Eye tracking during reading of affective lexical stimuli enables prediction of depression

This study investigated whether eye-tracking responses to expectancy violations and self-referential processing can objectively classify depression diagnosis and risk. Fifty participants (15 with Major Depressive Disorder, 35 healthy controls) viewed vignettes ending with a final critical word that was either congruent with a positive or negative outcome while eye movements were recorded at a sampling rate of 1000 Hz. Vignettes were presented in either third-person non-self-referential or first-person self-referential formats. All participants completed depression assessments using the Beck Depression Inventory-II (BDI-II) and Patient Health Questionnaire-9 (PHQ-9). Nineteen eye-tracking features were extracted from epochs, time-locked to critical word onset, including saccade dynamics, fixation patterns, pupil responses, and gaze entropy. Random Forest classifiers were trained using a 60-40 train-test split and 20-fold cross validation to predict clinical diagnosis, BDI risk, and PHQ-9 risk categories. Saccade-related measures consistently differentiated between groups, while pupil dilation responses indicated expectancy violation processing. Machine learning classification achieved 83.8% accuracy for clinical diagnosis, 81.0% for BDI risk, and 71.2% for PHQ-9 risk. Self-referential stimuli contributed more heavily to successful classification than non-self-referential stimuli, with saccade displacement and velocity measures being the most important features. Several eye-tracking features, including saccades, pupil dilation, and entropy impacted model performance. These findings demonstrate that eye-tracking responses to semantic expectancy violations can provide objective biomarkers for depression classification, potentially offering a valuable tool for clinical assessment when traditional self-report measures may be unreliable or unavailable.

Authors

Publication Details

Journal
PsyArXiv (OSF Preprints)
Published
2026-11-18
DOI
https://doi.org/10.31234/osf.io/wu2av
Primary Topic
Gaze Tracking and Assistive Technology
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Eye tracking during reading of affective lexical stimuli enables prediction of depression

Victoria P. Sharpe, Phillip C. Desrochers, Masoud Nateghi, Omer Refik Inan et al.
3 links
PsyArXiv (OSF Preprints)
Gaze Tracking and Assistive Technology
preprint

Eye tracking during reading of affective lexical stimuli enables prediction of depression

Victoria P. Sharpe, Phillip C. Desrochers, Masoud Nateghi, Omer Refik Inan, Gina R. Kuperberg, Bethany K. Bracken, Reza Sameni, Zeineb Bouzid, J. Douglas Bremner, Spencer K. Lynn, Lina Chitadze, Bradly Stone, Gabriella Cestero, Apoorva Vallamapati, Yi Yang, Cem Yaldiz, Rachel Bull, Arim Choi, Chuoqi Chen
preprint en

Abstract

This study investigated whether eye-tracking responses to expectancy violations and self-referential processing can objectively classify depression diagnosis and risk. Fifty participants (15 with Major Depressive Disorder, 35 healthy controls) viewed vignettes ending with a final critical word that was either congruent with a positive or negative outcome while eye movements were recorded at a sampling rate of 1000 Hz. Vignettes were presented in either third-person non-self-referential or first-person self-referential formats. All participants completed depression assessments using the Beck Depression Inventory-II (BDI-II) and Patient Health Questionnaire-9 (PHQ-9). Nineteen eye-tracking features were extracted from epochs, time-locked to critical word onset, including saccade dynamics, fixation patterns, pupil responses, and gaze entropy. Random Forest classifiers were trained using a 60-40 train-test split and 20-fold cross validation to predict clinical diagnosis, BDI risk, and PHQ-9 risk categories. Saccade-related measures consistently differentiated between groups, while pupil dilation responses indicated expectancy violation processing. Machine learning classification achieved 83.8% accuracy for clinical diagnosis, 81.0% for BDI risk, and 71.2% for PHQ-9 risk. Self-referential stimuli contributed more heavily to successful classification than non-self-referential stimuli, with saccade displacement and velocity measures being the most important features. Several eye-tracking features, including saccades, pupil dilation, and entropy impacted model performance. These findings demonstrate that eye-tracking responses to semantic expectancy violations can provide objective biomarkers for depression classification, potentially offering a valuable tool for clinical assessment when traditional self-report measures may be unreliable or unavailable.

PsyArXiv (OSF Preprints)
Good health and well-being
Gaze Tracking and Assistive Technology
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon