Syntactic complexity of written text as a structural linguistic marker of depressive symptoms

Language analysis methods have been increasingly explored for the detection of depressive symptoms. However, current approaches have largely focused on analysing the thematic content of language, while the underlying structure and grammatical composition of language remains underexamined. The current study assessed features of syntactic complexity in written text as potential linguistic markers of depressive symptoms.

Adult participants (n = 196) experiencing depressive symptoms completed four writing tasks (Personal Biography; Neutral Writing; Narrative Imagery; Letter to a Friend). Depressive symptoms were measured via the Patient Health Questionnaire-9 (PHQ-9). Linguistic features were extracted with the Tool for the Automatic Analysis of Syntactic Sophistication and Complexity (TAASSC). Bivariate correlations were used to assess associations between syntactic complexity features and depressive symptoms. Machine learning models developed with nested cross-validation were used to examine the combined value of syntactic complexity features in written text for predicting depressive symptoms.

Across writing tasks, bivariate correlations showed weak, negative associations between total PHQ-9 scores and both noun phrase complexity and syntactic sophistication. Best-performing machine learning models varied by writing task, and syntactic complexity features accounted for up to ∼7% of variance in depressive symptoms. Features of importance included syntactic sophistication, noun phrase complexity, and clause complexity.

Features of syntactic complexity within written text could serve as linguistic markers of depression. However, features were weakly associated with depression and varied across writing tasks. Combining features of syntactic complexity with other established linguistic markers may add value to language analysis methods used to detect depression.