This study examined attention allocation in 30 marksmen categorized into 3 skill levels ranging from expert to novice. Each shooter performed 336 shooting trials. Half of the trials were performed under an occluded-vision condition and the rest under regular, unoccluded conditions. Immediately after completion of a random subset of shots (96 trials), shooters estimated the actual location of each shot, and on a random subset of trials (48 trials), shooters gave retrospective verbal reports. A mixed 3 × 2 factorial analysis of variance revealed that the expert marksmen performed and estimated their shots more accurately than the intermediate and novice marksmen, the intermediates performed like the experts under the full-vision condition and like novices under the occluded-vision condition, and the experts reported attending more to nonvisual information while they estimated their shots than did the novices. The findings advance our understanding of the mechanisms mediating expertise.