This work is dedicated to exploring the process of metaphor comprehension. There are a number of cognitive theories addressing this issue, including the conceptual metaphor theory (Lakoff and Johnson, 1980), the salience imbalance theory (Ortony, 1979), the structure mapping theory (Gentner, 1983), the domain-interaction theory (Tourangeau and Sternberg, 1982), the attributive categorization theory (Glucksberg and Keysar, 1990) and the conceptual blending theory (Fauconnier and Turner, 1998, 2002). A critical review of these theories and their supportive empirical studies have revealed that all of them are in a degree applicable to explaining the comprehension of certain metaphors but not capable of working with the processing of others. What are the major factors that drive different mappings to be involved in processing different metaphors and affect the metaphor comprehension? This is the major question to be investigated in this research. Inspired by recent studies, a hypothesis is formulated: the cognitive processing mechanisms in comprehending a metaphor are largely influenced by the addressees’ pre-existing conceptual knowledge as reflected in their estimation of the conventionality and the aptness of the metaphor and the communicative context in which the metaphor arises. To test this hypothesis, a cross-cultural web-based experiment has been carried out to explore how three metaphors are comprehended under various role-play conditions by subjects whose pre-existing conceptual knowledge concerning these metaphors varies from each other. The metaphor The teacher is a candle was estimated by the Chinese subjects as the most conventional and apt teacher metaphor but the German subjects estimated it as a less conventional and less apt metaphor. The metaphor The teacher is a shepherd was estimated by the German subjects as the most conventional and apt, but by the Chinese subjects as less conventional and less apt. The metaphor The teacher is a captain was estimated by both Chinese and German subjects as a less conventional but apt metaphor. Under various role-play conditions (no role play, after the role play scenario with the positive development or after the role play with the negative development), the Chinese and the German participants were first asked to rate their affective impressions of the teacher metaphors on dimensions of the Self-Assessment Manikin (SAM) (Lang, 1980). Then they were required to rate how suitable thirty-three features selected from a pilot study are in describing the teacher metaphors. Altogether 180 complete valid data sets were collected from the participants from two German universities and two Chinese universities. The multivariate analysis of the SAM ratings and the cluster analysis and the network analysis of the feature ratings are summarized as follows: First, a greater consensus in both the SAM ratings and the feature ratings was shared among the subjects who regarded the metaphor as conventional and apt than those who regarded it as unconventional and inapt. Second, significant positive correlations between the topic and the vehicle concept were found in subjects’ SAM and feature ratings of the metaphor that they regarded as conventional and apt. In contrast, their SAM ratings and feature ratings of the metaphor, which they regarded as unconventional and inapt, shows no positive correlation between the topic and the vehicle of the metaphor. Instead, there was a tendency of high rating emergent features. Third, context exerted significant influence on subjects’ SAM and feature ratings. When the metaphor was provided in the role play with the positive development, a greater consensus in rating the SAM and the features appeared among the subjects who took the metaphor as unconventional and inapt. When the metaphor was provided under the condition of the role play with negative development, even the subjects who originally regarded the metaphor as conventional and apt seemed to lose their consensus in rating the SAM and the features. Such results largely confirm the hypothesis of the experiment and reveal that the comprehension of a metaphor is not a static process but rather a dynamic one that can be affected by both the pre-existing conceptual knowledge of the metaphor addressee and the context in which the metaphor arises. Based on the empirical findings, a dynamic theoretical view is formulated to explain the comprehension of metaphor through the integration of Wilson and Sperber’s (2004) relevance theory, Cowan’s (2005) working memory theory and relevant cognitive mapping theories. According to this view, the comprehension of metaphors can be segmented into various situations with the emphasis on the interplay of the addressee’s conceptual knowledge and the communicative context in which the metaphor appears. Depending on people‘s conceptual knowledge preexisting in their long-term memory and the communicative context in which a metaphor appears, the comprehension of the metaphor involves testing the contextual metaphoric assumption that is formulated through the ad-hoc interplay of the topic space, the vehicle space, and the contextual space generated in people‘s working memory. The more conventional and apt the metaphor appears to them, the less complicated are the mappings involved in drawing the contextual metaphoric assumption needed for comprehending the metaphor. This dynamic view of metaphor comprehension can explain why the German subjects had more difficulty than the Chinese subjects in comprehending the metaphor The teacher is a candle in this experiment. It can also provide a good solution to solve the debate among the current metaphor mapping theories and synthesize them in a plausible way. Furthermore, this cognitive metaphor research also suggests follow-up studies should be done in order to develop further the present dynamic view into a well-structured model of metaphor comprehension.