本文是针对美国总统竞选辩论语篇中的反讽的一项认知语用研究。作者采用Yus 提出的研究反讽识别和理解的认知语用基本框架（语境激活和最大可及性标准），并对其理论论述中界限不清、冗余的部分及忽视的部分进行了优化和修改：将语境激活模式中的七种语境信息中重合的部分合二为一，并提出Yus模式中的数量这个概念只能放在同一个反讽话语中理解，而不能同时放在多个反讽话语中；将最大可及性标准模式中的三个典型案例发展为四个。文章主要着眼于以下方面：一、根据Yus的基本认知框架研究反讽如何被识别和理解；二、美国总统辩论语篇中反讽的特点。本文既包括定性研究也包括定量研究。对美国总统辩论中的语料（总统辩论，副总统辩论，民主党辩论，共和党辩论。共73个辩论事件，57个反讽材料）进行定性和定量分析。研究发现美国总统辩论中反讽的使用并不少见。与其他语境信息相比，听者识别反讽时，绝大多数语言暗示和非语言行为经常作为辅助语境信息，这与Yus的先前预测是一致的；言者资料，百科知识，共有知识，语言暗示和非语言行为这几种语境信息被激活的比例较大。其中言者资料和百科知识这两种语境信息经常作为主导语境信息，而共有知识作为主导信息的比例较小。物理环境和先前语言这两种语境信息在该语篇中被激活的次数偏少。四种典型案例中，第一种比例最大，即听者大多数情况下可以快速识别反讽话语，这与该语篇的性质和特点有关。总统辩论非常严肃正式，一言一句都可能关乎选民对候选人的判断，所以说者为了避免听者的误解，会尽量为听话人提供足够的语境作支持来帮助听话人顺利获得其真正传达的意思。第二种典型案例即听者识别反讽较慢的情况较少，听者不能快速识别反讽很可能是说者为达到某种突出或增强反讽意味的语境效果故意而为之。第三典型案例即听者没能识别反讽和第四典型案例听者将非反讽话语误解为反讽话语的情况极少。另外，辩论语篇涉及双重听者：辩论对手和听众；这一特殊性使该语篇中的反讽话语呈现出了不同的特点。说者在使用了反讽之后，为确保其反讽意思被正确理解，常在反讽话语之后说明自己的真实意思。总之，对反讽的认知语用分析对语境如何在听者识别理解反讽的过程中起作用有很强的解释力。将修改过的Yus关于反讽的理论用于分析美国总统辩论语篇中的具体反讽话语实例是更合理、可行的，也便于操作。

This thesis investigates the recognition and interpretation of verbal irony in American presidential debates with a cognitive pragmatic approach. Yus’ theory of irony interpretation (contextual activation & criterion of optimal accessibility) is modified and adopted as the basic framework. Within the model of contextual activation, the overlaps among the seven contextual sources are dealt with, and the writer also points out this model can only be used to analyze the same ironic utterance rather than parallel cases. Within the criterion of optimal accessibility, the classification of three prototypical cases is developed into four cases. The study mainly focuses on the following aspects: 1) how the hearer interprets irony within modified Yus’ cognitive model; 2) the features of verbal irony in the discourse of formal political debates.The research employs both qualitative approach and quantitative approach to analyze ironic utterances (57 ironic language materials from 73 debating events including Presidential debates, Vice presidential debates, Democratic presidential debates and Republican presidential debates). It shows that the use of verbal irony is not a rare thing in American presidential debates. Compared to other contextual sources, biographical data, encyclopedic knowledge, mutual knowledge, nonverbal behavior and linguistic cues are frequently activated. Biographical data and encyclopedic knowledge often serve as the leading contextual source. Mutual knowledge often serves as a supportive source. Linguistic cues and nonverbal behavior have a very high frequency in the course of irony interpretation, but normally serve as supportive contextual sources. It is compatible with Yus’ prediction. By contrast, physical setting and previous utterances are far less common to be activated in this particular discourse.Furthermore, among the four prototypical cases, the majority of ironic utterances can be interpreted very fast (the first prototypical case) as the hearers possess adequate contextual sources which help them reach the intended meaning successfully. The ironic utterances comprehended relatively slowly (the second prototypical case) are probably deliberately produced by the speakers to highlight or enhance the ironic effect. The third and fourth prototypical cases are extremely rare in this kind of discourse. The special point about this discourse is that it contains two kinds of hearers—the opponent and the audience. Speakers tend to share as much background information as possible so as to avoid misunderstanding due to the fact that every word they say may impose great impact on the audience’s judgment.In conclusion, a cognitive pragmatic approach well explains how context fulfills its role in the course of irony interpretation. And the modified Yus’ model is appropriate and feasible relative to the original one in terms of the analysis of specific ironic utterances in interactive discourse.