Fanaticism and extremism are increasingly recognized as seminal to psychopathology and distress, especially considering the increase in political unrest and violence over the last decade. In the psychopathological literature, however, the cognitive style associated with extremism and overgeneralization has long been recognized as a risk factor for emotional distress, leading to both externalizing behavior (e.g. aggression) and internalizing pathology (e.g. depression). Despite its recognized importance, however, virtually no standardized measures of this cognitive style exist. Since direct inquiry about a respondent's Cognitive Rigidity, is likely to be biased, a text-analytical measure of extremism in spontaneous autobiographical narratives is proposed. In contrast to self-reports, naturally occurring speech often suggests cognitive proclivities towards overgeneralization, overconfidence or extremization. In this study, spoken autobiographical narratives were elicited from 483 participants, and contrasted with extensive mental health information using a hierarchical concordanced-keyword technique. The resulting corpus-based dictionary is context-sensitive, and exhibits significant correlations with measures of negative emotionality, with minimal association with response bias measures.