Productivity and safety are in a trade-off relationship, and the improvement of the safety management system of construction sites is a pressing issue. Therefore, it is important to know and analyze information about incidents at real construction sites. However, it is difficult to gather information about these incidents from workers' self-reports. Therefore, in this research, we take an approach to view the workers as the sensors distributed in construction site and detect these incidents with the reaction of the workers. Biological signals such as heart rate, sweating, and muscle activity are the signals generally used to detect an emotional reaction; however, requiring workers to attach electrodes to their body during work is not suitable. Thus, we focused on blinks since they can be detected without electrodes attached to the skin. This study aims to investigate changes that occur in human blinks during an incident at a construction site. For safety purposes, this study used VR technology to simulate an incident at a construction site. During the simulation, an image of the subject's eyes was taken by the camera installed in the head-mounted display. The results of this study suggest that humans who face an incident have lower blink rates because they gaze at the cause of the incident.