Background: The relationship between rapid automatized naming (RAN) and reading has been examined extensively, but there is still a question as to the reason for the relationship. To reduce involvement of alphanumeric and vocabulary knowledge in RAN, and to possibly reduce phonological demands, researchers have suggested using the RAN sizes that have only two items, that is, naming shapes according to two sizes: 'little' or 'big'. However, despite the RAN sizes advantages, as far as we know, this test is not widely used. The purpose of the current study was to examine whether RAN sizes in kindergarten predict reading speed and accuracy toward the end of first grade among Hebrew speaking children. Methods: A total of 113 Hebrew-speaking children were assessed twice. In preschool, we collected matrix reasoning, phonological awareness and RAN sizes data. In first grade, these data were recollected along with reading data. Results: Kindergarten RAN sizes accounted for 4% of the variance in reading accuracy and 4% of the variance in reading speed. Phonological awareness accounted for 4% of the variance in reading accuracy only. In first grade, RAN sizes explained 19% of the variance in reading speed, with no significant contribution to accuracy. Phonological awareness still explained 4% of the variance in accuracy. Conclusions: The results from the current study support the claim that RAN is a unique longitudinal predictor of early reading, over and above phonological awareness. Moreover, rapid size naming may be used to test RAN with less reliance on alphanumeric and vocabulary knowledge. Finally, RAN sizes may be used to predict reading speed in Hebrew.