: Learning material in multimedia learning environments is frequently made up of text and pictures. Many students, however, have difficulties learning from such material successfully. In order to support students’ learning, we conceptualized a deep level strategy and a surface level strategy for learning from text and pictures. In an experimental study we investigated whether students who exercised the deep level strategy subsequently learned more successfully than students who exercised the surface level strategy. No significant differences in learning success were found between students who previously exercised different strategies. In order to better understand why no differences were observable, we analyzed think aloud protocols taken from the students during learning. On the basis of this analysis we determined the strategies students actually applied during learning. In many cases students applied a different strategy than they had exercised. As expected, students who actually ...