["Journal of Computer Assisted Learning, Volume 42, Issue 3, June 2026. ", "\nABSTRACT\n\nBackground\nAlthough embodied pedagogical agents are often viewed as an effective strategy for improving multimedia learning, it remains unclear whether they outperform simple signalling methods and through which mechanisms—cognitive, social, or both—they facilitate learning.\n\n\nObjectives\nThis study compared the effects of five cue‐based instructional strategies (embodied pedagogical agent, hand gesture cue, colour cue, arrow cue and no cue) on learning perception, attention distribution, and learning performance to examine the unique role of embodied pedagogical agents and the mechanisms underlying their effects.\n\n\nMethods\nCollege students learned about neural transmission through a multimedia lesson, supported by embodied pedagogical agents, hand gesture cues, colour cues, arrow cues, or no cues, while their eye movements were recorded.\n\n\nResults and Conclusions\nAlthough the colour cue group showed the highest attention to relevant learning elements, the embodied pedagogical agent group achieved the best learning outcomes, surpassing all other cue groups. Mediation analyses indicated that embodied pedagogical agents not only directly improved learning performance but also indirectly enhanced it by increasing attention to the materials and reducing perceived cognitive load. In addition, the dissociation between attentional guidance and learning outcomes suggests that mechanisms beyond attention may be involved, potentially including social processes.\n\n\nImplications\nThis study demonstrates the advantage of embodied pedagogical agents over other simple cues and provides a more nuanced understanding of how they influence multimedia learning. These findings extend insights from social agency theory and the signalling principle.\n\n"]