This study examined the impacts of adding emotional design features to a multimedia lesson (color alone, anthropomorphism alone, or color & anthropomorphism together) on college students' affective processes (measured by ratings of experienced emotion during learning), cognitive processes (measured by eye-tracking metrics), and learning outcomes (measured by retention and transfer test scores). One-hundred students were randomly assigned to watch a short multimedia lesson in one of four conditions: no emotional design, colorful emotional design, anthropomorphism emotional design, and colorful and anthropomorphism emotional design. The study results showed that compared to the no emotional design group, the colorful and anthropomorphism emotional design group showed the higher positive emotion rating (d = 0.726), the shortest time to first fixation on an emotional design area (d = -0.877), the longest fixation duration on emotional design areas (d = 0.640), and the best transfer test score (d = 0.679). In contrast, the anthropomorphism emotional design group outperformed the no emotional design group only on rating of positive emotion, and the colorful emotional design group outperformed the no emotional design group only on transfer test score. The results show that two emotional design features are more effective than one in multimedia lessons. A structural equation model indicated that positive emotion (tapping affective processing) and fixation duration (tapping cognitive processing) mediated the pathway from emotional design to learning performance. These results partially support the Cognitive-Affective Model of E-Learning.