The Danger of Interpreting Detection Differences between Image Categories
The Danger of Interpreting Detection Differences between Image Categories
复制标题
解释图像类别之间的检测差异的危险
DOI：
10.31234/osf.io/pvmex
复制
发表时间：
2018
影响因子：
3.1
通讯作者：
M. Peelen
中科院分区：
文献类型：
--
作者：
Surya Gayet;T. Stein;M. Peelen
Gomes和他的同事（2018年）使用中断连续闪光抑制（b-CFS;一种感知抑制技术）表明，人类观察者在检测蛇的图像（构成进化上古老的威胁）方面比鸟类更有优势。在他们的研究中，蛇和鸟的图像被过滤，以包含粗尺度或细粒度的信息。蛇的优先检测依赖于粗尺度（而不是细粒度）的信息，这被认为是支持存在一个进化上古老的皮层下通路专门用于蛇检测。在这里，我们提出的关注，蛇和鸟类的图像固有的视觉特征不同，这可能会强烈影响检测时间在b-CFS。例如，蛇的图像比鸟的图像具有更大的周长与表面的比率。重要的是，这些视觉特征不是蛇特有的，因为它们与许多非威胁性物体类别共享。为了说明这一点，我们比较了自行车和汽车图像之间的检测时间;非威胁性图像类别的视觉特征不同，但检测不太可能利用进化上古老的专用皮层下通路。观察者在探测自行车方面比探测汽车更有优势。模仿Gomes等人报道的蛇鸟差异，这一优势是由图像中的粗尺度（而不是细粒度）信息驱动的。因此，两个非威胁性的，语义匹配的刺激类别之间的视觉特征的差异足以产生与蛇与鸟观察到的完全相同的发现模式。我们的结论是空间频率特异性检测的差异，在b-CFS不能明确地归因于加工途径的差异。
Using breaking continuous flash suppression (b-CFS; a perceptual suppression technique), Gomes and colleagues (2018) showed that human observers have an advantage in detecting images of snakes (constituting an evolutionarily old threat) over birds. In their study, images of snakes and birds were filtered to contain either coarse-scale or fine-grained information. The preferential detection of snakes relied on coarse-scale (rather than fine-grained) information, which was taken as support for the existence of an evolutionarily old subcortical pathway dedicated to snake detection. Here, we raise the concern that images of snakes and birds inherently differ in their visual characteristics, which can strongly affect detection times in b-CFS. Images of snakes, for instance, have a larger perimeter-to-surface ratio than images of birds. Importantly, these visual characteristics are not snake-specific, as they are shared with many non-threatening object categories. To illustrate this point, we compared detection times between images of bicycles and cars; non-threatening image categories that differ in visual characteristics but for which detection is unlikely to capitalize on an evolutionarily old dedicated subcortical pathway. Observers exhibited an advantage for detecting bicycles over cars. Mirroring the snake-bird differences reported in Gomes et al., this advantage was driven by the coarse-scale (rather than fine-grained) information in the images. Hence, differences in visual characteristics between two non-threatening, semantically matched stimulus categories suffice to produce the exact same pattern of findings as observed with snakes versus birds. We conclude that spatial frequency-specific detection differences in b-CFS cannot be unequivocally attributed to differences in processing pathways.