The primary aim of the current study was to identify the strongest independent predictors of reading comprehension using word reading, language and memory variables in a normal sample of 180 children in grades 3-5, with a range of word reading skills. It was hypothesized that orthographic processing, receptive vocabulary and verbal working memory would all make independent contributions to reading comprehension. The contributions of reading speed, receptive grammatical skills, exposure to print, visuospatial working memory and verbal learning and retrieval (a measure of longer-term retention) were also investigated. Working memory tasks that required the processing and storage of numerical and spatial material were used. One of the numerical working memory tasks was based on the number span task developed by Yuill, Oakhill, and Parkin "British Journal of Psychology," 1989, 80, 351-361. A visuospatial equivalent of that task was developed from the forward Corsi block task [Corsi, "Abstracts International," 1973, 34, 891]. The results revealed that, after controlling for age and general intellectual ability, the word reading and the language variables had a much stronger relation with reading comprehension than the memory variables. The strongest independent predictor of reading comprehension was orthographic processing since it captured variance in both word reading, language skills and verbal working memory. The forward Corsi task and performance on a measure of verbal learning and retrieval each made small independent contributions to reading comprehension but the contribution of verbal working memory was not significant. It was concluded that tasks measuring the interplay between short-term and long-term memory, in which new information is combined with information already stored in long-term memory, may better predict reading comprehension measured with the text available than working memory tasks which only have a short-term memory component.