Abstract
Information processing in the human brain is highly parallel. One prediction of this parallel processing is that features of different objects may be combined erroneously. Using a backward masking paradigm, we showed recently that features presented at different retinotopic locations can be combined. (Otto, Ogmen, & Herzog, 2006, Journal of Vision, 6(10), 1079–1086). Here, we show that such feature “mis-bindings” occur not only across retinotopic maps but also across orientation maps. We presented a central Vernier that was offset either to the left or right. On either side of this Vernier, a sequence of flanking lines devoid of any Vernier offset followed. Flanking lines were successively shifted further away in space and successively rotated by a slight orientation difference. A motion percept of two orbital streams of lines was elicited whereas the central Vernier itself was largely rendered invisible. However, the Vernier offset was perceived in the stream of flanking lines. If one of the flanking lines was offset itself, this offset was integrated with the offset of the central Vernier even when the orientation difference between the central Vernier and the flanking line was more than 30°. Hence, features presented at substantially different locations and orientations can be bound together. We suggest that these feature “mis-bindings” are not errors of the visual system but part of a computational strategy.
Thomas Otto was supported by Swiss National Science Foundation (SNF)