Despite the widespread use of rapid automatized naming (RAN) tasks in reading and mathematics research, little is known about the direction of their relation. Thus, in this longitudinal study, we examined the direction of the relation between RAN and reading/mathematics accuracy and fluency. One-hundred and eighty-three Grade 1 Chinese children from Shanghai (84 girls and 99 boys; M[subscript age] = 85.98 months, SD = 3.76) were followed until Grade 5 and were assessed annually on RAN, reading, and mathematics. The results of standard cross-lagged analyses indicated that the effects were unidirectional (earlier RAN predicted future reading and mathematics fluency), specific (earlier RAN predicted further growth only in fluency outcomes), and consistent (with one exception, earlier RAN predicted every subsequent reading and mathematics fluency). Theoretical implications of these findings are discussed in light of the language in which the study was conducted and the developmental span covered.