Previous analyses of released tobacco industry documents succeeded in unearthing certain instances of industry deception and manipulation that deepened the threat to public health posed by tobacco use. None of these prior studies, however, sampled documents from the database in any systematic fashion, nor did they employ theoretically motivated or empirically validated tools of linguistic science. Thus the pervasiveness of deceptive and manipulative practices within the document set cannot currently be estimated. Moreover, subtle discourse strategies of deception and manipulation are most likely to be revealed only by fine-grained linguistic analyses that compare across texts. As an illustration, equivocation as a deceptive practice can be indexed by relative frequency of modifiers and hedges like """"""""sort of,"""""""" """"""""nearly,"""""""" and """"""""in a manner of speaking."""""""" Strategies relating to audience manipulation can be revealed by comparing use of suasive language (""""""""because,"""""""" """"""""obviously"""""""") across messages intended for differing readers. Much of this type of language analysis can now be accomplished by computerized text analysis. The overall objective of the proposed study, then, is to treat the tobacco industry documents as a corpus of language to be analyzed by means of accepted tools of applied and forensic linguistics and rhetorical analysis. More specifically the following activities will be undertaken: (1) Create 50 sets of cross-audience text samples, i.e., similar topics directed toward varying internal and external audiences; (2) Create 50 sets of cross-draft text samples, i.e., tracing evolution of key documents from conception to commentary to distribution; (3) Apply to these text samples selected linguistic indices of deception and obfuscation, e.g, relative frequency of agentless passive constructions such as """"""""Nicotine levels were augmented.""""""""; (4) Apply to these text samples selected linguistic indices of audience adaptation and manipulation, e.g., a commonality index containing terms of cooperation, inclusion; (5) Identify contrasting patterns of these linguistic markers across drafts and across audiences; (6) Disseminate findings broadly to the public health, legal, and applied linguistics communities. It is anticipated that significant linguistic variation across audiences or across drafts will reveal language strategies by which information about adverse health effects of smoking was deceptively expressed to citizen audiences, or by which responsibility for adverse effects was deflected from the tobacco industry. These language strategies may persist even in contemporary tobacco industry statements or in documents of other industries responsible for potential health risks (e.g., pesticide manufacture). The public health benefits of this research therefore are to expose such language practices to scrutiny and thus to guard against future occurrences.
