According to the embodiment principle, students learn better when they engage in task-relevant sensorimotor experiences during learning, such as gesturing or manipulating objects. Students may benefit from enacting movements themselves and/or observing them performed by others. Embodied instruction supports learning by offloading thinking to the physical world (i.e., reduced cognitive load) and by drawing analogies between abstract concepts and meaningful actions (i.e., increased generative processing). Prior research has identified a wide range of promising embodiment methods – using gestures to represent math concepts or to trace important elements of diagrams; manipulating concrete (or virtual) objects to understand stories, math concepts, molecular structures, or physics principles; and designing visualizations that present lessons from the learner’s perspective.

A pedagogical agent (PA) is a computerized humanlike onscreen image whose purpose is to guide the learner’s attention and facilitate learning. In this chapter, the main findings from research on pedagogical agents are reviewed and a meta-analysis is used to reveal the specific role of PAs on learning outcomes, yielding an effect size of g = .45 on transfer performance and g = .23 on retention performance. We also examine explanations for why PAs could have a positive effect on learning performance based on social agency theory, the embodiment principle, and the signaling principle. In addition, this chapter summarizes the main moderating variables for the effect of PAs, including characteristics of the learner such as the learner's prior knowledge, characteristics of the learning environment such as the subject matter, and characteristics of the PA such as voice and gesture.

Social cues can create a sense of partnership between learners and the instructor, which then motivates learners to engage in generative processing. The personalization principle is that people learn better when multimedia lessons are presented in a conversational or polite style, rather than a formal or direct style. The voice principle is that people learn better when multimedia lessons are presented in a human voice rather than computer-generated voice. The image principle is that people do not necessarily learn better when the image of the instructor or a virtual agent is visible on the screen. Finally, the embodiment principle is that people learn better when onscreen pedagogical agents engage in human-like behaviors, such as gesture, eye-contact, or facial expressions.