Receiving a Ph.D. in Linguistics yesterday from Stanford: Tyler Schnoebelen, on the linguistics of affect.
Emotions are relational: Positioning and the use of affective linguistic resources
The dissertation abstract (or you can watch a three-minute video summary here, starting at 1:26:00):
Understanding expressions of emotion means understanding how people use collections of linguistic resources to position themselves, their audiences, and their topics relative to one another. Expressions of emotion are not just internal states made visible. They are positional: reflecting, creating, and changing relationships. I demonstrate how positioning works in three case studies.
I begin with an emotional telephone conversation between two friends. This shows how affect is communicated at every linguistic level as well as how to integrate what’s happening in a situated context with larger, community-wide patterns. In particular, I show how speakers take and avoid agentive positions and how they deploy a variety of linguistic resources to increase or decrease the emotional immediacy of what they’re saying.
Next, I analyze little from a variety of directions to show how speakers use it in positioning. Experimental work reveals linguistic factors that are significant for judgments of confidence, likeability, overall emotional intensity, and positivity/negativity. In information theoretic terms, little carries very little information when it’s modifying a common collocate. As we might expect, there are strong reactions when you pair little with a non-collocate (how’s your little project? is negative in a way that how’s your project? is not). But it turns out that little‘s presence/absence with common collocates also makes a difference: they are much more positive with little than without.
I also describe the distribution of little across six different corpora, focusing on who is using it, with whom, and in what sort of situations. Two of these corpora have clear hierarchies that allow me to show the role of power. Parents and children use little very differently, as do people with different education backgrounds in the ICSI meeting corpus. I then go on to problematize gender. At a coarse level, women use little more than men, but this rather misses the point because of the significant interactions between speaker, audience, and topic.
Finally, I offer a comprehensive analysis of emoticons. I uncover emoticon ”dialects” having to do with whether people use noses or not
: – ) vs. : )
This demonstrates who is using which emoticons and when, but the bigger project is to describe the affective dimensions structuring the emotional universe of Twitter. Across hierarchical cluster analyses, factor analyses, and topic modeling, positive/negative emerges as an important dimension as we’d expect. But so do some more unusual ones: immediacy, teasing, and flirting, each of which is also positional.
Direct queries to Tyler at firstname.lastname@example.org.