Subliminal facial primes bias emotion judgments during naturalistic videos in major depressive disorder

Emotional information is typically encountered in dynamic and context-rich environments, yet mood-congruent bias in major depressive disorder (MDD) has largely been studied using static stimuli. Subliminal emotional information modulations on behavioral and neural responses are investigated using a novel dynamic emotional video processing paradigm. In total, 126 participants (MDD: n = 60, Control: n = 66) completed an fMRI task in which happy or sad videos were presented while subliminal facial expressions (happy/sad; 16.7 ms) were flashed every second. Participants categorized video emotion (happy/sad/neutral) and rated subjective affect after each video. Outside the scanner, participants narrated positive and negative autobiographical events; speech features were extracted and tested for associations with questionnaire measures moderated by group. Across participants, emotionally congruent video-prime pairings were associated with faster and more accurate emotion judgments and increased activation in visual cortical regions during both happy and sad video processing. Incongruent pairings elicited broader recruitment of frontal, parietal, and temporal. Compared to healthy controls, MDD showed slower responses and increased activation in right occipito-temporal regions during video viewing. MDD show reduced accuracy during sad-prime/happy-video relative to congruent happy trials, while HC showed no prime effect on accuracy. No group-by-condition interaction was observed at the neural level. After correction for multiple comparisons, speech-questionnaire associations were not significantly moderated by group. Depression is characterized by reduced processing efficiency and increased perceptual-attentional demands during unconscious emotional conflict, particularly when happy videos are disrupted by sad unconscious stimuli. This supports an unconscious mood-congruent processing bias.