Thirty grade 4 subjects were individually tested on a digit span test of working memory capacity (Case and Kurland) and the Reading Span Test (Daneman and Carpenter). The Reading Span Test was administered using sentences at a grade 2, grade 4, and grade 6 reading level. It was predicted that, as the decoding demands of the stimulus sentences in the Reading Span Test were increased, the amount of working memory capacity available for other reading tasks such as comprehension would decrease relative to a subject's working memory span as measured on the digit span test. This hypothesis was confirmed , providing support for the predictions of the LaBerge and Samuels model of learning to read. The results further support the Case, Kurland and Goldberg argument that working memory capacity is a function of the interaction between the amount of central capacity required for processing operations and the amount of working memory capacity left over for storage. However, these conclusions need to be qualified by the observation that sentence reading time and word meaning knowledge were not controlled in this study. (Author/PN)