Abstract A common assumption in multimedia design is that audio‐visual materials with pictures and spoken narrations lead to better learning outcomes than visual‐only materials with pictures and on‐screen text. The present study questions the generalizability of this modality effect. We explored how modality effects change over time, taking into account study strategies during learning and the modality of the final performance measure. Eighty‐four university students (Mage = 21.4) studied learner‐paced visual‐only or audio‐visual multimedia presentations and answered written and oral retention and transfer questions immediately after learning and after 1 day. There was no performance difference between the audio‐visual and the visual‐only groups immediately after learning, but after 1 day, the visual‐only group had significantly higher scores on three of four outcome measures. This reversed modality effect was independent of test modality, but both groups scored higher on written than on oral questions. While both groups spent on average 33 min studying, the visual‐only group went through the materials at a faster pace and repeated more slides. In sum, results contradict common multimedia design recommendations and instead suggest that learner‐paced presentations should include on‐screen text. Benefits of on‐screen text could be due in part to the successful use of reading strategies.