Embodied cognition is essentially the brain’s processing of multimodal information, and thus heterogeneous modalities like images are same as language, which can be directly involved in the cognitive representation of metaphors. Reexamining the process of metaphor representation, we find that the embodied representation of multimodal metaphor involves three different dimensions: formal expression, concept and cognitive context. Therefore, the role of “body” in cognition cannot be simply generalized: “sensory” experience can produce perceptual imitation effects at the level of formal expression’s construal operation, which affects the direction of multimodal metaphors; by means of diverse formal expressions, “body structure” becomes a different meaning through the cognitive operation of “taking images and assigning shapes”. The “scene” forms a cognitive context in the interaction with the body, and then gives the symbol a new connotation.