Producing affective language: experimental and corpus based approaches

How does the way we feel influence the way we speak? Currently, most of the evidence of the relationship between language and emotion comes either from corpus studies or from studies addressing the more pragmatic aspects of language use. For example, a recent large scale study (N = 4.754, Tackman et al., 2019) showed a small but reliable relationship between depression and self-referential language. At the other end; happy speakers have been shown to use more abstract language in their narratives (Beukeboom & Semin, 2006) and are less polite in their request (Forgas, 1999). The work of Kempe et al. (2013) is somewhat of an exception, because they show experimentally that emotion affects addresses the choice of referential expression. Here, we summarize a number of studies that probe the role of emotion in language production at the level of linguistic choice, in both experimental and corpus studies. Specifically, we will present work on to the role of emotion in lexical choice in image descriptions, on politeness in dialogue, and we will discuss our corpus based work to study role of affect and perspective on the linguistic realization of sports narratives. Surprisingly, many of the studies we will discuss fail to find a direct influence of emotion on the language production system. This disparity between the well-established influence of emotion on language characteristics (e.g., its abstractness or its self-referential nature) and the limited amount of evidence for a direct link between affect and lexical choice or realization will be discussed.