Mechanisms of visual feature binding

Abstract

Visual feature binding is the method by which coherent objects and scenes are perceived. Advances in the science of perception have indicated that visual features such as colour, motion, and orientation are to some extent, processed separately in primate early visual cortex. However, the mechanism by which these features are integrated remains unclear. Phenomenologically, the process of binding features to form objects appears to be an efficient and automatic process. Some research also shows a high temporal resolution for binding features together, in addition to populations of neurons that jointly code for features. However, dominant theories of feature binding and the majority of the binding literature indicate that the feature binding process is severely limited by a relatively low temporal resolution, especially when compared to other perceptual properties such as feature detection. To identify and resolve the discrepancy in the feature binding literature, I investigate the feature binding process and its inter-relationship with perceptual surface segregation. Surface segregation has been postulated as the method by which features can be rapidly bound together, giving them impression of a high temporal resolution. In Chapter 2, displays are used that alternate between two arrays of differently coloured, oppositely moving dots. The alternation frequency is modified in order to gauge the temporal resolution of binding. This is combined with surface segregation cues such as coherent motion, consistency of dot configuration, and colour. In Chapter 3, coloured, oriented gratings are used to investigate colour-orientation binding. Angular separation, spatial and temporal coincidence, and stimulus presentation duration are varied. Across these experiments, a number of these surface segregation cues are manipulated in order to measure the corresponding effects on feature binding, perceptual interpretation of the stimulus, and its neural representation. The results of the psychophysical experiments indicate that feature binding, surface segregation, and temporal integration are inextricably linked. These findings are reinforced by data gathered through functional magnetic resonance imaging (fMRI) of human subjects. Both surface segregation and feature pairs were found to modulate neural activity in early visual cortex, providing evidence that similar neural substrates are recruited for both feature binding and surface segregation. Overall, the two complementary sets of experiments using stimulus conjunctions of colour-motion and colour-orientation stimuli provide converging evidence and insight into the dynamics of the underlying binding mechanisms. A discussion of the implications of the research follows, concluding that rapidly formed surface representations can be maintained across presentation intervals by temporal integration. Attentional selection of one feature (e.g. orientation) can then be used to boost the response to the paired feature (colour) in order to identify and extract the correct feature pairing. Based on the known properties of the visual system, several potential neural mechanisms are proposed that are consistent with both the psychophysical and neural data, in addition to suggested future directions for the study of visual feature binding.