Basic prediction mechanisms as a precursor for schizophrenia studies

McGruer, Fiona (2017) Basic prediction mechanisms as a precursor for schizophrenia studies. PhD thesis, University of Glasgow.

Abstract

Traditionally, early visual cortex (V1-3) was thought of as merely a relay centre for feedforward retinal input, providing entry to the cortical visual processing steam. However, in addition to feedforward retinal input, V1 receives a large amount of intracortical information through feedback and lateral connections. Human visual perception is constructed from combining feedforward inputs with these feedback and lateral contributions. Feedback connections allow the visual cortical response to feedforward information to be affected by expectation, knowledge, and context; even at the level of early visual cortex. In Chapter 1 we discuss the feedforward and feedback visual processing streams. We consider historical philosophical and scientific propositions about constructive vision. We introduce modern theories of constructive vision, which suggest that vision is an active process that aims to infer or predict the cause of sensory inputs. We discuss how V1 therefore represents not only retinal input but also high-level effects related to constructive predictive perception.
Visual illusions are a ‘side effect’ of constructive and inferential visual perception. For the vast majority of stimulus inputs, integration with context and knowledge facilitates clearer, more veridical perception. In illusion these constructive mechanisms produce incorrect percepts. Illusory effects can be observed in early visual cortex, even when there is no change in the feedforward visual input. We suggest that illusions therefore provide us with a tool to probe feedforward and feedback integration, as they exploit the difference between retinal stimulation and resulting perception. Thus, illusions allow us to see the changes in activation and perception induced only by feedback without changes in feedforward input. We discuss a few specific examples of illusion generation through feedback and the accompanying effects on V1 processing.

In Schizophrenia, the integration of feedback and feedforward information is thought to be dysfunctional, with unbalanced contributions of the two sources. This is evidenced by disrupted contextual binding in visual perception and corresponding deficits in contextual illusion perception. We propose that illusions can provide a window into constructive and inferential visual perception in Schizophrenia. Use of illusion paradigms could help elucidate the deficits existing within feedback and feedforward integration. If we can establish clear effects of illusory feedback to V1 in a typical population, we can apply this knowledge to clinical subjects to observe the differences in feedback and feedforward information.
Chapter 2 describes a behavioural study of the rubber hand illusion. We probe how multimodal illusory experience arises under varying reliabilities of visuotactile feedforward input. We recorded Likert ratings of illusion experience from subjects, after their hidden hand was stimulated either synchronously or asynchronously with a visible rubber hand (200, 300, 400, or 600ms visuotactile asynchronicity). We used two groups, assessed by a questionnaire measuring a subject’s risk of developing Schizophrenia - moderate/high scorers and a control group of zero-scorers. We therefore consider how schizotypal symptoms contribute to rubber hand illusory experience and interact with visuotactile reliability. Our results reveal that the impact of feedforward information on higher level illusory body schema is modulated by its reliability. Less reliable feedforward inputs (increasing asynchronicity) reduce illusion perception. Our data suggests that some illusions may not be affected on a spectrum of schizotypal traits but only in the full schizophrenic disorder, as we found no effect of group on illusion perception.

In Chapter 3 we present an fMRI investigation of the rubber hand illusion in typical participants. Cortical feedback allows information about other modalities and about cognitive states to be represented at the level of V1. Using a multimodal illusion, we investigated whether crossmodal and illusory states could be represented in early visual cortex in the absence of differential visual input. We found increased BOLD activity in motion area V5 and global V1 when the feedforward tactile information and the illusory outcome were incoherent (for example when the subject was experiencing the illusion during asynchronous stimulation). This is suggestive of increased predictive error, supporting predictive coding models of cognitive function. Additionally, we reveal that early visual cortex contains pattern representations specific to the illusory state, irrespective of tactile stimulation and under identical feedforward visual input.

In Chapter 4 we use the motion-induced blindness illusion to demonstrate that feedback modulates stimulus representations in V1 during illusory disappearance. We recorded fMRI data from subjects viewing a 2D cross array rotating around a central axis, passing over an oriented Gabor patch target (45°/ 135°). We attempted to decode the target orientation from V1 when the target was either visible or invisible to subjects. Target information could be decoded during target visibility but not during motion-induced blindness. This demonstrates that the target representation in V1 is distorted or destroyed when the target is perceptually invisible. This illusion therefore has effects not only at higher cortical levels, as previously shown, but also in early sensory areas. The representation of the stimulus in V1 is related to perceptual awareness. Importantly, Chapter 4 demonstrated that intracortical processing can disturb constant feedforward information and overwrite feedforward representations. We suggest that the distortion observed occurs through feedback from V5 about the cross array in motion, overwriting feedforward orientation information.

The flashed face distortion illusion is a relatively newly discovered illusion in which quickly presented faces become monstrously distorted. The neural underpinnings of the illusion remain unclear; however it has been hypothesised to be a face-specific effect. In Chapter 5 we challenged this account by exploiting two hallmarks of face-specific processing - the other-race effect and left visual field superiority. In two experiments, two ethnic groups of subjects viewed faces presented bilaterally in the visual periphery. We varied the race of the faces presented (same or different than subject), the visual field that the faces were presented in, and the duration of successive presentations (250, 500, 750 or 1000ms per face before replacement). We found that perceived distortion was not affected by stimulus race, visual field, or duration of successive presentations (measured by forced choice in experiment 1 and Likert scale in experiment 2). We therefore provide convincing evidence that FFD is not face-specific and instead suggest that it is an object-general effect created by comparisons between successive stimuli. These comparisons are underlined by a fed back higher level model which dictates that objects cannot immediately replace one another in the same retinotopic space without movement.

In Chapter 6 we unify these findings. We discuss how our data show fed back effects on perception to produce visual illusion; effects which cannot be explained through purely feedforward activity processing. We deliberate how lateral connections and attention effects may contribute to our results. We describe known neural mechanisms which allow for the integration of feedback and feedforward information. We discuss how this integration allows V1 to represent the content of visual awareness, including during some of the illusions presented in this thesis. We suggest that a unifying theory of brain computation, Predictive Coding, may explain why feedback exerts top-down effects on feedforward processing. Lastly we discuss how our findings, and others that demonstrate feedback and prediction effects, could help develop the study and understanding of schizophrenia, including our understanding of the underlying neurological pathologies.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Keywords: Predictive coding, schizophrenia, illusions, cortical feedback, motion induced blindness, rubber hand illusion, fmri, flashed face distortion.
• Subjects: B Philosophy. Psychology. Religion > BF Psychology; R Medicine > RC Internal medicine > RC0321 Neuroscience. Biological psychiatry. Neuropsychiatry
• Colleges/Schools: College of Science and Engineering > School of Psychology
• Funder's Name: Economic and Social Research Council (ESRC)
• Supervisor's Name: Muckli, Professor Lars
• Date of Award: 2017
• Depositing User: Ms Fiona McGruer
• Unique ID: glathesis:2017-8236
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 01 Jun 2017 08:27
• Last Modified: 18 Jul 2017 12:53
• URI: https://theses.gla.ac.uk/id/eprint/8236