Numerous studies have investigated the language that PTSD
sufferers use in “trauma narratives” describing their trau-
matic experiences (for a review, see (O’Kearney and Perrott
2006)), but only a handful of these studies have used au-
tomated analysis of language. The Linguistic Inquiry Word
Count (LIWC) (Pennebaker et al. 2007) has been used to
analyze the narratives of 28 female assault victims being
treated for chronic PTSD, and found that the LIWC cog-
nitive words category was inversely correlated with post-
treatment anxiety, and that social adjustment was negatively
relatedtonegativeemotionwordsanddeathwords(Alvarez-
Conrad, Zoellner, and Foa 2001).
We conduct a LIWC analysis of the PTSD and non-PTSD
tweets to determine if there are differences in the language
usage of PTSD users. We applied the LIWC battery and
examined the distribution of words in their language. Each
tweet was tokenized by separating on whitespace. For each
user, for a subset of the LIWC categories, we measured the
proportion of tweets that contained at least one word from
that category. Specifically, we examined the following nine
categories: first, second and third person pronouns, swear,
anger, positive emotion, negative emotion, death, and anxi-
ety words. Second person pronouns were used significantly
less often by PTSD users, while third person pronouns and
words about anxiety were used significantly more often. Un-
surprisingly, given the stylistic and topical difference be-
tween tweets and trauma narratives, we do not observe the
same trends as Alvarez-Conrad et al. (2001), but the fact that
significant differences in language use exist is an encourag-
ing demonstration of the effectiveness of our data.