LOCALIZING THE SPATIAL LOCALIZATION SYSTEM: Helmholtz or Gibson?

McCloskey and a team of authors (1995) documented a fascinating deficit involving the localization of objects. When subject A.H. reaches for objects placed in different positions, she is often wrong about direction but not amount. An object located 30° to her left, say, might cause her to reach 30° to the right. The authors demonstrated that her motor abilities are flawless and that the errors are genuinely in vision. But precisely where in the visual localization process does the deficit occur? 1 suggest that an analysis of how objects are normally localized may point to the source of error. I begin with a description of two separate visual localization systems, and then discuss how the nature of the deficit seems to implicate one of these systems over the other. Both localization systems can deal with the inherent ambiguity of a single retinal location. The ambiguity is easily appreciated by considering a simple example in which the image of an object is centered on the fovea. This retinal image can be caused by an object that is straight ahead of the person's nose, but only if the person is fixating the object. If the fixation is to the left, then the same retinal image will be produced by an object to the person's left. That is, one retinal image position could be produced by any position in the world, and one position in the world can produce an image anywhere on the retina. The first proposed solution to this problem is one of Helmholtz's unconscious inference models (1866/1962). We owe the retinal ambiguity to the fact that we are movable observers; our eyes can move, our head can move, and so forth. Consequently, if we include information about the current state of the body in our visual calculations, we can uniquely recover the position of the object that produced the ambiguous retinal image. According to this model, then, the ambiguity is resolved by considering the extraretinal information about the position of the eyes in the head along with the retinal information. If head position is held constant, the simple cancellation method (eye position retinal position) will reconstruct the correct position of the object in the world. The second model is best known from the work of Gibson (1959). Gibson's general solution to ambiguity was to look over a wider area of the retina, rather than to