Graham Rawle’s paste-up novel Woman’s World successfully develops two vivid main characters—Roy and Norma. The employment of first person point of view of the character Norma ensures readers the actual existence of this female character in the story world; while with the unfolding of the story, readers finally realize the fact that the two main characters are integrated into a double-dressing male character Roy and Norma is imagery even in the fictional world. Based on Culpeper’s comprehensive cognition model for dramatic characters, a character cognition model for multimodal literary works is developed. In the new model, the presentation of characters is no longer the product of only language but the coordination of verbal mode and non-verbal modes such as image, typography, layout and color. To explore readers’ interpretation and appreciation of the two characters, the role of multimodal metaphor in the coordination of verbal mode and non-verbal modes are specifically studied in terms of the naming of the two characters, the description of their appearances, and the presentation of their speeches and thoughts. It is concluded that both verbal mode and non-verbal modes are utilized in the meaning construction and characterization of the novel, scheme questioning and schema reconstruction strategies are needed to obtain an accurate interpretation of such complex characters, and multimodal stylistic features such as the use of pictures from commercial advertisements and the utilization of peculiar typography, layout and color play a significant facilitative role in the accurate interpretation and appreciation of the two main characters.