NEURONAL PROCESSING OF NATURAL IMAGES

在提议的研究中，我们计划采用一种相对较新的

In the proposed research, we plan to employ a relatively new experimental paradigm-termed "free viewing"-in which a monkey freely views a collection of natural images, to investigate the firing patterns of pairs of simultaneously recorded neurons in striate cortex under conditions approximating those occurring naturally. Our goal is to obtain answers to three categories of questions. First, are the receptive field properties of V1 neurons, measured during free viewing, consistent with the receptive field properties of the same cells measured using classical techniques? An answer to this question will help us to determine the extent to which classically measured receptive field properties can account for the behavior of primary visual cortex under natural conditions. Second, is the activity of V1 neurons during free viewing consistent with the predictions of a hypothesis for sparse coding of natural images? In this framework, an efficient code for natural scenes is achieved by representing visual images with the smallest number of simultaneously active cells. The free viewing paradigm will provide an opportunity to test several specific predictions of this hypothesis. Third, do simultaneously recorded neurons in V1 exhibit synchronous responses to natural image features that are consistent with a hypothesis for perceptual grouping? A large body of evidence has demonstrated that cortical neurons engage in stimulus dependent synchronous firing. The proposed experiment will determine if, and in response to what stimulus features, synchronous activity occurs during the free viewing of natural images. To answer these questions, we will conduct two types of measurements from pairs of single units in area V1 of awake behaving macaques monkeys. First, while the animals perform a simple fixation task, we will quantitatively map each cell's receptive field properties using i) drifting sinewave gratings, and ii) the methods of reverse correlation analysis. Second, while the animals perform a match-to-sample task, in which they visually scan a set of natural images to identify a cued object, we will measure the joint activity of pairs of neurons in response to these stimuli. These data will allow us to address the questions posed above, and thus improve our understanding of the behavior of primary visual cortex under natural conditions.