Abstract
A surprising amount of information can be gathered in just a single glance at a scene. In a few hundred milliseconds, we can gather the gist of a scene and obtain critical information about the environment. How do expectations about objects, and prior knowledge about their typical locations, sizes, or materials affect this rapid information acquisition? Here we investigate eye movements in an experiment in which we manipulate observers’ expectation about the object’s material properties. Five naïve participants’ gaze was tracked while they performed Two-Interval-Forced-Choice (2IFC) tasks comparing the stiffness and (in a separate block) the lightness of gelatinous blobs. The stimuli were 3D, computer-rendered animations showing wobbling, diffusely-reflecting objects hanging from the ceiling. Both the stiffness and lightness of the objects varied in 5 perceptually distinct steps. We trained a linear classifier on landing positions to differentiate between the two perceptual tasks. Results indicate that the first fixation for both tasks is at the center of the object. However, by the 4th fixation, the landing position can be used to reliably classify the task (1st: 50.57%, 2nd: 60.11%, 3rd: 67.73%, 4th: 74.94%, 5th: 69.77% classification accuracy with chance performance at 50%). For lightness judgments, fixations tended to be near the brightest region on the object, while for stiffness judgments, fixations tended to gravitate towards regions of higher motion energy. Our results show that observers’ expectations enable them to adopt an optimal sampling strategy. Ultimately, identifying where people fixate on an object will allow us to understand which cues are relevant in the perception of material qualities in dynamic scenes.
Acknowledgement: This research was supported by a Sofja Kovalevskaja Award from the Alexander von Humboldt Foundation, sponsored by the German Federal Ministry for Education and Research.