In this study, we attempt to clarify the link between rapid automatized naming (RAN) and reading. We performed an experiment with dyslexic children in which we compared two versions of the RAN task: a continuous versus a discrete version. in the continuous version, ail items to be named were presented simultaneously on a sheet of paper (classic RAN), whereas in the discrete version items were presented successively on a computer screen. The discrete version reduces the involvement of visuo-attentional processes (e.g., eye movement control). In each version of the task, three types of stimuli were compared. Stimuli were more or less automatized (i.e., digits, letters and objects). The results showed that dyslexic children had comparable deficits in both versions of the task suggesting that visuo-attentional factors cannot by themselves explain the deficits observed in RAN. Moreover, the comparison of the three types of stimuli was not in favor of an hypothesis according to which a general deficit in skill automatization underlies dyslexia. Instead, our results suggest that children with dyslexia have problems accessing phonological representations rapidly and automatically, which is a fundamental skill required for learning to read.