Abstract
Humans execute multiple saccades every second to sample information from the environment, creating complex scanpaths across many objects and locations. There is divergent evidence as to whether humans are aware of the locations and objects they have fixated across these sequences of saccades. While previous research explicitly asked participants to report the locations or objects of their fixations, we approached the problem from a different perspective by measuring perceptual confidence, to gauge instead whether people know how much information they have about objects in the world. If people retain a representation of how much information they have acquired about objects across fixations, there should be a link between fixations, perceptual confidence and perceptual performance.
Participants viewed an array of five real-world, everyday objects that were presented at any random angle from 360º of possible viewpoints. After 1500ms, participants were asked to choose two objects to make a perceptual report on: this choice was a proxy measure for perceptual confidence. They then made a perceptual report by rotating a presented object to match the viewpoint they remembered. This report object could be chosen or non-chosen, with a higher probability of presenting a chosen object.
Participants were more likely to choose objects they had fixated, and they reported the orientation of these objects more accurately. This demonstrates they had more perceptual confidence for items they had more precise information about. We also calculated information uptake for chosen vs non-chosen items, assuming a Gaussian window of information uptake around fixations; participants relied primarily on foveal vision, using a window with 1.6 degrees SD to guide choices.
Overall this study suggests that humans retain an accurate representation about the precision of information gained during fixations. Knowing how much information you have about objects in the world may be more useful than remembering exact scanpaths.