The aim of the reported study was to test the hypothesis that text-picture combinations aid learning because pictures reduce the learners’ need to actively generate mental images on their own. This, in turn, should free up cognitive resources which can be used for other cognitive processes associated with learning, resulting in better performance compared to a text-only condition. This hypothesis was confirmed in an experiment based on a 2x2 design with picture presentation prior to text (yes vs. no) and visuo-spatial secondary task (with vs. without) as independent variables: Learners without pictures showed a higher load of the visuo-spatial sketchpad (i.e., higher interference with the secondary task) during text processing than learners with pictures, presumably because the former generated mental images during reading. This interpretation is supported by the finding that pictures were especially helpful for learners with low imagery capacity. The implications of these results are discussed.