During everyday scene viewing, the information received by the visual system is frequently disrupted: Objects are often occluded by other objects, and sensory processing is interrupted by eye, head, or body movements. The visual system is extremely efficient at correcting for these interruptions and in establishing object correspondence and perceptual continuity. At the end of this correspondence process, the visual system is left with two representations of an object: The initial representation and the one acquired after the disruption. In the present dissertation, I investigated the mechanisms by which the visual system reconciles these discontinuous inputs to give us a perception of a smooth and stable visual world.
To achieve this, I ran four experiments in which participants were presented with a colored saccade target, and instructed to remember its color before executing the saccade. On some trials, the color of the saccade target was changed to a new value during the saccade. Participants were asked to report either the pre- or post-saccadic color value in a continuous report task. Object continuity was manipulated in two ways. The target blanking paradigm served as the main manipulation of stability: On half the trials, the target was removed from the screen during the saccade, disrupting object continuity. In addition, the magnitude of color change was used as a secondary manipulation of visual stability. The color report data were fit with probabilistic mixture models. First, there was no evidence for integration of pre- and post-saccadic feature values into a composite representation. Instead, on a majority of trials participants could successfully retain and report both pre- and post-saccadic states of the target object. Further, these two states dynamically interacted with each other, resulting in their feature values systematically shifting toward each other. Lastly, when reporting the pre-saccadic color, participants were more likely to incorrectly report the post-saccadic color under conditions of visual stability versus instability, supporting a probabilistic overwriting mechanism. Together, these results are more consistent with an object-based model, rather than an image-based model of representational updating. Although the present study only focused on transsaccadic updating mechanisms, similar mechanisms are likely to be functional in many common situations where the visual system needs to establish perceptual continuity across disruptions and changes.