Purchase this article with an account.
Rudiger von der Heydt, Fangtu T. Qiu; Figure-ground, Proto-objects, and selective attention: understanding the neural mechanisms. Journal of Vision 2007;7(9):346. doi: https://doi.org/10.1167/7.9.346.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
As the Gestalt psychologists noticed, attention interacts with figure-ground organization: figures draw attention, while shapes of the ground seem to be ignored. Mechanisms of figure-ground organization have recently been demonstrated in areas V1 and V2 of the visual cortex. However, their role in selective attention remains unclear. In area V2 of the monkey visual cortex many cells are selective for border ownership, responding to the same local edge more strongly when the edge is part of a figure on one side of the receptive field than the other. We found that a majority of these neurons were also influenced by attention. Tests in which the border between two figures was placed in the receptive field revealed asymmetrical attention effects, with enhancement of responses on one side or suppression on the other. The side of attention enhancement tended to coincide with the neuron's preferred side of border ownership. This indicates that the same neural circuits that create border ownership selectivity also provide the structure for selective attention. We propose a model in which ‘grouping cells’ integrate co-circular contours, and, via feedback, increase the gain of the corresponding contour neurons. This creates selectivity for side of figure, as observed in the border ownership tests. Supporting the feedback model, we found that border ownership signals show pronounced hysteresis (O'Herron and von der Heydt, this meeting). We explain the attention effects by assuming that top-down attention signals activate selected grouping cells, thus enhancing the responses of the corresponding contour neurons. Our results demonstrate an intermediate cortical stage in which features are grouped to larger entities (proto-objects), providing an interface between the initial local feature representation and later object-related stages of processing.
This PDF is available to Subscribers Only