Embodying Affect in Narrative Discourse: a facial EMG study on simulating affective language in context.

B. 't Hart, M.E. Struiksma, Anton van Boxtel, J.J.A. van Berkum

Research output: Working paperAcademic


In this experiment we examine the role of facial muscle activation (fEMG) in the online comprehension of affective language. fEMG research has shown that the corrugator supercilii (‘frowning muscle') is a reliable indicator of affective valence in processing all manner of stimuli, with activation indicating negative affect and deactivation indicating positive affect (e.g., Larsen, Norris, Cacioppo, 2003). Research into embodied language processing has previously used this method to show simulation effects at the lexical and phrasal level. Crucially, the inhibition of specific facial muscle activation resulted in impaired or delayed processing, suggesting a causal role in language comprehension (e.g., Foroni & Semin, 2009, Niedenthal et al., 2009, Havas et al., 2010). The current experiment extends the work mentioned above to richer, multi-sentence narratives. Apart from increasing ecological validity, richer narratives also present interesting opportunities to explore a more complex situation, where valence is no longer unambiguously positive or negative (as in the typical lexical/phrasal level experiment), but a layered phenomenon involving different perspectives. For example, although a particular event may have positive valence for the protagonist in focus (e.g., something good happens to him or her in the story), it may in the end have negative valence for the reader (e.g., when something good happens to a character that the reader has cause to dislike). Using fEMG, we investigated whether the simulation of a particular linguistic stimulus (such as "Mark was happy when he found out he'd won the lottery") could be influenced by the evaluative context, by having 60 participants read 64 short narratives that orthogonally contrasted protagonist valence with evaluative reader valence. In each narrative, we manipulated reader valence by depicting a protagonist as likeable or dislikeable, after which this person experienced a positive or negative event. Our hypothesis was that the likeability manipulation would tease out protagonist and reader valence and that we would see this reflected in the fEMG response. For instance, we predicted that something bad happening to a disliked character should result in a positive fEMG pattern rather than a negative one. We used a mixed model linear regression analysis to build a growth curve model of the fEMG response; this analysis preserved information regarding the temporal development of the response. We found a significant effect of the evaluative context on the fEMG response pattern at the affective event. We found that positive events resulted in increased corrugator activity (negative affect) and negative events in decreased corrugator activity (positive affect) when the character was dislikeable. Moreover, the time course of the response reveals that this influence is immediate. This result not only confirms that evaluative appraisal of affectively salient language is an integral part of language processing (e.g., schadenfreude in the case of a disliked character experiencing something negative), but also constrains ideas on how facial muscles are involved in language processing at the protagonist level.
Original languageEnglish
Publication statusUnpublished - 2015


Dive into the research topics of 'Embodying Affect in Narrative Discourse: a facial EMG study on simulating affective language in context.'. Together they form a unique fingerprint.

Cite this