Long Bria, Störmer Viola S, Alvarez George A
Department of Psychology, Harvard University, Cambridge, MA,
Department of Psychology, University of California, San Diego, CA, USA.
J Vis. 2017 Jun 1;17(6):20. doi: 10.1167/17.6.20.
While substantial work has focused on how the visual system achieves basic-level recognition, less work has asked about how it supports large-scale distinctions between objects, such as animacy and real-world size. Previous work has shown that these dimensions are reflected in our neural object representations (Konkle & Caramazza, 2013), and that objects of different real-world sizes have different mid-level perceptual features (Long, Konkle, Cohen, & Alvarez, 2016). Here, we test the hypothesis that animates and manmade objects also differ in mid-level perceptual features. To do so, we generated synthetic images of animals and objects that preserve some texture and form information ("texforms"), but are not identifiable at the basic level. We used visual search efficiency as an index of perceptual similarity, as search is slower when targets are perceptually similar to distractors. Across three experiments, we find that observers can find animals faster among objects than among other animals, and vice versa, and that these results hold when stimuli are reduced to unrecognizable texforms. Electrophysiological evidence revealed that this mixed-animacy search advantage emerges during early stages of target individuation, and not during later stages associated with semantic processing. Lastly, we find that perceived curvature explains part of the mixed-animacy search advantage and that observers use perceived curvature to classify texforms as animate/inanimate. Taken together, these findings suggest that mid-level perceptual features, including curvature, contain cues to whether an object may be animate versus manmade. We propose that the visual system capitalizes on these early cues to facilitate object detection, recognition, and classification.
虽然大量研究聚焦于视觉系统如何实现基本层面的识别，但对于它如何支持物体之间的大规模区分，如区分有生命物体和实际大小等方面的研究较少。先前的研究表明，这些维度反映在我们的神经物体表征中（康克莱和卡拉马扎，2013），并且不同实际大小的物体具有不同的中级感知特征（朗、康克莱、科恩和阿尔瓦雷斯，2016）。在此，我们检验这样一个假设：有生命物体和人造物体在中级感知特征上也存在差异。为此，我们生成了动物和物体的合成图像，这些图像保留了一些纹理和形状信息（“纹理形状”），但在基本层面上无法识别。我们将视觉搜索效率作为感知相似性的指标，因为当目标在感知上与干扰物相似时，搜索速度会变慢。在三个实验中，我们发现观察者在物体中找到动物比在其他动物中更快，反之亦然，并且当刺激物简化为无法识别的纹理形状时，这些结果依然成立。电生理证据表明，这种混合生命性的搜索优势出现在目标个体化的早期阶段，而不是在与语义处理相关的后期阶段。最后，我们发现感知曲率解释了部分混合生命性的搜索优势，并且观察者使用感知曲率将纹理形状分类为有生命/无生命。综上所述，这些发现表明，包括曲率在内的中级感知特征包含了物体是有生命还是人造的线索。我们提出，视觉系统利用这些早期线索来促进物体的检测、识别和分类。
