Mitigation is the modification of language in response to social or cognitive challenges (stressors) in contexts of linguistic interaction (Martinovski, Mao, Gratch, & Marsella 2005). Previous mitigation research has been largely from social perspectives, addressing the word or utterance levels of language. This dissertation presents an empirical study of mitigating language resulting from both a cognitive stressor (degree of uncertainty) and social stressor (degree of imposition) in Spanish discourse, and the impacts of mitigation on interaction and interlocutors. The tripartite approach includes a: (1) quantitative analysis of discourse markers associated with mitigation (speaker-discourse, speaker-listener, and epistemic markers); (2) qualitative discourse analysis, relying on concepts from the Conversation Analysis framework; and (3) qualitative analysis of interlocutors’ perceptions of mitigation, using metalinguistic conversations. The results are discussed considering prior research on mitigation, politeness theories, and Clark’s (2006) model of ‘language use’ to address information types, interlocutor roles, and mutual knowledge. In addition, Caffi and Janney’s (1994) ‘anticipatory schemata’ and Pinker’s (2007) social psychological perspective of indirect language inform the theoretical framework. Results indicate that: (1) Mitigation devices vary depending on contextual factors prompting mitigation, significantly fewer speaker-listener markers are shown as evidence of mitigation, and epistemic markers, which are commonly analyzed mitigation devices, are infrequent overall in these data. These results provide evidence against the assumption that mitigation is associated with increased use of linguistic devices; (2) Two mitigating discourse structures were found, depending on the degree of uncertainty. Within contexts of high-imposition, the Co-reconstruction structure (CRS) is found in contexts with uncertainty and the Non-linear structure (NLS) is in contexts with certainty; and (3) The listeners’ metalinguistic comments indicate that the CRS, compared to the NLS, is preferred. Also, the results indicate how interlocutors address cognitive, social, and emotional stressors in interaction. Considering all analyses, a unifying definition of mitigation in discourse is provided. This phenomenon is characterized as the postponement of both confirmed knowledge and negotiation of the interlocutor relationship. This research contributes the first experimental investigation of mitigation as the result of cognitive and social stressors, and also the first systematic analysis of mitigation in Spanish discourse. / text