Perception is influenced by predictions about the sensory environment. These predictions are informed by past experience and can be shaped by exposure to recurring patterns of sensory stimulation. Predictions can enhance perception of a predicted stimulus, but they can also suppress it by favoring novel and unexpected sensory information that is inconsistent with the predictions. Here we investigate the role of prediction on visual perception by measuring how different types of predictions affect perceptual selection. We measure perceptual selection using binocular rivalry, a bistable phenomenon in which incompatible images presented separately to the two eyes result in perceptual alternation between the two images over time.

In one study, we show that by creating predictive priors through repeated exposure to arbitrary sequence of orientation gratings, subjects were more likely to perceive the grating that matched the orientation that was consistent with the predictive context. That is, observers were more likely to see what they expected to see, compared to the likelihood of perception of the unexpected stimulus.

In another study, we examine how other studies in the literature have reported the opposite effect of prediction on visual perceptual selection when the predictive prior is created with natural image sequences. We investigate if a relationship exists between these opposing effects of prediction. We measure this by studying effects of prediction from orientation gratings generated in our first study compared with predictions from natural images within the same individuals. We found no relationship between these two types of prediction. One possible explanation of these results is that the weighting of visual predictions differs between stimuli of different complexity. Unfortunately, although we found no correlation between these two types of predictions, we also did not replicate the effects of prediction seen previously for either type of stimuli.

Finally, we describe how binocular rivalry can be adapted for experiments outside of the lab that will be used to investigate how the constraints on conscious perception guided by visual priors may be relaxed under the influence of psychedelics in a clinical setting.

Anatomical, physiological, and psychophysical approaches have revealed a great deal about the structure and function of the visual system. Non-invasive techniques based on nuclear magnetic resonance supplement these approaches by allowing the possibility of recording from the awake, behaving human brain. Here, we present two experiments using different non-invasive techniques to answer questions about the neural basis of human vision. In Chapter 2, we examine the relationship between γ-aminobutyric acid (GABA), perceptual suppression, and amblyopia (“lazy eye”). In amblyopia, abnormal visual experience during development leads to an enduring loss of visual acuity in adulthood. Physiological studies in animal models suggest that intracortical GABAergic inhibition may mediate visual deficits in amblyopia. To better understand the relationship between visual cortical GABA and perceptual suppression in persons with amblyopia (PWA), we employed magnetic resonance spectroscopy (MRS) to quantify GABA levels in both PWA and normally-sighted persons (NSP). In the same individuals, we obtained psychophysical measures of perceptual suppression for a variety of ocular configurations. In PWA, we found a robust negative correlation between the depth of amblyopia (the difference in visual acuity between the amblyopic and non-amblyopic eyes) and GABA concentration that was specific to visual cortex and was not observed in a sensorimotor cortical control region. Moreover, lower levels of visual cortical GABA were associated with weaker perceptual suppression of the fellow eye by the amblyopic eye and stronger suppression of the amblyopic eye by the fellow eye. Taken together, our findings provide evidence that intracortical GABAergic inhibition is an important component of the pathology of human amblyopia and suggest possible therapeutic interventions to restore vision in the amblyopic eye through enhancement of visual cortical GABAergic signaling in PWA. In Chapter 3, we use functional magnetic resonance imaging (fMRI) to identify cortical and subcortical components of the visual system. In particular, we localized many topographically-organized visual cortical areas as well as the lateral geniculate nucleus (LGN) and the magnocellular and parvocellular subdivisions thereof. We report the extension of this M/P localization procedure to fMRI data acquired with multiband acceleration. We also examined thalamocortical coupling between the LGN subdivisions and visual cortical areas but found no effects of stimulus type or spatial attention.

Perception consists of the brain's single best interpretation of the sensory world at a given moment in time. Multiple channels of visual input - be they from the two eyes or from the many parallel visual pathways that originate as early as the retina - must be reconciled to arrive at a unified percept. The fact that this must occur in roughly real time as the visual scene changes poses special challenges and constraints. I investigated two classes of visual processes relevant for the perception of time-varying visual stimuli: prediction, with a probable neural substrate in early visual cortical areas, and parallel processing in the magnocellular (M) and parvocellular (P) pathways.

In Chapters 2 and 3, I asked how prediction and parallel pathways, respectively, contribute to perceptual selection using dynamic binocular rivalry stimuli. In binocular rivalry, incompatible images presented to the two eyes result in just one of the images being selected for awareness at any given time. This bistability makes rivalry a useful tool for the study of perceptual selection. In Chapter 2, we found that predictive context in the form of an unambiguous rotating grating biased perceptual selection during subsequent rivalry towards the expected next grating in the rotation sequence, compared to an orthogonal grating. This provided evidence that a prediction-like process influences perceptual selection during rivalry between gratings, which other work has shown is likely resolved at early stages of visual processing.

In Chapter 3, we studied spatial, temporal, luminance, and chromatic factors influencing perceptual selection during interocular switch rivalry. In this type of rivalry, flickering orthogonal gratings are periodically exchanged between the two eyes, resulting in either the perception of a fast, regular alternation between orthogonally oriented gratings (similar to the display presented to a single eye) or a slow, irregular alternation, a percept that requires integration across the two eyes over time. We found that stimuli biased toward the M pathway increased the prevalence of fast, regular alternations, while stimuli biased toward the P pathway increased the prevalence of slow, irregular alternations. This finding suggested that the M and P pathways can make distinct contributions to perception during binocular rivalry and led us to propose a new framework for understanding perceptual selection during interocular switch rivalry.

Physiological measurement of activity in the M and P pathways can lead to greater understanding of how these pathways contribute to perceptual experience, but methods for measuring functional signals from the M and P pathways of humans have been lacking. Therefore, in Chapter 4, we developed a procedure for functionally mapping the M and P subdivisions of human LGN, the site where these pathways are most clearly segregated, using functional magnetic resonance imaging (fMRI). We observed a gradient of more M-like to more P-like responses across the LGN. Importantly, this gradient had a spatial layout consistent with known LGN anatomical organization. This new method for localizing the M and P subdivisions of the LGN provides a way forward for investigating the function of these pathways in human visual perception, in both healthy and clinical populations.

In summary, prediction and parallel processing are two classes of mechanisms that contribute to perception of dynamic visual stimuli. Here we have shown how such mechanisms operating at low levels of the visual system can help resolve competition between percepts, thereby affecting the contents of visual awareness. In addition, we developed a method for the physiological study of the M and P LGN subdivisions in the human brain, which is a promising technique for the future investigation of the roles of the M and P pathways in human visual perception, among other applications.

Improvement of cognitive function is of great value to many aspects of society. However, identifying robust procedures for training cognitive processes with generalizable benefits remains elusive. Here we present a novel attention regulation training paradigm that incorporates skill application in multiple learning environments. We hypothesized that our training procedure would enhance goal-oriented cognitive function and cognitive control mechanisms. We evaluated training effects using both computerized assessments and self-report questionnaires designed to probe cognitive control in everyday life. We tested for specificity of training effects by employing multiple active control conditions and for generalizability by using assessments that were significantly different from the training tasks as well as a set of secondary self-report questionnaires to explore the extent of far transfer. Training substantially improved performance on multiple tasks that involve goal-oriented cognitive functioning and cognitive control, and these gains were often specific to the group that received attention regulation training and also applied learned skills in varied environments.

Attention is a well-studied and complex topic that covers many fields of research. Effects of attention are ubiquitous throughout the brain and can be differentiated among sensory modalities (e.g., visual vs. auditory), volition (e.g., exogenous vs. endogenous), and application (e.g., covert vs. overt). Over the past few decades, researchers have proposed many computational models of visual attention, and with the rise of machine learning tools, more have been proposed to solve computer vision problems as well. In this dissertation, I will focus on a specific subset of models that place visual attention within a probabilistic inference framework in which humans utilize attention to infer the current state of the world from noisy sensory information. Across three experiments, I propose and evaluate computational models of visual attention that address endogenous spatial attention, feature and spatial attention during covert visual search, and bottom-up and top-down attention during free viewing of natural images. Each model builds upon the previous one in an effort to understand the influence of common principles across different tasks and applications.

In the first experiment, I propose a computational model of spatial attention that uses a dynamic pooling mechanism to simulate receptive field changes that have been observed in neurophysiological studies of endogenous spatial attention. The model can be viewed as a spatial prior over a region of the visual field that reduces uncertainty in visual processing by enhancing the local spatial resolution. By reproducing well-characterized perceptual phenomena observed in visual crowding literature, we conclude that reduction in the spatial uncertainty of encoded feature representations relieves crowding. This decrease in uncertainty influences crowding mainly by increasing the redundancy of encoded representations, with effects on fidelity playing a more limited role. In the second experiment, I extend this model by incorporating spatial attention into a hierarchical generative model to simulate a covert visual search task for digits among non-digit distractors. The generative model learns top-down priors over digit features, and these priors disambiguate among low-level target and distractor features during search to highlight regions that are likely to contain the target. By spatially attending predicted target locations that were generated with or without the use of top-down priors, we show a benefit of using top-down priors on downstream target classification accuracy that is greater than the improvement from spatial attention alone. Finally, in the third experiment, I introduce a model of bottom-up and top-down attention at multiple levels of feature complexity and spatial scale to account for gaze behavior in a free-viewing experiment across many categories of natural images. As an extension of the second experiment, priors in this experiment influence bottom-up as well as top-down attention. In this case, bottom-up attention is measured as the surprise relative to priors (in an information-theoretic sense) when viewing a scene. By learning priors within as well as across categories, the results demonstrate that surprise from category-specific priors over high-level features best accounted for gaze behavior across the majority of scene types.