Assessment of cognitive negativity bias, psychopathogenic processes and speech temporal patterns using computerized text-analysis of personal narratives

In this dissertation, three text-analytical studies elucidate aspects of the relationship between speech patterns and personality characteristics. Text analysis provides a unique lens into the personality of the speaker and can potentially complement information obtained from clinical interviews or self-report questionnaires. Text-analysis has several notable advantages over self-report inventories, including minimal constraints on the individual's response, non-obtrusiveness and high reliability. The corpus of narratives used in these studies was collected from 483 undergraduate students who related a story concerning an interpersonal conflict with a significant other. In Study 1, I describe the development and validation of a text-analytical measure for positive and negative emotions. This measure considers the immediate context of target emotion words, and thus ensures that the words are used in their emotional sense. Concordance analysis of candidate emotion words determined the criteria by which these words were excluded or included in the dictionary. Text analysis indices based on these dictionaries of emotional language were more highly correlated with various measures of mental health than indices based on existing text-analytical dictionaries. In Study 2, I describe the development and validation of a text-analytical measure of the tendency to overgeneralize and exaggerate. Concordance analysis again determined exclusion and inclusion criteria of candidate superlatives. Indices based on this dictionary were related to measures of general psychological distress and dysphoria independently of the text-analytical indices of emotionality from Study 1. In Study 3, I report the use of Lagged Co-occurrence Analysis (LCA) to assess the temporal pattern of words in the stream of speech. LCA quantifies the sequential dependence among words within a narrative, and Study 3 shows that these patterns are psychologically meaningful by relating them to gender. Gender differences were found in the temporal pattern of the first pronoun "I", such that women refrain longer than men from saying the pronoun "I" after having previously used it. Overall, the three studies use text-analysis of personal narratives to extract psychologically-relevant information in a fashion that may reside outside the conscious awareness and control of the speaker.