非表示:
キーワード:
-
要旨:
A common denominator in all vision tasks is scene segmentation: what is fore- and background, which visual components belong to the same or different entities? In prior studies we used a bi-stable stimulus that can either be perceived as separate local components or as a global Gestalt. fMRI and TMS showed that posterior parietal cortex (PPC) was selectively and causally involved in global Gestalt perception (Grassi et al., 2016; Zaretskaya et al., 2013).
Here we employed three additional such local vs. global bi-stable stimuli. Importantly, we found that apart from the classification of the two possible percepts into local (ungrouped, component) versus global (grouped, Gestalt), the two possible perceptual interpretations can alternatively be classified according to a second dimension: the complexity or sophistication of the interpretation, i.e. default (simple) versus non-default (complex, high-level, sophisticated). As these two dimensions overlapped differentially across the four stimulus classes (for two stimuli, global coincided with complex, in another two with simple), we were able to identify whether parietal cortex involvement reflected grouping, or the complexity of the perceptual interpretation.
We found that the involvement of parietal cortex reflected the level of sophistication of the visual interpretation rather than grouping into a single Gestalt. For all four stimuli, we found activity pattern that was highly similar for the contrast of default (simple) vs. non-default (complex, sophisticated) perceptual interpretations. It consistently and prominently involved posterior parietal cortex. Also consistent with previous findings, we found for all stimuli strong early visual cortex deactivations during sophisticated perceptual interpretations. Mid-level regions such as LOC or motion regions were differentially involved with each stimulus class and percept-type. Our results lead us to suggest that PPC is not necessarily involved in mere grouping toward global Gestalt, but instead more generally it is involved in generating the more complex, high-level, or more sophisticated perceptual interpretation of a given stimulus. The activation of high-level dorsal areas (PPC) and the concurrent deactivation of early visual areas during high-level perceptual interpretations is in line with predictions from generative models of visual perception, also known as predictive coding theory, but not with attention. Our findings suggest a generic mechanism for scene segmentation with the PPC as its anatomical substrate.
