Identification of the marks of psychic trauma in spoken language: Definition of the “SPLIT-10” diagnostic scale

The search for understanding psychological trauma has grown considerably over the past fifteen years, leading to a real conceptual revolution at the crossroads of psychiatry, psychology, neurobiology, sociology, and anthropology, among others. However, despite the wealth of semiological descriptions, the under-diagnosis and late diagnosis of post-traumatic disorders, at the stage of intense suffering, remain numerous. This could be because the traumatic origins of the disorders remain unclear, due to their clinical characteristics–that is, the “unspeakable experience” of dissociation in language–, or because the healthcare system and the networks of practitioners come up against conceptual impasses that undoubtedly reflect the psychotraumatic process present even in theoretical discourse, to the point of rendering it ineffective. Now is the time to build a new model. Based on a linguistic methodology, the standardized computerized and manual study of the speech of psychically injured patients recently enabled us to define the notion of traumatic psycholinguistic syndrome (SPLIT). Our new perspective aims to overcome the diagnostic and therapeutic obstacles that too many people with mental health injuries still face. More objective than semiological and psychometric approaches, linguistic markers pave the way for the digital phenotyping of post-traumatic stress disorder and make it possible to better assess the recommended care. After discussing the pioneering work in the literature, we build a psycholinguistic tool allowing for the identification of psychically injured subjects. The exploratory analysis material includes two corpora of traumatic event narratives called “Bataclan” (n=20 collected among survivors of the Paris attacks in 2015) and “Afghanistan War” (n=15 collected among French soldiers deployed), which are matched to a control group. The narratives were transcribed and segmented into clauses and the following linguistic characteristics were analyzed: disfluencies (silent pauses, hesitation pauses, vocalic lengthenings, incomplete words, incomplete utterances, contiguous word repetitions). Narrative coherence, certain lexical fields (concerning death, emotions, etc.), spatio-temporal references, references to the person (personal and generic pronouns in particular), and non-literal language were also taken into account. In order to confirm the validity of the SPLIT-10 scale, we tested it on two additional corpora of traumatic narratives: the “