Quantitative Determination and Analysis of Visual Receptive Fields.

A unique approach to obtaining nonbiased quantitative information on visual receptive fields is examined. This approach involves placing the animal in a feedback and control loop, and inverting the pattern recognition process. In a preliminary application, receptive fields are determined via a random scanning technique. The system moves predetermined stimuli around on a cathode ray tube viewed by the animal, while recording extracellularly from retinal ganglion fibers in the tectum. Digitized response waveforms are saved and analyzed by extensive software to characterize the receptive field. In the primary application, the receptive field is determined by a stochastic technique using response feedback. In iterative fashion, initially random patterns are modified by the algorithm enabling the animal to generate visual patterns optimizing some measure of the response. Characteristics of the receptive field as determined by the two different techniques are contrasted in many runs, and are compared to classically determined receptive field characteristics. The detailed nature of the responses to stimuli is examined. The problem of optimizing the second technique of receptive field determination by response feedback is explored via simulation. The response feedback technique is shown to converge successfully on receptive field patterns in actual physiological runs. The capability of the technique to provide unique results is discussed. In addition to providing a check on the response feedback technique, the results of the random scan technique demonstrate the separability of independent units on the basis of pulse height. Also, many units were found to undergo rhythmic bursting.