The Language of Trauma: A Comparative Analysis Using Natural Language Processing and Traditional Text Analysis Methods
Abstract
Post traumatic stress disorder (PTSD) is associated with alterations in the recall of autobiographical memories (AMs) regarding the traumatic event, but methodology of studies investigating traumatic and non-traumatic narratives in the context of PTSD vary. There is little consensus regarding the linguistic differences between different narrative valences and the influence of PTSD-symptomatology outside of clinical populations. We collected 450 narratives regarding traumatic, negative non-traumatic and neutral AMs of 150 healthy participants and their post-traumatic stress measured by the PTSD checklist for DSM-5 (PCL-5). For a comprehensive analysis of differences in linguistic patterns, we examined syntactic, lexical, semantic, and discourse-level features of the narratives. Compared to neutral narratives, traumatic narratives were linguistically richer, more PTSD-congruent in semantic content, less coherent, and were subjectively experienced as more emotionally intense and vivid. Trauma and negative narratives were largely similar in their linguistic and structural features, with differences confined to word count, death-related language, coherence, and semantic alignment with hyperarousal and intrusion symptom domains. Post-traumatic stress did not moderate any of the association between narrative categories and linguistic features, but higher PCL-5 scores predicted use of death-related words and how temporarily close participants felt to the traumatic event. Although future research should replicate these findings in clinical comparison samples, the present study provides a comprehensive attempt to characterize the language of trauma narratives.
Related
- Predicting One-Year Psychiatric Rehospitalization Using Natural Language Processing and Machine Learning Methods Based on Clinical Texts
- Do thoughts actually drive symptoms? Testing cognitive theories of posttraumatic stress with natural language processing in daily diaries
- Natural Language Processing for Risk Stratification of Emergency Department Patients with Chest Pain
- Combining natural language processing with patient-reported outcome measures scores to investigate the association of pandemic regulations with anxiety in children
- Using natural language processing analysis to investigate daily-life positive autobiographical memories
BibTeX
@misc{groer2026language,
title = {{The Language of Trauma: A Comparative Analysis Using Natural Language Processing and Traditional Text Analysis Methods}},
author = {Giuliano Groer and Laurin Plank and Armin Zlomuzica},
howpublished = {PsyArXiv},
year = {2026},
month = sep,
doi = {10.31234/osf.io/fjt7w_v1},
url = {https://osf.io/preprints/psyarxiv/fjt7w_v1/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - The Language of Trauma: A Comparative Analysis Using Natural Language Processing and Traditional Text Analysis Methods
AU - Giuliano Groer
AU - Laurin Plank
AU - Armin Zlomuzica
PY - 2026
DA - 2026/09/30
PB - PsyArXiv
DO - 10.31234/osf.io/fjt7w_v1
UR - https://osf.io/preprints/psyarxiv/fjt7w_v1/
AB - Post traumatic stress disorder (PTSD) is associated with alterations in the recall of autobiographical memories (AMs) regarding the traumatic event, but methodology of studies investigating traumatic and non-traumatic narratives in the context of PTSD vary. There is little consensus regarding the linguistic differences between different narrative valences and the influence of PTSD-symptomatology outside of clinical populations. We collected 450 narratives regarding traumatic, negative non-traumatic and neutral AMs of 150 healthy participants and their post-traumatic stress measured by the PTSD checklist for DSM-5 (PCL-5). For a comprehensive analysis of differences in linguistic patterns, we examined syntactic, lexical, semantic, and discourse-level features of the narratives. Compared to neutral narratives, traumatic narratives were linguistically richer, more PTSD-congruent in semantic content, less coherent, and were subjectively experienced as more emotionally intense and vivid. Trauma and negative narratives were largely similar in their linguistic and structural features, with differences confined to word count, death-related language, coherence, and semantic alignment with hyperarousal and intrusion symptom domains. Post-traumatic stress did not moderate any of the association between narrative categories and linguistic features, but higher PCL-5 scores predicted use of death-related words and how temporarily close participants felt to the traumatic event. Although future research should replicate these findings in clinical comparison samples, the present study provides a comprehensive attempt to characterize the language of trauma narratives.
ER -