Thirty 8–11-year-old children were administered tests of rapid naming (RAN letters and digits) and reading-related skills. Consistent with the hypothesis that RAN predicts reading because it assesses the ability to establish arbitrary mappings between visual symbols and verbal labels, RAN accounted for independent variance in exception word reading when phonological skills were controlled. Response timing analysis of different components of RAN digits and letters revealed that neither average item duration nor average pause duration were unique predictors of reading skill. However, the number of pauses on digit naming predicted unique variance in exception word reading. Moreover, better readers paused more strategically than poorer readers (e.g. more often at the ends of lines). We suggest that rapid automatised naming may in part reflect differences in strategic control that are a result of differences in reading practice and experience.