Perceptual Learning of Image Texture: A Neuroscience Approach using Isotrigon Textures

Abstract

This thesis outlines the current state of the use of artificial image-textures in psychophysical studies, and new directions in perceptual learning of isotrigon image texture. This thesis uses new and innovative methods, such as Amazon Mechanical Turk (an online testing platform) for data collection. This thesis also utilises Principal Component Analysis to explore whether learning to discriminate isotrigon textures heavily loaded (weighted) on to one factor transfers to learning of textures loaded onto a different factor. Because factor loadings can represent channels, this will give us insight into the pathways that may be used in learning image texture. In Chapter 2 this thesis explores whether isotrigon image textures can be learned rapidly, in multiple learning sessions over the course of a single day. This work was based on a previous study by our lab, which measured image texture learning over a longer time-span, of 6 weeks, where significant learning was observed. That learning met some of the criterial for Perceptual Learning. Work done in this thesis indicates that the participants learned discriminate the textures over time with practice at the task during the course of single day. This suggests that our ability to discriminate between different texture types may not be innate after all, and may be learned, possibly by recruiting a greater number of innate channels to improve discrimination performance and speed. In Chapter 3 we conducted a follow up study focusing on the mechanisms underlying texture learning. If each factor represents a single mechanism, then learning one texture that has a strong loading for a given factor should also improve performance for another texture strongly loaded onto those mechanisms. This should be the case, even if participants have never seen any of the other texture type. This thesis suggests that a relatively small number of mechanisms may be responsible for learning complex image structure. In Chapter 4 this thesis examines the perception of image textures in Parkinson's Disease. The implication of visual processing in Parkinson's disease is explored. This thesis shows that isotrigon texture discrimination is impaired in Parkinson's Disease, and puts forward several key texture types as being worthy of future study. In Chapter 5, this thesis further explores location based perceptual learning using isotrigon textures. Specifically, we study whether participants can learn the difference between odd and even textures, and whether this learning is location-dependent in the visual field. That is, whether learning to perceive an image texture type at a specific location in the visual field transfers to other regions of the visual field. Location-dependency is a hallmark of perceptual learning of difficult tasks. This phenomena was observed in our study. We also examined the retention of what was learned, with lengthy retention being typical of perceptual learning, which was also observed here and in Chapter 2. This thesis adds to the literature by examining the learning of image texture, an area of little study. It is also innovative as it focuses on perceptual learning, especially new methods for accessing that in Visual Area 2 by using isotrigon image textures, which is a first. Future research directions include: investigation of the mechanisms behind perceptual learning using a larger data set to drill down into how the time-course affects learning; and the use of isotrigon texture discrimination as part of a diagnostic test for presence of neurological diseases; and an investigation into the limits of perceptual learning of isotrigon image textures. Show more