Foreground, midground, and restoration: EEG and preference responses to depth-controlled rural scenes

This is the authors' abstract. We don't add key points for this paper.

Introduction Exposure to natural and rural landscapes has been associated with stress reduction and psychological restoration. However, conventional photograph-based experiments often provide limited control over individual landscape elements and their positions within the visual field. This study examined whether depth-controlled, AI-generated rural landscape stimuli could support stress recovery and clarify how foreground and midground elements relate to landscape preference and EEG-based responses. Methods Depth estimation, layer masking, and FLUX-based image editing were used to construct 27 controlled rural landscape stimuli. Selected foreground and midground elements were modified while the background context was kept relatively stable. Thirty participants completed a stress-induction task followed by rural landscape viewing and a duration-matched blank-screen control condition. Systolic blood pressure, diastolic blood pressure, heart rate, EEG beta/alpha (β/α) stress index, and POMS mood scores were measured at baseline, after stress induction, and after recovery. Landscape preference ratings and image-level EEG responses were analyzed for the rural landscape stimuli. Results Compared with blank-screen viewing, rural landscape viewing produced significantly greater recovery in all seven stage-level outcomes after Holm correction. Rural-minus-blank recovery differences were −4.45 mmHg for systolic blood pressure, −3.93 mmHg for diastolic blood pressure, −5.85 bpm for heart rate, and −0.10 for the EEG β/α stress index. Within the rural landscape stimuli, foreground water was associated with higher preference ratings (estimate = 0.500, SE = 0.084, p < 0.001), whereas foreground road was associated with lower preference (estimate = −0.354, SE = 0.084, p < 0.001). EEG-based image-level associations were weaker. Foreground water and forest-based context were associated with lower β/α values in exploratory models, but the matched water-versus-road EEG contrast was not significant after correction. Discussion Validated AI-generated rural images can serve as controlled stimuli in environmental psychology research. The findings suggest that rural landscape viewing may promote stronger physiological, EEG-based, and affective recovery than passive blank-screen viewing, while visual preference and EEG-based restoration may follow partly distinct depth-layered pathways. Given the exploratory design, the depth-layer findings should be interpreted cautiously and require confirmation in more systematically balanced stimulus sets.