Bias and sensitivity in the haptic perception of geometry

Our ability to recognize and manipulate objects relies on our haptic sense of the objects' geometry. But little is known about the acuity of haptic perception compared to other senses like sight and hearing. Here, we determined how accurately humans could sense various geometric features of objects across the workspace. Subjects gripped the handle of a robot arm which was programmed to keep the hand inside a planar region with straight or curved boundaries. With eyes closed, subjects moved the manipulandum along this virtual wall and judged its curvature or direction. We mapped their sensitivity in different parts of the workspace. We also tested subjects' ability to discriminate between boundaries with different degrees of curvature, to sense the rate of change of curvature, and to detect the elongation or flattening of ellipses. We found that subjects' estimates of the curvature of their hand path were close to veridical, and did not change across the workspace though they did vary somewhat with hand path direction. Subjects were less accurate at judging the direction of the hand path in an egocentric frame of reference, and were slightly poorer at discriminating between arcs of different curvature than at detecting absolute curvature. They also consistently mistook flattened ellipses and paths of decreasing curvature (inward spirals) for circles-and mistook arcs of true circles for arcs of tall ellipses or outward spirals. Nevertheless, the sensitivity of haptic perception compared well with that of spatial vision in other studies. Furthermore, subjects detected curvature and directional deviations much smaller than those that actually arise for most reaching movements. These findings suggest that our haptic sense is acute enough to guide and train motor systems and to form accurate representations of shapes.