Attractive and repulsive serial biases in visual cognition

Abstract

Our eyes are the primary gates to the visible world around us. Yet, whatever we perceive involves a translation of the patterns of the photons hitting our retinas into mental representations. It could be argued that visual perception is merely an interpretation of the physical world, which raises the question of how reliable this translator is. Fischer and Whitney (2014) found that the perception of an oriented Gabor patch was influenced by the orientation of previously presented Gabor patches, a phenomenon that they called serial dependence. They argued that serial dependence reflects the visual system's assumptions about continuity: The visual world around us is generally constant from moment to moment and does not suddenly change despite large changes in viewpoint, occlusion or lighting conditions. Our perceptual systems employ this predictability to overcome potential perceptual noise and maintain perceptual stability. Subsequent research has revealed that the perception of a variety of other features, including shape, motion coherence, numerosity, facial identity, and even stimulus ensembles, is systematically biased toward recent information. Almost all of the studies mentioned above involved paradigms where the stimuli causing the serial dependence (the inducers) were attended, and therefore do not address whether items that we ignore can cause serial dependence biases. In the papers in this thesis, we studied the role of the to-be-ignored items in forming biases in our perceptual decisions. In Paper I here, we used a visual search for an oddly oriented line among distractors to demonstrate that the to-be-ignored items can form serial dependence in perceptual decisions in addition to the attended inducer. Notably, the repulsive bias occurs even when the distractors and targets are remarkably dissimilar, which distinguishes this from the well-known tilt illusion and tilt adaptation. Furthermore, our results demonstrate that explicit reports of stimulus features are not required for serial dependence. The findings in Paper I suggest that perception accounts for both attended and ignored stimuli in preserving the visual world's continuity to a greater extent than previously thought. In Paper II we demonstrated that visual search can result in serial-dependence biases in the perceived orientation of a stimulus that is unrelated to the search task. Our findings also revealed that both attention and similarity between the search stimuli (distractors and targets) and the test item play a significant role in forming serial dependence in perceptual decisions. Finally, our findings in Paper III demonstrated that similar biases to those reported in Papers I and II can occur from a single inducer line upon a set of test lines as previously observed from a set of lines upon a single test line. Secondly Paper III shows that when the inducer is similar to the test, it produces an attractive bias, but it creates a repulsive bias when these items are dissimilar. This means that the inducer creates opposing biases based on the similarity of the perceptual history content to the current stimulus. These attractive and repulsive biases can occur simultaneously when more than one test item is present. Overall, our results show that attention and proximity in feature space play a crucial role in shaping serial dependence biases. We showed that biases introduced by attended versus ignored visual search items (targets and distractors) influenced the general perceptual decisions, and furthermore, we showed that even attended items can produce repulsive biases in perceptual decisions.