Perceptual systems often use multiple cues to encode one dimension of the stimulus. However, because many of those cues either use different inputs or different transformations, multiple cues are usually encoded separately, i.e. in different streams or modules. The separation is evident in different sensory modalities, but it is also seen within sensory modalities in the form of different processing streams. This segregation of processing then raises the question of how the results of these different cues are brought back together. The present experiments are focused on the binding of visual edges that may be defined by different attributes. The minimum distance between two edges required to see them as separate was measured, as a function of different pairings of attribute-defined edges (such as luminance, colour, texture, flicker), and as a function of retinal eccentricity (Chapter 3). Because no systematic effects remained once the effects of the single attributes were removed from performance in integration conditions (i.e. when the two edges are defined by different attributes), it is unlikely that integration occurs via an integration-specific module. In Chapters 4 to 6, orientation after-effects were measured for global patterns composed of local edges defined by alternating attributes, for example, a near-vertical grating could be composed of horizontal rows of colour-defined edges interleaved with rows of motion-defined edges. Conditions included several combinations of luminance-, colour-, texture and motion-defined edges, or blank interleaves. Results show that transfer occurs even when adaptation and test attribute pairs are different, suggesting that by default recalibration of orientation mechanisms spread to affect all attributes involved in orientation discrimination, including those not present in the adaptation stimuli. In Chapter 7, some participants could simultaneously maintain two orientation after-effects in different directions when defined by different attribute pairs (for example: left oblique luminance-colour, and right oblique texture-motion). Stimuli were designed to be ambiguous in orientation to mechanisms encoding single-attributes edges or attribute-independent edges, so any after-effect seen would require the work of integration mechanisms. Thus, integration appears to occur in parallel, within the attribute modules. A framework is presented that may resolve many discrepancies across integration research fields.
展开▼
