Individual differences and implicit language: personality, parts-of-speech and pervasiveness

Individual differences and implicit language: personality, parts-of-speech and pervasiveness Jon Oberlander (J.Oberlander@ed.ac.uk) School of Informatics, University of Edinburgh 2 Buccleuch Place, Edinburgh, EH8 9LW UK Alastair J. Gill (A.Gill@ed.ac.uk) School of Informatics, University of Edinburgh 2 Buccleuch Place, Edinburgh, EH8 9LW UK Abstract Dewaele and Furnham predict that in oral language Extraverts prefer to produce what they term im- plicit language. They use: more pronouns, adverbs and verbs; and fewer nouns, adjectives and prepo- sitions. However, communication in a computer- mediated environment, such as e-mail, might dis- rupt these preferences. Also, other personality di- mensions, such as Neuroticism, may be related to implicitness. The study exploited an existing cor- pus of e-mail texts written by native English speak- ers of known personality. Stratified corpus com- parison used n-gram-based techniques from statisti- cal natural language processing, to compare relative frequencies of use of (sequences of) parts-of-speech. Implicitness effects were found, and Neuroticism ap- peared to have a clearer impact than Extraversion. Personality and language Individuals differ in the way they speak and write. Some of those differences are systematic, and can be attributed to apparently deeper differences, such as personality traits, like Extraversion and Neuroti- cism. Extraversion is a trait strongly related to interpersonal interaction and sociability, whereas, Neuroticism, or Emotional Stability, is related to internal emotional states, rather than interaction. In the past, it has been found that both these per- sonality traits do significantly influence an individ- ual’s language production behaviour in a variety of contexts (Pennebaker and King, 1999; Dewaele and Furnham, 1999). Recent work has investigated e- mail text, and suggested that there are characteristic sequences of words and punctuation associated with each end of both dimensions (Extravert or Neurotic) (Gill and Oberlander, 2002, 2003). However, Mehl and Pennebaker (2003) note that linguistic style is more consistently described by its syntactic component, than by content. So, it could be that the relative use of different parts-of-speech (POSs) is a more important indicator of personality than the relative use of words or strings of words. The work by Dewaele and Furnham suggests that, at least for Extraversion, there are real effects to be found in spoken language, at the level of POSs. In their account, implicit language involves a preference for pronouns, adverbs and verbs, whereas explicit language involves a preference for nouns, adjectives and prepositions. Heylighen and Dewaele (2002) suggest that Extraversion leads to implicitness due to greater visual-spacial capacities, and this is part of an overall preference for informal language. How- ever, this work leaves open whether or not implicit- ness effects will be found for Neuroticism. Gill and Oberlander’s work suggests that formality may also be a factor in Neurotic language behaviour, because the reduced resources of high Neurotics do not en- able detailed language planning. But that work did not investigate implicitness in patterns of POS use. It would therefore be interesting to know whether Dewaele and Furnham’s ‘Implicit-Extravert hypoth- esis’ applies in the genre of e-mail text—a genre close to spoken language—and if so, how. To address this question, the rest of this paper is structured as follows. First, we give some back- ground to help frame implicitness hypotheses that gives POS predictions for both Extraversion and Neuroticism. We then present the stratified cor- pus comparison methods used in analysing POS use in the e-mail corpus. Results were somewhat unex- pected, in that implicitness predictions appear to be confirmed for Neuroticism, but not for Extraversion. We discuss possible ways of resolving the issue. Background Two personality traits Extraversion and Neuroticism are traits which are common to the two major trait theories of person- ality: Eysenck’s three factor model (Eysenck and Eysenck, 1991); and the five factor model developed by Costa and McCrae (Costa and McCrae, 1992) and others. They are described as follows: High Extraverts are said to be sociable, easy-going, and optimistic, and to take chances. Low Extraverts (or Intro- verts) are said to be quiet, and reserved, and to plan ahead, and dislike excitement. High Neurotics are said to be: anxious, worrying, over-emotional, and frequently depressed. Low Neurotics are said to be: calm, even-tempered, controlled, and unworried (Eysenck and Eysenck, 1991).