Recognition of rotated objects and cognitive offloading in dogs _ Lonardo et al.

We trained pet dogs on a visual discrimination task between two geometrical shapes presented on a touchscreen. We subsequently tested the dogs on the rotated (±45°, ±90°, ±135°) versions of the same stimuli. We analysed dogs' accuracy and head tilts both at the group and at the individual level. During the recognition test, all dogs performed significantly above chance level with upright stimuli, consistently with the strict learning criteria they had to meet before entering the test phase. Females were more accurate than males. With rotated stimuli, on average dogs' accuracy dropped to chance level and we did not find evidence for a human-like mental rotation effect of the angle of stimulus rotation on performance. Indeed, dogs' accuracy was not modulated by the angular disparity between stimulus and reference, nor by the direction of stimulus rotation (clockwise or counter clockwise). With rotated stimuli, the female advantage disappeared and we found no sex difference in the accuracy.
At the individual level, only 3 (38%) out of 8 dogs performed differently from chance level with rotated stimuli. Of these, one male and one female dog performed above chance level and one female dog performed significantly below chance level. We interpret these data as evidence that dogs use individual strategies to solve the mental rotation task and that some of them were likely focusing their attention on local parts of the stimuli rather than considering the stimuli globally.
In our controlled setting, in which dogs had to lay their head on a chinrest before observing the stimuli, we found that dogs only rarely exhibited head tilting. Rather, they made small adjustments of their head position on the chinrest following the stimuli onset (average amplitude of head tilt observed after stimulus onset: ca. 5°; range: 0°- ca. 26°). Although dogs' head tilts after the stimulus onset followed the same direction as the stimulus rotation and were wider for wider stimulus rotations, there was no evidence that tilting their head influenced dogs' accuracy, probably due to the challenging task and relatively small sample size. We conclude that these data do not support the hypothesis that dogs, similarly to humans, try to offload the demands of the mental rotation task by using their bodies. This work paves the way to the study of cognitive offloading in non-human animals.

本研究对宠物犬进行视觉辨别训练，使其在触摸屏上区分两种几何形状。随后，我们对犬只进行了旋转（±45°，±90°，±135°）版本相同刺激物的测试。在分析中，我们从群体和个体层面考察了犬只的准确性及头部倾斜情况。在识别测试过程中，所有犬只均显著高于随机水平，符合其在进入测试阶段前必须严格遵循的学习标准。雌犬的准确性高于雄犬。面对旋转刺激，犬只的平均准确性降至随机水平，且未发现刺激旋转角度对表现产生类似人类心理旋转效应的证据。实际上，犬只的准确性并未受到刺激与参考点之间角度差异的调节，亦未受到刺激旋转方向（顺时针或逆时针）的影响。在旋转刺激下，雌犬的优势消失，且未发现性别在准确性方面存在差异。在个体层面，仅有3只（占总数的38%）犬只在旋转刺激下表现出与随机水平不同的行为。其中，一只雄犬和一只雌犬的表现高于随机水平，而另一只雌犬的表现则显著低于随机水平。我们解读这些数据为犬只使用个体策略解决心理旋转任务的证据，其中一些犬只可能更专注于刺激物的局部而非全局考虑。在我们的控制环境中，犬只在观察刺激物前需将头部置于下巴托上，我们发现犬只很少出现头部倾斜现象。相反，它们会在刺激物出现后对下巴托上的头部位置进行微小调整（刺激物出现后平均头部倾斜幅度约为5°；范围：0°-约26°）。尽管犬只在刺激物出现后的头部倾斜方向与刺激物旋转方向一致，且在更大幅度的刺激旋转中更为明显，但并未发现头部倾斜对犬只的准确性有影响，这可能是由于任务难度较大以及样本量相对较小所致。我们得出结论，这些数据不支持犬只与人类类似，试图通过身体动作来减轻心理旋转任务的要求。本研究为非人类动物认知卸载的研究开辟了道路。