Previous investigations of visual object recognition have found that some views of both familiar and unfamiliar objects promote more efficient recognition performance than other views. These views are considered as canonical and are often the views that present the most information about an object's 3-D structure and features in the image. Although objects can also be efficiently recognised with touch alone, little is known whether some views promote more efficient recognition than others. This may seem unlikely, given that the object structure and features are readily available to the hand during object exploration. We conducted two experiments to investigate whether canonical views existed in haptic object recognition. In the first, participants were required to position each object in a way that would present the best view for learning the object with touch alone. We found a large degree of consistency of viewpoint position across participants for both familiar and unfamiliar objects. In a second experiment, we found that these consistent, or canonical, views promoted better haptic recognition performance than other random views of the objects. Interestingly, these haptic canonical views were not necessarily the same as the canonical views normally found in visual perception. Nevertheless, our findings provide support for the idea that both the visual and the tactile systems are functionally equivalent in terms of how objects are represented in memory and subsequently recognised.
Generalization between canonical and non-canonical views in object recognition