Worse than bad, better than good

This article investigates text-based emotional contagion in online customer feedback, focusing on the role of evaluative language in Twitter (now X) complaints and compliments. Specifically, it examines the impact of evaluative language on readers’ attribution of positive and negative emotions in these messages, as well as the emotions they themselves experience when reading them. The study focuses on Dutch-language Twitter posts in which customers publicly complain about or praise the Belgian national railway company. We report on a questionnaire experiment with a 2 x 2 design in which the valence (positive or negative) of corpus-based Twitter stimuli and the presence/absence of evaluative language were manipulated. Participants were first asked to report their own emotional states while reading the messages and were then asked to assess positive and negative emotions expressed in the same messages. The results show, first, that complaints containing evaluative language led to higher perceived sadness and lower perceived happiness than complaints without evaluative language. Second, while evaluative language did not affect readers’ emotional responses to compliments, self-reported sadness was higher for complaints with evaluative language than for those without. Third, participants reported greater surprise in response to compliments than to complaints overall. Finally, a significant positive correlation was found between the intensity of perceived emotions and that of self-reported emotions. Together, these findings provide novel evidence for emotion attribution and text-based emotional contagion in online customer feedback.