Linguistic features of suicidal thoughts and behaviors: A systematic review. (July 2022)

• Record Type: Journal Article
• Authors: Homan, Stephanie; Gabi, Marion; Klee, Nina; Bachmann, Sandro; Moser, Ann-Marie; Duri', Martina; Michel, Sofia; Bertram, Anna-Marie; Maatz, Anke; Seiler, Guido; Stark, Elisabeth; Kleim, Birgit
• Is Part Of: Clinical psychology review. Volume 95(2022)
• Publication Date: 2022-07
• Subjects: Suicidal thoughts and behavior -- Linguistic features -- Prosody -- Lexicon -- First-person singular -- (morpho)syntax; Clinical psychology -- Periodicals; Psychology, Pathological -- Periodicals; Psychotherapy -- Periodicals; Psychology, Clinical -- Periodicals; Electronic journals; 616.89
• DOI: 10.1016/j.cpr.2022.102161
• Languages: English
• ISSNs: 0272-7358

Abstract

Abstract: Language is a potential source of predictors for suicidal thoughts and behaviors (STBs), as changes in speech characteristics, communication habits, and word choice may be indicative of increased suicide risk. We reviewed the current literature on STBs that investigated linguistic features of spoken and written language. Specifically, we performed a search in linguistic, medical, engineering, and general databases for studies that investigated linguistic features as potential predictors of STBs published in peer-reviewed journals until the end of November 2021.We included 75 studies that investigated 279, 032 individuals with STBs (age = 29.53 ± 10.29, 35% females). Of those, 34 (45%) focused on lexicon, 20 (27%) on prosody, 15 (20%) on lexicon and first-person singular, four (5%) on (morpho)syntax, and two (3%) were unspecified. Suicidal thoughts were predicted by more intensifiers and superlatives, while suicidal behaviors were predicted by greater usage of pronouns, changes in the amount of verb usage, more prepend and multifunctional words, more nouns and prepositions, and fewer modifiers and numerals. A diverse field of research currently investigates linguistic predictors of STBs, and more focus is needed on their specificity for either suicidal thoughts or behaviors. Highlights: Reliable prediction models for suicide risk are lacking. Promising predictors could be linguistic features. Those can be derived i.e., from therapy recordings, suicide notes, and blog posts. Lexicon, prosody, and first-person singular pronoun features predict suicidal risk. Syntax remains to be investigated.