Abstract
In our work with late-onset vision patients in Project Prakash, our results suggest that, immediately following sight onset, subjects exhibit an overly fragmented interpretation of their visual world. It is tempting to infer that this hyper-fragmentation is at least partially responsible for their deficits in object recognition.
To test the hypothesis that hyper-fragmentation is a debilitating condition for object processing, we presented shapes made up of dots, parametrically varying the resolution of the shape contour and embedded in noise. Normally sighted subjects were excellent at finding the shape when allowed to view the entire figure, as expected, under all parameters tested. However, when viewed through a series of apertures (mimicking a hyper-fragmented view), performance dropped precipitously.
This experiment leaves open at least one confounding interpretation. Normal vision is not usually performed through an aperture, so the aperture may be particularly disruptive in vision. Haptic exploration, on the other hand, is routinely performed through the “aperture” of the finger tips. Normal subjects who performed a haptic version of the same task exhibited poor performance as expected. Since haptic object identification may be unnatural for these subjects, we also tested completely blind subjects, for whom haptic exploration is the norm. Even these subjects exhibited extremely poor performance at low resolutions. Both normal and blind subjects, however, were excellent at identification of high resolution shapes in the haptic condition as well as the visual aperture condition.
We conclude, therefore, that the ability to identify an object depends not so much on the simultaneous perception of all components of the shape, but rather on the presence of cues indicating the interconnectivity of the fragments and on the ability to parse these cues.