In forest harvesting, operators must visually monitor the terrain, machinery, the stand and the trees they are cutting in order to plan, evaluate and adjust their tasks. To exploit increasing opportunities to automate these tasks and create decision support systems it is essential to understand not only what forestry workers do, but also what they look at and why they focus on specific aspects during specific tasks. This knowledge may also aid operator training and knowledge transfer between age and experience groups. Eye-tracking (ET) is therefore a potentially valuable technique that may facilitate both extraction of implicit knowledge and elucidation of operators' information acquisition strategies. However, real world ET-recordings are sensitive to environmental variations and analyzing them is time consuming. Thus, the aims of this study were to examine the utility of a head-mounted eye-tracking system in forest harvesting machines in a natural setting and obtain information on operators' visual behavior (gaze patterns) during harvesting. The output from the eye-tracker was affected by large head movements, changes in illumination and (possibly) vibrations. The gaze pattern analysis revealed that the operators looked at the harvester head or forest most of the time, but their gaze behaviors varied during different harvesting operations. They looked at the monitor, canopy and falling trees less frequently during first thinning than during second thinning and final felling. The results suggest that some harvesting information is gathered in advance to get an overview and plan the work, but most eye movements closely follow actions.