Abstract
Human gaze behavior provides rich insights into the innermost states of our minds, reflecting perception, attention, and even intentions. Saliency maps are models that predict fixation locations, created to generalize across a wide range of visual scenes but validated using only few participants. Generalisations across individuals are generally implied. Here, this implied generalisation across people, not images, was focused on by assessing gaze behavior of 8,325 participants, including children. Using a single, yet feature-rich image, shortcomings were found in the ability to predict over this diverse sample in each of 21 models. Strikingly, models performed well on participants aged 18-35, but poorly for other age groups, with children in particular. Modelling and understanding gaze behavior beyond college-student like samples thus requires an approach which incorporates knowledge on differences in gaze behavior across the lifespan or at least validation on (age)-diverse samples.