Starting with the premise that conventional metaphor (e.g., ’time is money’) is inadequate to reveal the nature of metaphor, the present study attempts to explore the cognitive process of novel metaphor (e.g., ’metaphor is a solar eclipse.’) understanding. With Sperber & Wilson’s relevance theory (1986/1995) as the general conceptual framework, we maintain that novel metaphor understanding is a process of seeking relevance, involving three sub-tasks: identification of explicature (metaphoricity), construction of contextual assumptions and recovery of implicatures, which are mainly based on non-demonstrative inferences. To testify our assumption, we conducted a two-stage experiment, based on the data collected from newspapers, magazines and academic articles.This thesis starts with one of the key issues of metaphor understanding: metaphoricity identification. Unlike the prevalent notion designed to distinguish literal utterances from metaphorical ones, ’metaphoricity’ in this thesis is defined as the degree to which an utterance is metaphorical, and the metaphoricity of an utterance is assumed to be manifested in three dominant dimensions: (i) incongruity between Topic and Vehicle; (ii) freshness of a particular Vehicle,given the particular Topic domain; and (iii) associativeness in Vehicle terms. In light of the above criteria, a questionnaire of metaphoricity judgement is conducted, whose results validate the tripartite of metaphoricity identification.The second part of the thesis explores the cognitive process of novel metaphor understanding. As an integral part of metaphor understanding, metaphor-related knowledge is argued to be represented as schemas and/or images, which might account partly for the efficiency and accuracy of metaphor understanding. For the firstsubtask of the process - explicature identification, we argued that novel metaphors are acts of belief, contributing to explicatures by loosening their lexical concepts. For the second subtask, the construction of contextual assumptions is constrained by several factors, such as (i) the activation of mutual manifest assumptions; (ii) the balance between cognitive adequacy and processing efficiency and (iii) the capability of working memory and encyclopedic memory of the hearer. For the third subtask -implicature recovery, the hearer is supposed to construct ’emergent structures’ by the integration of Topic and Vehicle as two mental spaces, in addition to making bridging inferences to obtain the intended interpretations of novel metaphors. On the basis of the above procedures, we attentatively develop a model of metaphor understanding.Related to the cognitive process of novel metaphor understanding is the controversial issue of ’metaphor and effort’, which is tackled by exploring the relationship between metaphoricity of an utterance and its processing difficulty in metaphor understanding. We found through experiments that the more metaphorical an utterance, and the more diversified contextual assumption available to the hearer, the more likely for him to be diverted from the intended interpretations, and thus, more difficult for him to derive the correct understanding. As a result, we conclude that novel metaphor understanding is effort-demanding, and that the experimental results are compatible with the prediction made by relevance theory about ’metaphor and effort’.Though we have made a preliminary study of novel metaphor understanding, more substantial research on novel (rather than conventional) metaphor is believed to shed greater light on some really fundamental problems in metaphor studies.