A study published in Cognition & Emotion provides evidence that the general use of negative words in natural language are associated with elevated symptoms of depression and anxiety over time, regardless of individual mental health status.
Depression and anxiety are two of the most prevalent psychological disorders, prompting research into early detection and symptom monitoring using everyday language. Studies suggest that language, especially the frequency of negatively valenced words, can reveal psychological states. Izabela Kaźmierczak and colleagues conducted this study to explore whether patterns in language use, particularly in real-life settings, could serve as reliable markers for these mental health conditions.
Previous studies have shown mixed results regarding positive language usage in relation to symptoms of depression. To clarify these results, the researchers examined both positive and negative language across participants, including individuals with clinically diagnosed depression and non-depressed controls. This work contributes to an emerging area that utilizes natural language processing to improve mental health diagnostics.
The researchers recruited 40 participants (22 women and 18 men), ages 22 to 44, who had recently undergone a significant negative life event. Participants were carefully selected to include those not currently in therapy, allowing researchers to examine language sentiment without intervention effects. Each participant was assessed for depression using the SCID-I diagnostic module, which categorized them into depressed and non-depressed groups.
Additionally, the Hospital Anxiety and Depression Scale (HADS) measured the intensity of their depressive and anxiety symptoms. Participants were recorded describing critical life events three times over a year, each session lasting about an hour. These sessions were spaced out to align with key stages of emotional adjustment: the initial shock period (within two months of the event), a period of inner crisis (around five months later), and a phase of reconstruction (eight or more months post-event).
Throughout each session, participants described their experiences in structured interviews. These interviews were recorded, transcribed, and later analyzed for sentiment using sentiment dictionaries designed for the Polish language. The dictionaries contained thousands of words labeled for positive or negative valence.
Across all sessions, the study amassed 1,440 narratives. The sentiment analysis, focusing on both positive and negative word use, allowed researchers to track shifts in sentiment across time and across participants with varying levels of depression and anxiety symptoms.
The results showed that participants who generally used more negative words in their language also had higher levels of depression and anxiety symptoms, a trend that was evident across both clinically depressed and non-depressed groups. This association between negative word use and symptom severity was strong, highlighting that persistent use of negatively valenced language could indicate more intense mental health challenges. In contrast, positive language use was associated with slightly lower levels of depression and anxiety, though this link was weaker compared to that of negative language.
The researchers also found that changes in the frequency of positive or negative word use over time did not correspond with changes in depression or anxiety symptoms. This suggests that a stable pattern of language use, rather than short-term fluctuations in word sentiment, may be a more reliable indicator of mental health status.
These findings underscore the potential of negative language patterns as a marker for assessing depression and anxiety severity.
One limitation noted by the authors is the study’s reliance on spoken language, which may differ from written communication, especially in online contexts.
The research, “Natural language sentiment as an indicator of depression and anxiety symptoms: a longitudinal mixed methods study”, was authored by Izabela Kaźmierczak, Adrianna Jakubowska, Agnieszka Pietraszkiewicz, Anna Zajenkowska, David Lacko, Aleksander Wawer, and Justyna Sarzyńska-Wawer.