergarten until Grade 3. In Kindergarten and Grade 1, they were assessed on measures of RAN, phonological awareness, and letter knowledge. In Grade 3, they were assessed on measures of reading accuracy (for the English sample) and reading fluency (for the Greek sample). The results of logistic regression and receiver operating curve (ROC) analyses indicated that RAN was a significant predictor of an individual's risk for reading difficulties when assessed in Grade 1 in English and in both Kindergarten and Grade 1 in Greek. Although the logistic models (including RAN as one of the predictors) accurately identified the majority of the children who turned out to be poor readers, they also over-identified as poor readers a large number of children who did not manifest reading difficulties later on. (Contains 4 tables and 2 figures.)