Where objects come from: Attention, segmentation, and textons Academic Article uri icon

abstract

  • It is well established that the units of attention are not merely spatial but closely relate to perceptual objects. While much work in the field of object-based attention (OBA) is aimed at finding these units, their relationship to the basic features (textons) that guide the segmentation of visual stimuli into 'objects' is largely unexplored. Here we bridge this gap for one of the most conspicuous features of early vision, namely orientation. Much work in the segmentation literature suggests that orientation-based texture segmentation (OBTS) is guided by orientation gradients, and our previous work also suggests that it is also significantly mediated by texture flow curvatures. In addition, this work makes the surprising prediction that the flow of attention should *not* depend on the general direction (i.e. the 'grain') of the texture - in contradiction to previous findings in the OBA literature (Avrahami, 1999). To address this contradiction and to reveal the relationship between attention, objects, OBTS, and the orientation texton, we employ both the cueing and divided-attention paradigms on various orientation-defined textures (ODTs), both uniform (one 'object') and discontinuous (two 'objects'). Contrary to previous studies, we find that the texture's 'main direction' has no effect: attention flows just as readily with vs. against the 'grain' of ODTs. At the same time, texture-defined discontinuities have a major effect: attention flows less readily across texture-defined boundaries which are defined by both orientation and curvature. These effects replicated across multiple paradigms and dependent measures, and also held for jittered ODTs, wherein the effects must be due to the global structure and not local good continuation. We conclude that uniform ODTs are single objects from an attentional point of view, while discontinuous ODTs are processed as multiple objects. This work reveals how the 'objects' of OBA are formed from simple visual features.

publication date

  • January 1, 2010