Edge- and region-based processes of 2nd-order vision

Abstract

The human visual system is sensitive to 2nd-order image properties (often called texture properties). Spatial gradients in certain 2nd-order properties are edge-based, in that contours are effortlessly perceived through a rapid segmentation process. Others, however, are region-based, in that they require regional integration in order to be discriminated. The five studies reported in this thesis explore these mechanisms of 2nd-order vision, referred to respectively as segmentation and discrimination. Study one compares the segmentation and discrimination of 2nd-order stimuli and uses flicker-defined-form to demonstrate that the former may be subserved by phase-insensitive mechanisms. In study two, through testing of a neuropsychological patient, it is shown that 2nd-order segmentation is achieved relatively early in the visual system and, contrary to some claims, does not require the region termed human “V4”. Study three demonstrates, through selective adaptation aftereffects, that orientation variance (a 2nd-order regional property) is encoded by a dedicated mechanism tuned broadly to high and low variance and insensitive to low-level pattern information. Furthermore, the finding that the variance-specific aftereffect is limited to a retinotopic (not spatiotopic) reference frame, and that a neuropsychological patient with mid- to high-level visual cortical damage retains some sensitivity to variance, suggests that this regional property may be encoded at an earlier cortical site than previously assumed. Study four examines how cues from different 2nd-order channels are temporally integrated to allow cue-invariant segmentation. Results from testing a patient with bilateral lateral occipital damage and from selective visual field testing in normal observers suggest that this is achieved prior to the level of lateral occipital complex, but at least at the level of V2. The final study demonstrates that objects that are segmented rapidly by 2nd-order channels are processed at a sufficiently high cortical level as to allow object-based attention without those objects ever reaching awareness.