That's frowned upon: Using facial EMG to track evaluation and simulation during affective language processing

This thesis presents research on the interface between language and emotion. Specifically, it investigates the way language describing emotions and emotional events is understood. It does so within the theoretical framework of grounded cognition. The main aim in the experiments is an examination of the claims made by theories of grounded language comprehension that multi-modal simulation of the meaning of linguistic stimuli takes place during (affective) language comprehension. One common way in which simulation of emotionally salient language is often measured is through facial electromyography (EMG), specifically of the corrugator supercilii, i.e. the frowning muscle. A number of studies have shown that during online processing of various types of language stimuli with a positive or negative affective meaning, the corrugator involuntarily and robustly reflects the same valence with either increased activation (negative affect) or decreased activation (positive affect). This correspondence in terms of valence is taken as an indicator of language-driven simulation.
The series of three experiments presented here looks at how such a simulation account of affective language processing meshes with the fact that, as readers or listeners, we also relate emotionally to much of the language we process in everyday life. Rather than using emotionally unambiguous stimuli, these experiments use short narratives that manipulate the likeability of the main character primarily by describing them engaged in moral or immoral behaviour. This, in turn, influences the reader's evaluation of the language describing the emotions and the emotional events that this character experiences. In some cases then, the corrugator activity predicted by simulation and evaluation contradict each other. For a character evaluated as immoral, simulation would predict positive corrugator activity while processing langauge describing this character as 'happy' or experiencing a positive event. At the same time, this immoral character experiencing positive emotions or events would be evaluated by the reader as negative and thus could also lead to negative corrugator activity.
The three studies showed that, first of all, readers immediately evaluate the described behavior of story characters. We found robust patterns of corrugator activity congruent with the moral status of the character for such behavioural descriptions, but not when the moral status of a character was simply asserted as a given rather than described narratively. Regardless of the presence or absence of a corrugator response to the character manipulation, we repeatedly found clear evidence that the moral status of the character impacted corrugator activity during subsequent processing of language describing their emotions or emotional events. More concretely, there was no differential corrugator activity in response to affective state adjectives (e.g., happy vs. sad), and reduced patterns of activity for descriptions of emotional events when they referred to immoral characters. For moral characters, however, we found clear, congruent differential patterns of corrugator activity.
These results clearly speak against an interpretation of corrugator activity specifically, and facial EMG more generally, in terms of langauge-driven simulation. This thesis presents clear evidence of the impact that evaluation has on such a peripheral psychophysiological process during online (affective) language processing. Such findings also have implications for an account of language comprehension in terms of simulation, suggesting that simulation is more contextually constrained than previously thought.