Justus Liebig University, Giessen, Germany.
Skidmore College, NY, USA.
Cognition. 2019 Aug;189:167-180. doi: 10.1016/j.cognition.2019.04.006. Epub 2019 Apr 12.
Shape-deforming processes (e.g., squashing, bending, twisting) can radically alter objects' shapes. After such a transformation, some features are due to the object's original form, while others are due to the transformation, yet it is challenging to separate the two. We tested whether observers can distinguish the causal origin of different features, teasing apart the characteristics of the original shape from those imposed by transformations, a process we call 'shape scission'. Using computer graphics, we created 8 unfamiliar objects and subjected each to 8 transformations (e.g., "twisted", "inflated", "melted"). One group of participants named transformations consistently. A second group arranged cards depicting the objects into classes according to either (i) the original shape or (ii) the type of transformation. They could do this almost perfectly, suggesting that they readily distinguish the causal origin of shape features. Another group used a digital painting interface to indicate which locations on the objects appeared transformed, with responses suggesting they can localise features caused by transformations. Finally, we parametrically varied the magnitude of the transformations, and asked another group to rate the degree of transformation. Ratings correlated strongly with transformation magnitude with a tendency to overestimate small magnitudes. Responses were predicted by both the magnitude and area affected by the transformation. Together, the findings suggest that observers can scission object shapes into original shape and transformation features and access the resulting representational layers at will.
变形过程(例如挤压、弯曲、扭曲)可以彻底改变物体的形状。在这种转变之后,一些特征是由于物体的原始形状,而另一些特征则是由于变形,但很难将两者区分开来。我们测试了观察者是否可以区分不同特征的因果起源,将原始形状的特征与变形施加的特征分开,我们称之为“形状分裂”。我们使用计算机图形学创建了 8 个不熟悉的物体,并对每个物体进行了 8 种变形(例如“扭曲”、“膨胀”、“熔化”)。一组参与者始终如一地命名变形。第二组参与者根据原始形状或变形类型将卡片上描绘的物体排列成类。他们几乎可以完美地做到这一点,这表明他们可以轻易区分形状特征的因果起源。另一组使用数字绘画界面来指示物体上哪些位置发生了变形,他们的反应表明他们可以定位由变形引起的特征。最后,我们参数化地改变了变形的幅度,并要求另一组对变形的程度进行评分。评分与变形幅度强烈相关,并且有低估小幅度的趋势。响应由变形的幅度和受影响的面积共同预测。总之,这些发现表明,观察者可以将物体形状分裂为原始形状和变形特征,并根据需要访问由此产生的表示层。