2026 ECNP Congress

PS03-3149

Language profiles in young help seekers and their association with psychopathology

Chiara Colli — C. Bonvivento, S. Peri, C. Perlini, M. Mauri, A. Marini, M. Nobile, P. Brambilla · Fondazione IRCCS Ca' Granda Ospedale Maggiore Policlinico-, Department of Neurosciences and Mental Health, Milan, Italy

2026 ECNP Congress

Background

Language represents one of the most sophisticated human abilities; it is a central feature of human cognition and social interaction, and its disruption has long been recognized as a key manifestation of psychopathology. The study of language in psychiatry is now being revitalized by computational methods that allow for the objective quantification of linguistic features (1). This is particularly relevant in developmental psychiatry, where symptoms are often polymorphic and do not fit neatly within clear diagnostic boundaries. In this context, a transdiagnostic approach may better capture such heterogeneity, and linguistic analysis may offer a promising avenue for the early identification of risk markers (2).

Aims

This study investigated spontaneous narrative language in a transdiagnostic sample of children and adolescent help seekers. Our aim was to identify linguistic profiles and examine their relationship with psychopathology.

Methods

We quantified nine standardized linguistic variables by analyzing spontaneous speech in 318 participants, including 250 help-seeking youths (mean age 14.45 ± 2.38 years) and 68 controls (mean age 14.46 ± 4.46 years) Variables includedMacro-linguistic features(i.e., Global coherence errors, Local coherence errors, Lexical informativeness, Flunecy) and Micro-linguistic features (Syntactic completeness, Paragrammatic errors, Semantic errors, Phonological errors, Words).

The participants’ total scores at the Child Behavior Checklist/Adult Behavior Checklist (CBCL/ABCL), evaluating emotional, attention and behavioral problems, were also entered into the analyses.

Principal component analysis (PCA) was performed to reduce intercorrelated linguistic features into latent dimensions. We then applied k-means clustering to identify homogeneous linguistic subgroups. Between-cluster comparisons were conducted using non-parametric tests (Kruskal-Wallis and Wilcoxon).

Results

From PCA performed on nine linguistic variables, four principal components (PCA) were retained, explaining 67.8% of the total variance (PC1 = 27.98%, PC2 = 16.54%, PC3 = 11.83%, PC4 = 11.49%). K-means clustering identified three clusters (Cluster 1: n = 101; Cluster 2: n = 80; Cluster 3: n = 137), which were mainly differentiated by two PC: PC1, capturing productivity and informativeness, and PC2, reflecting completeness and a lower level of errors. Cluster 1 was characterized by low PC1 scores, with relatively preserved accuracy, and showed intermediate levels of psychopathology (CBCL/ABCL total score = 53.6). Cluster 2 showed low PC2 scores and exhibited the highest level of psychopathology, with significantly higher CBCL/ABCL total scores (57.0) than Cluster 3, which showed the lowest symptom levels (52.3). Cluster 3 was characterized by high scores on both PC1 and PC2.

Kruskal–Wallis tests indicated no cluster differences in age (χ² = 3.52, p = 0.17) and a difference in CBCL total score across the 3 clusters (χ² = 14.06, p < 0.001). Bonferroni-adjusted post-hoc Wilcoxon comparisons showed that Cluster 2 scored significantly higher than Cluster 3 on CBCL total score (p = 0.00043).

Conclusions

Overall, these results suggest that linguistic patterns can map onto levels of clinical vulnerability, supporting the potential use of language as a transdiagnostic biomarker, particularly during the developmental period. The study of spontaneous language measures can therefore provide valuable support for the clinical profiling of youths at risk for future mental health problems.

References

References

  1. Nour, M. M., & Huys, Q. J. M. (2023). Natural Language Processing in Psychiatry: A Field at an Inflection Point. Biological Psychiatry. Cognitive Neuroscience and Neuroimaging, 8(10), 979–981. https://doi.org/10.1016/j.bpsc.2023.08.001
  2. Abi-Dargham, A., Moeller, S. J., Ali, F., DeLorenzo, C., Domschke, K., Horga, G., Jutla, A., Kotov, R., Paulus, M. P., Rubio, J. M., Sanacora, G., Veenstra-VanderWeele, J., & Krystal, J. H. (2023). Candidate biomarkers in psychiatric disorders: State of the field. World Psychiatry: Official Journal of the World Psychiatric Association (WPA), 22(2), 236–262. https://doi.org/10.1002/wps.21078