A corpus-based analysis of potential linguistic indicators of corporate deception in tobacco industry documents

Abstract Introduction: To more fully understand the impact of specific language attributes on deception in corporate communication, we used a stratified random sample of tobacco industry documents to test a small number of individual potential automated linguistic indicators of corporate deception: cognitive-emotional verbs, allness and superlative terms, nonbinding verbs, and group mentality. Methods: Texts from the Tobacco Documents Corpus were categorized by audience addressed (industry internal/external) and company of origin, and compared for incidence of the deceptive language indicators. Results: Cognitive-emotional verbs were strongly associated with likely deception and most prevalent for external audiences and documents produced by the Tobacco Institute, an industry front group. Cognitive-emotional verbs include believe, think, seem, feel and realize, in opposition to action verbs (e.g. throw). Discussion: Linguistic measurements of cognitive-emotional verbs may help pinpoint deceptive and misleading corporate communication. Verb choice and the use of different classes of verbs may reflect corporate linguistic deception.