Face processing and recognition has been one of the most productive research areas in cognitive science over the past four decades, and in most studies images of real faces are the focus of inquiry. Owing to the proliferation of technology in social media in recent years, we have witnessed a significant surge of using line-drawn faces and expressions along with their real-face counterparts for purpose of communication. Here in two experiments we examined how line-drawn faces may differ from real faces in terms of identity and emotional expression processing. In Experiment 1, we used the part-whole task and showed that, compared to real faces, line-drawn faces were processed in a more part-based manner similar to non-face objects (i.e., houses). In Experiment 2, we tracked participants' eye movements while they performed a delayed matching-to-sample task, in terms of expressed emotion, where images of either real or line-drawn faces were used as the sample. In addition, we also examined the role a verbal label may play in identifying the facial expression that matched the description. We did this to test the idea whether facial expressions of line-drawn face were in general more symbolically coded than real faces such that a verbal label would be more effective in retrieving those expressed by line-drawn faces. The results indicated that while line-drawn faces differed from real faces in terms of identity processing, they may be quite similar in terms of expression processing. Furthermore, compared to real faces, providing a verbal label failed to offer any additional help locating the matched expression from line-drawn faces, after controlling for the potential speed-accuracy tradeoff with inverse efficiency scores. This might explain why it has become a common practice to exaggerate portrayed expression in line-drawn faces: To overcome the inherently vague signals of emotional expression.