The authors investigated whether practice in reading fluency had a causal influence on the relationship between working memory (WM) and text comprehension for 155 students in Grades 2 and 4 who were poor or average readers. Dysfluent readers were randomly assigned to repeated reading or continuous reading practice conditions and compared with untreated dysfluent and fluent readers on posttest measures of fluency, word identification, vocabulary, and reading comprehension. Three main findings emerged: (a) The influence of WM on text comprehension was not related to fluency training, (b) dysfluent readers in the continuous-reading condition had higher posttest scores than dysfluent readers in the other conditions on measures of text comprehension but not on vocabulary, and (c) individual differences in WM better predicted posttest comprehension performance than word-attack skills. In general, the results suggested that although continuous reading increased comprehension, fluency practice did not compensate for WM demands. The results were interpreted within a model that viewed reading comprehension processes as competing for a limited supply of WM resources that operate independent of fluency. (Contains 9 tables and 2 notes.)