Singh Manish
Department of Psychology, Rutgers University-New Brunswick, Piscataway, NJ 08854, USA.
Psychol Sci. 2004 Jul;15(7):454-9. doi: 10.1111/j.0956-7976.2004.00701.x.
Mechanisms of contour completion are critical for computing visual surface structure in the face of occlusion. Theories of visual completion posit that mechanisms of contour interpolation operate independently of whether the completion is modal or amodal--thereby generating identical shapes in the two cases. This identity hypothesis was tested in two experiments using a configuration of two overlapping objects and a modified Kanizsa configuration. Participants adjusted the shape of a comparison display in order to match the shape of perceived interpolated contours in a standard completion display. Results revealed large and systematic shape differences between modal and amodal contours in both configurations. Participants perceived amodal (i.e., partly occluded) contours to be systematically more angular--that is, closer to a corner--than corresponding modal (i.e., illusory) contours. The results falsify the identity hypothesis in its current form: Corresponding modal and amodal contours can have different shapes, and, therefore, mechanisms of contour interpolation cannot be independent of completion type.
在面对遮挡时,轮廓完成机制对于计算视觉表面结构至关重要。视觉完成理论假定,轮廓插值机制的运行与完成是模态还是非模态无关,从而在两种情况下生成相同的形状。在两个实验中,使用两个重叠物体的配置和修改后的卡尼兹配置对这一同一性假设进行了检验。参与者调整比较显示的形状,以匹配标准完成显示中感知到的插值轮廓的形状。结果显示,在两种配置中,模态和非模态轮廓之间都存在巨大且系统的形状差异。参与者认为非模态(即部分被遮挡)轮廓在系统上比相应的模态(即虚幻)轮廓更有棱角——也就是说,更接近一个角。结果证伪了当前形式的同一性假设:相应的模态和非模态轮廓可以有不同的形状,因此,轮廓插值机制不能独立于完成类型。