Evaluation of a free-viewing task to measure distinct negative and positive biases in depression

Abstract

Attentional bias has traditionally been inferred through the measurement of reaction-time-based tasks. Eye-tracking offers a way to measure attention bias directly, and free-viewing tasks with intricate stimuli presentations may capture the complexities and dynamics of attention bias in ways previous modalities have not. The present study developed a free-viewing task using a data-driven stimuli selection process. Two free-viewing tasks were created using sad and neutral stimuli, and happy and neutral stimuli to tease apart the distinct effects of negative and positive bias. In this study, eye tracking data was collected and analyzed from n = 130 participants using mixed-effect and generalized linear models. Results revealed the interaction term (depression severity and stimuli valence) influenced dwell time on emotional stimuli, such that with an increase of 1 SD in depression severity (7.87 points on the Beck Depression Inventory-II), participants spent less 60 ms less time viewing sad stimuli and 25 ms less viewing happy stimuli. A significant interaction of depression severity and valence also influenced participant’s latency to first fixation. Increased depression severity (1 SD) was associated with increased odds of being slower to fixate on stimuli when it was sad (OR = 1.10) and when it was happy (OR = 1.03). There was no effect of depression severity or stimuli valence on latency to first fixation, nor an effect of depression severity on the proportion of trials where the first fixation was emotional or proportion of trials where dwell time for emotional areas of interest (AOIs) exceeded neutral. Internal consistency for emotional dwell time was high for both tasks (omega = .95 and .94 for the sad and happy versions, respectively), and split-half reliability for the outcomes was overall strong. Findings suggest depressed individuals may interact with stimuli differently at various levels of depression severity. Implications for future research are discussed.