Item

Abstract

Reviews data from physiological studies that give some indication of what temporal cortical neurons might represent. If learning indeed modifies the properties of these neurons, then the snapshots of a cell's life provided by these standard single-unit studies can tell us something about what is learned. The authors present evidence that perceptual learning may provide a useful framework for understanding inferotemporal (IT) neurons. They argue that when novel patterns of sufficient complexity are regularly encountered, long-term changes occur in the connectivity of IT cells that implicitly incorporate these experiences. The authors do not suggest that the activity of these specialized cells ever replaces the activity of cells in earlier visual areas, but rather that it offers compact and reliable representations that can be used for rapid perceptual analysis and that can be associated with both actions and other neuronal representations. More important, the authors think that this experience-dependent modification is not a special process that is explicitly turned on and off, but is instead always operational during active (and perhaps even passive) visual processing. Thus object learning and object recognition may not be so easily distinguished.