We examined the contribution of working memory capacity to the development of children's reading comprehension. We present data from three waves of a longitudinal study when the children were 7 years (Grade 1), 8 years (Grade 2) and 9 years (Grade 3). Two questions were raised: The first question concerned the developmental changes of the relative contribution of working memory in predicting reading comprehension compared to vocabulary and decoding skills. The second question explored to what extent reading comprehension could be predicted by working memory capacity measured at a prior time. At the end of each grade, reading comprehension, nonword reading, vocabulary knowledge and working memory capacity were assessed. To test the first question, the predictive power of working memory capacity was compared to vocabulary and decoding skills by performing concurrent multiple-regression analyses in each grade. The results showed that working memory capacity emerged as a direct predictor of reading comprehension in Grade 3. To address the second question, we performed multiple-regression analyses predicting reading comprehension from working memory, nonword reading, and vocabulary measured at a prior time. In these analyses, the autoregressive effect was taken into account to separately assess the unique contribution of each predictor to the development of later reading comprehension. The results showed that Grade 1 vocabulary and Grade 2 working memory had additional effects on Grade 3 reading comprehension after the autoregressive effect of reading comprehension had been accounted for. These findings support the idea that, as word recognition becomes automated throughout the early grade levels, working memory becomes an important determinant of reading comprehension. There is also evidence that working memory capacity directly influences the development of reading comprehension skills. The direction of the causal flow is discussed.