It has been evidenced that in attention-window tasks, the participants fixate on the center of a screen while inspecting two stimuli that appear at the same time in parafoveal vision. Such tasks have successfully been used to estimate a person’s breadth of attention under various conditions. While behavioral investigations of visual attention have often made use of response accuracy, recent research has shown that the pupil size can also be used to track shifts of attention to the periphery. The main finding of previous studies is that the harder the evaluation of the stimuli becomes, e.g., because they appear farther away from the central fixation point, the stronger the pupils dilate. In this paper, we present experimental data suggesting that in an attention-window task, the pupil size can also be used to assess whether the participants attend to static, non-moving, or dynamic, moving stimuli. That is, regression models containing information on presentation mode (static vs. dynamic) and the visual angle between spatially separated stimuli better predict accuracy of perception and pupil dilation than model without these sources of information. This finding is useful for researchers who aim at understanding the human attentional system, including potential differences in its sensitivity to static and dynamic objects.