Psycholinguistic Features of Depression-Related Discourse in Online Communities: A Comparative Corpus-Based Study

Abstract

The language of depression found in social media communities offers insights into the emotional, cognitive, and social experiences of its users. This study aimed to compare the psycholinguistics of depression-related discourse with non-depression discourse on Reddit to shed light on the differences between the two discourses in a quantitative and cross-sectional comparative corpus-based study. The first set of texts was in English and numbered 7731. Three thousand seven hundred and sixty-one depression-related and three thousand eight hundred and eighty-nine non-depression texts remained in the analytical corpus after 81 exact duplicates had been removed. Document-level linguistic features were extracted and were normalized as occurrences per 100 words. Lexical diversity, personal pronouns, emotional vocabulary, terms related to cognitive processes, social language, terms associated with time orientation, negation, and absolutist terms were analyzed. Mann-Whitney U tests, Cliff's effect size, and Benjamini-Hochberg corrections were used to assess group differences. Depressive texts had significantly more lexical diversity and were significantly longer. They also had more first-person singular pronouns, negative emotions, sadness, anxiety, fear, insight, causation, social interaction, and absolutist language. The greatest differences occurred in negative emotion, sadness, social interaction, and absolutist expressions. By contrast, second-person pronouns, present-focused language, future-focused language and normalized negations were found to be more common in non-depression discourse. The results show that the Reddit comments on depression have a unique and explainable psycholinguistic profile. They help with corpus-based research on mental health and ethically designed digital monitoring.