Language with character: A stratied corpus comparison of individual dierences in e-mail communication

Abstract

The goal is to learn more about language production capacities by studying individual dierences between adults. We focus on the links between personality traits and language use in e-mail communication. We previously gathered a corpus of email messages written by individuals of known personality, as measured by Eysenck’s EPQ-R instrument. Here, it is argued that top-down content analysis techniques are not sucient to reveal expected patterns in the data. We therefore use a bottom-up stratied corpus comparison to isolate linguistic features associated with dierent personality types, via both word and part-of-speech n-gram analysis. It is shown that: Extraversion is associated with linguistic features involving uency, positivity and implicitness; Neuroticism with self-concern, negativity and implicitness; and Psychoticism with creativity and detachment. To explain these ndings, the paper discusses a model which links aect to language production processes.