Abstract
Perceptual stability is a critical problem for the visual system because visual information that reaches the eyes is often noisy and discontinuous over time due to changes in illumination or viewpoint, among others. A proposed solution to this problem is the continuity field, a mechanism that promotes perceptual stability by generating serially dependent appearance – making similar objects nearby in space and time appear more alike than they actually are (Fischer and Whitney, 2014; Cicchini, et al., 2014; Liberman et al., 2014). The visual system seems to therefore sacrifice accuracy for the sake of stability; however, such a mechanism could quickly reduce accuracy to a level that is maladaptive. In the current study, we ask if the continuity field can accommodate the competing goals of visual accuracy and stability, depending on the demands of a particular task. To test this, we measured the strength of serial dependence when subjects performed a task with a high working memory load. Subjects viewed a sequence of random target faces drawn from a continuous distribution of morphed faces. After each target face, subjects were cued to adjust a random test face to match the identity of the target face they just saw or the target face on the previous trial (high working memory load). In another condition, subjects were cued to match only the most recent face (low working memory load). We found that the strength of serial dependence was diminished in the trials with a high working memory load compared to the trials with a lighter load. When the task required working memory for past objects, the continuity field adjusted to reduce serially dependent perception. This result suggests that the continuity field does not depend on working memory and can balance the need to represent images accurately with the need to stabilize perception.
Meeting abstract presented at VSS 2016