Identifying hazardous situations is a complex and multidimensional cognitive process that requires the proper allocation of workers' attention. Eye-Tracking technologies provide a viable option for studying construction workers' attentional allocation and for linking attention to their hazard-identification capabilities. The objective of the study is to use eye-movement measures to determine which types of hazards construction workers miss, ignore, or perceive to be insignificant. In order to achieve this goal, 31 construction workers participated in a controlled laboratory experiment in which they searched for hazards in images of 35 real construction-site scenarios while a head-mounted EyeLink II tracked their eye movements. The results showed differences in the participants' attentional distributions and that the hazard identification of workers with low and high hazard-identification skills stems from the types of hazard-not the number of hazards-within the scenarios. Further investigation on five images revealed that at-risk workers dwelt on imminent danger (e.g., workers in dangerous areas) rather than spreading their attentional efforts searching for sources of non-obvious hazards, including electrical hazards, housekeeping hazards and fall-protection-system-related hazards. The results of this experiment can thus support personalized safety training that targets at-risk workers.