The present eyetracking study examined the influence of emotions on learning with multimedia. Based on a 2 × 2 experimental design, participants received experimentally induced emotions (positive vs. neutral) and then learned with a multimedia instructional material, which was varied in its design (with vs. without anthropomorphisms) to induce positive emotions and facilitate learning. Learners who were in a positive emotional state before learning had better learning outcomes in comprehension and transfer tests and showed longer fixation durations on the text information of the learning environment. Although anthropomorphisms in the learning environment did not induce positive emotions, the eyetracking data revealed that learners' attention was captured by this design element. Hence, learners in a positive emotional state who learned with the learning environment that included anthropomorphisms showed the highest learning outcome and longest fixation on the relevant information of the multimedia instruction. Results indicate an attention arousing effect of expressive anthropomorphisms and the relevance of emotional states before learning.