Abstract: The epistemological status of the visual grammar established by Kress & van Leeuwen\1\ has long been challenged, especially the meaning of spatial relations, which is the primary element of visual grammar.This thesis argues that the meaning of visual space comes from crossdomain mapping based on human experience.Therefore, spatial metaphor, which does the mapping from spatial elements to semiotic meaning, plays a significant role in establishing the epistemological status of visual grammar.In visual discourse, spatial metaphor constructs social reality and interpersonal relations on the one hand, and organizes different elements into a coherent whole on the other.Meanwhile, the choice and configuration of spatial elements is not arbitrary, but motivated by discourse strategies.The authors realize their specific discourse purposes through the choice, transformation and configuration of multimodal semiotic resources.