Abstract
Introduction: People coordinate their eye, head, and body movements to gather information from a dynamic environment while maximizing reward and minimizing biomechanical and energetic costs. Such natural behavior is not possible in a laboratory setting where the head and body are usually restrained and the tasks and stimuli used often lack ecological validity. Therefore, it’s unclear to what extent the principles of fixation selection derived from lab studies, such as inhibition-of-return (IOR), apply in a real-world setting. Results: Participants performed nine real-world tasks, including driving, working at a desk, grocery shopping, and building a lego set (N recordings, 169). We measured fixations using a Tobii Pro Glasses 2 mobile eye tracker. We found that spatial and temporal IOR were absent in all tasks. Instead, participants most often re-fixated on what they had just viewed. Saccade latencies were shorter preceding re-fixations than forward saccades. We propose that participants orient their heads toward objects of interest and then minimize eccentric fixations in order to conserve oculomotor effort. Correspondingly, we observed center biases in the distributions of fixation location and duration (relative to the head orientation). A model that generates scanpaths by randomly sampling these distributions reproduced the spatial and temporal return phenomena seen in the data. The model also explained the existence of distinct 3-fixation sequences of durations for forward and return saccades. The amount of the orbit used in a given task traded off with fixation duration, consistent with an oculomotor cost that interacts with task demands. Conclusion: In naturalistic settings, people orient their heads toward objects of interest and then repeatedly return their gaze to those objects, fixating longer than at other locations. Conservation of oculomotor effort (“laziness”) explains these behaviors parsimoniously, demonstrating that oculomotor costs fundamentally shape how people extract and act on relevant visual information from the environment.