Many models have hypothesized that multimedia comprehension requires the concurrent processing of verbal and visuospatial information by limited information processing systems. However, in spite of the emphasis devoted to the concurrent processing of verbal and visuospatial information, little research has so far investigated the specific role played by verbal and visuospatial abilities in multimedia comprehension. The present paper aims to study the specific involvement of verbal and visuospatial working memory in multimedia learning. Ninety-two students (39 boys and 53 girls) from a middle school in a small city in the northeast of Italy, were asked to learn new information on the physical and social geography of Germany from a hypermedium. Participants were also required to perform a reading comprehension test, two short-term memory tasks, and two working memory tasks, which assessed either verbal or visuospatial memory. The results support the hypothesis that both verbal and visuospatial working memory sub-components play a role in hypermedium processing, but with specific and distinct functions: the performance in the verbal working memory task was able to predict the semantic knowledge children can derive from hypermedia learning. In contrast, visuospatial working memory seemed to have lower connections to the semantic knowledge children derived from a hypermedia document, whereas its role emerged in the ability to construct a representation of the document structure.