Identifying structural linguistic markers of depression in written text: A narrative review of language analysis methods

• Existing research on linguistic markers of depression has focused on content-based features, with fewer studies examining structural aspects of written language. • Structural linguistic markers such as syntactic complexity and referential cohesion are theoretically relevant to depression but remain empirically underexplored. • Evidence from related clinical populations shows that structural language measures can capture cognitive vulnerability, providing proof-of-concept for future depression research. • Current findings on structural markers in depression are preliminary, and further empirical work is required before clinical applications can be established. • A range of Natural Language Processing tools and pipelines are available for extracting structural linguistic features, offering practical pathways for advancing this research area. The study of ‘linguistic markers’ has gained prominence as a method of exploring how depressive symptoms are reflected in language use. In the context of depression, existing research has predominantly focused on perspectives of analysis related to language meaning and thematic content (e.g., semantics). In contrast, techniques that examine the underlying structure and grammatical composition of language (e.g., syntax) remain a novel application for identifying text-based markers of depression. Digital phenotyping enables the efficient collection of written text samples through digital, remote, and low-cost methodologies, making written language a practical and scalable target for research. Coinciding with advancements of natural language processing, this creates the timely opportunity to conduct the automated analysis of text samples. As such, this paper provides an overview of language analysis techniques for the examination of linguistic markers of depression. In particular, we focus on perspectives of structural language analysis, highlighting syntactic complexity and referential cohesion as potential markers, and the insight these features may provide into the linguistic manifestations of depressive symptoms. We discuss the clinical translation of content-based markers of depression through digital phenotyping and just-in-time adaptive interventions, and the potential benefit of incorporating structure-based markers of language. Finally, we discuss automated language analysis tools for their use in clinical and research settings.