Although phonological awareness (PA) and rapid automatized naming (RAN) are confirmed as early predictors of reading in a large number of orthographies, it is as yet unclear whether the predictive patterns are universal or language specific. This was examined in a longitudinal study across Grades 1 and 2 with 1,120 children acquiring one of five alphabetic orthographies with different degrees of orthographic complexity (English, French, German, Dutch, and Greek). Path analyses revealed that a universal model could not be confirmed. When we specified the best-fitting model separately for each language, RAN was a consistent predictor of reading fluency in all orthographies, whereas the association between PA and reading was complex and mostly interactive. We conclude that RAN taps into a language-universal cognitive mechanism that is involved in reading alphabetic orthographies (independent of complexity), whereas the PA–reading relationship depends on many factors like task characteristics, developmental status, and orthographic complexity.