Ralph G. Hale

Disentangling boundary extension and normalization of view memory for scenes

ABSTRACT Boundary extension (BE) is a memory error for close-up views of scenes in which participants tend to remember a picture of a scene as depicting a more wide-angle view than what was actually displayed. However, some experiments have yielded data that indicate a normalized memory of the views depicted in a set of scenes, suggesting that memory for the previously studied scenes has become drawn toward the average view in the image set. In previous studies, normalization is only found when the retention interval is very long or when the stimuli no longer appear to represent a spatial expanse. In Experiment 1, we examine whether normalization can influence results for scenes depicting a partial view of space and when the memory test occurs immediately following the study block by manipulating the degree of difference between studied close-up and wide-angle scenes. In Experiment 2, normalization is induced in a set of scenes by creating conditions expected to lead to memory interference, suggesting that this may be the cause of view normalization. Based on the multi-source model of BE, these scenes should be extended during perception (Intraub, H. (2010). Rethinking scene perception: A multisource model. Psychology of Learning and Motivation, 52, 231–265). In Experiment 3, we show that BE is indeed observable if the same scenes are tested differently, supporting the notion that BE is primarily a perceptual phenomenon while normalization is a memory effect.

Contrast sensitivity indicates processing level of visual illusions.

A nearly linear contrast response function (CRF) is found in the lower level striate cortex whereas a steep, nonlinear increase at lower contrasts that gradually increases toward response saturation for higher contrasts is found in the higher level extrastriate cortex. This change of CRFs along the ventral cortical pathway indicates a shift from stimulus- and energy-dependent coding at lower levels to percept- and information-dependent coding at higher levels. The increase of nonlinearity at higher levels optimizes the extraction of perceptual information by amplifying responses to the ubiquitous low-contrast inputs in the environment. We used this difference of CRFs between lower and higher levels, particularly at lower contrasts (.0 to .30), as a tool to investigate examples of 2 lower level (simultaneous brightness and simultaneous tilt) and 2 higher level (Poggendorff and Ponzo) illusions. As predicted, the Poggendorff and Ponzo illusions yielded strong nonlinear increases in their CRFs compared to the more linear functions found for the simultaneous-brightness and simultaneous-tilt illusions. We conclude that the Poggendorff−Ponzo illusions rely more heavily on high-level, percept-dependent cortical processing than do the simultaneous-brightness−simultaneous-tilt illusions and, more generally, that differences between contrast-dependent changes may be a useful tool in determining the relative level of cortical processing of many other visual effects.

Using object color diagnosticity to influence access to semantic information in a boundary extension paradigm.

Individuals consistently remember seeing a more wide-angle version of a previously viewed scene than actually existed. The multi-source model of boundary extension (BE) (Intraub, 2010) suggests many sources of information contribute to this visual memory error (e.g., amodal perception at the view boundaries, semantic information about the scene). The color diagnosticity of an objects color is known to affect object recognition with poorer recognition for atypically vs. typically colored objects. If atypically colored objects lead to poorer recognition (i.e., reduced availability of semantic information) then, according to the multi-source model, a less precise initial encoding should lead to greater BE. Scenes classified as having high or low color diagnostic objects were used as stimuli. Low color diagnostic stimuli (i.e., color non-diagnostic) and two versions of high color diagnostic scenes were tested. Typical and atypical versions of the high color diagnostic scenes were made by changing the central objects color only. Scenes were presented for either 46 or 250 ms followed by a mask, and then immediately presented again for test. Observers first identified the central object, then gave a BE rating. As expected, recognition was significantly poorer for atypically colored vs. typically colored and color non-diagnostic scenes at the 46 ms duration, while there were no differences at 250 ms. BE occurred for all conditions. Most importantly, the BE was greatest at 46 ms for scenes with atypically colored objects. This finding suggests the reduced availability of semantic information for this condition lead to a less precise memory representation leading to increased boundary extension. At 250 ms where recognition was similar, semantic information would have been equally available leading to consistent BE ratings across the three types of colored scenes. These findings provide further insight into the role of object recognition and semantic information on boundary extension. Meeting abstract presented at VSS 2015.