This study employed a 2 × 2 × 3 mixed experimental design. The required sample size was estimated using G*Power 3.1, with the effect size set at f² = 0.25, α = 0.05, and 1 − β = 0.80. The power analysis indicated that at least 17 participants were required per group. From October 1, 2024, to October 1, 2025, a total of 831 copies of the Center for Epidemiologic Studies Depression Scale (CES–D) were randomly distributed to university students at ******. Participants were subsequently invited from among those who provided valid responses to take part in the experimental study. A total of 64 participants were ultimately included and assigned to two groups: the experimental group (subclinical depression group) and the control group (nondepressed group).The inclusion criteria were as follows: (1) participants did not meet the DSM-5/ICD-10 diagnostic criteria for a depressive episode and had no history of depressive episodes; (2) all participants had normal or corrected-to-normal vision and were right-handed; (3) individuals with organic brain diseases or diagnosed psychiatric disorders were excluded; and (4) participants in the experimental group had CES–D scores ≥ 16, whereas those in the control group had CES–D scores < 16 (Radloff, 1977).All experimental stimuli were selected from the China Affective Picture System (CAPS) and the International Affective Picture System (IAPS). An independent sample of 75 university students who did not participate in the formal experiment rated the images for valence and arousal. After balancing the valence and arousal ratings across emotional categories, 240 facial images and 240 scene images were selected. For each image type, the stimulus set consisted of 96 positive, 96 negative, and 48 neutral images. All images were resized to 260 × 300 pixels.Two principal emotional-image configurations were used for both facial and scene stimuli: (1) two positive images, one negative image, and one neutral image were randomly selected; or (2) one positive image, two negative images, and one neutral image were randomly selected. The four images were randomly presented in the upper-left, upper-right, lower-left, and lower-right quadrants of the stimulus display.Eye movements were recorded using an EyeLink 1000 Plus eye-tracking system (SR Research Ltd., Canada) at a sampling rate of 1000 Hz. Movements of each participant’s left eye were tracked and recorded during the task. A nine-point calibration procedure was conducted before the experiment and after each interruption. For successful calibration, participants were required to maintain fixation for at least 200 ms within a 1.5 mm × 1.5 mm window surrounding the fixation point. If calibration failed three times, the procedure was repeated until successful, and data collection commenced only after successful calibration. Visual stimuli were presented on a 19-inch Dell LCD monitor (P1917S; resolution: 1024 × 768 pixels; refresh rate: 60 Hz). Participants were seated approximately 60 cm from the monitor, with head movements restricted using a chin rest.Functional near-infrared spectroscopy data were recorded using a portable NIRSport system in continuous-wave mode at a sampling rate of 7.81 Hz and at two wavelengths (760 and 850 nm). The optode cap contained eight emitters (light sources) and seven detectors, forming a total of 22 measurement channels. The source–detector distance was fixed at 3.0 cm. Optodes were positioned according to the international 10–20 system using an EEG cap (EASYCAP, Herrsching, Germany).