Purpose: Several cross-sectional studies have found a correlation between paired-associate learning (PAL) and reading (e.g. Litt et al., 2013; Messbauer & de Jong, 2003, 2006). These findings suggest that verbal learning of phonological forms is important for reading. However, results from longitudinal studies have been mixed (e.g. Lervåg & Hulme, 2009; Horbach et al. 2015). The present study investigated the possibility that the mixed results may be a result of a conflation of accuracy and speed. It is possible that PAL is a stronger correlate of reading accuracy than speed (Litt et al., 2013). The present study attempted to disentangle this issue by measuring reading accuracy and speed separately. Method: 134 Danish children were followed from Grade 0 to Grade 5. In Grade 0, the children completed tests of PAL, RAN, letter knowledge, and phonological awareness. In Grade 5, they completed tests of reading comprehension and isolated sight word reading accuracy and speed. Results: PAL predicted unique variance in sight word accuracy, but not speed. Furthermore, PAL was indirectly linked to reading comprehension through sight word accuracy. RAN correlated with both accuracy and speed, and these effects translated into indirect effects on reading comprehension. Phonological awareness and letter knowledge did not contribute uniquely to Grade 5 reading after control for PAL and RAN. Conclusions: PAL and RAN predict separate aspects of sight word reading: accuracy and speed. Both aspects appear relevant for reading comprehension. We discuss the lack of long-term prediction from phonological awareness and letter knowledge.