Neural model for processing the influence of visual orientation on visually perceived eye level (VPEL)

Abstract

An individual line or a combination of lines viewed in darkness has a large influence on the elevation to which an observer sets a target so that it is perceived to lie at eye level (VPEL). These influences are systematically related to the orientation of pitched-from-vertical lines on pitched plane(s) and to the lengths of the lines, as well as to the orientations of lines of ‘equivalent pitch’ that lie on frontoparallel planes. A three-stage model processes the visual influence: The first stage parallel processes the orientations of the lines utilizing 2 classes of orientation-sensitive neural units in each hemisphere, with the two classes sensitive to opposing ranges of orientations; the signal delivered by each class is of opposite sign in the two hemispheres. The second stage generates the total visual influence from the parallel combination of inputs delivered by the 4 groups of the first stage, and a third stage combines the total visual influence from the second stage with signals from the body-referenced mechanism that contains information about the position and orientation of the eyes, head, and body. The circuit equation describing the combined influence of n separate inputs from stage 1 on the output of the stage 2 integrating neuron is derived for n stimulus lines which possess any combination of orientations and lengths; Each of the n lines is assumed to stimulate one of the groups of orientation-sensitive units in visual cortex (stage 1) whose signals converge on to a dendrite of the integrating neuron (stage 2), and to produce changes in postsynaptic membrane conductance ( g i) and potential ( V i) there. The net current from the n dendrites results in a voltage change ( V A) at the initial segment of the axon of the integrating neuron. Nerve impulse frequency proportional to this voltage change signals the total visual influence on perceived elevation of the visual field. The circuit equation corresponding to the total visual influence for n equal length inducing lines is V A=Σ V i/[ n+( g A/ g S)], where the potential change due to line i, V i, is proportional to line orientation, g A is the conductance at the axon's summing point, and g S= g i for each i for the equal length case; the net conductance change due to a line is proportional to the line's length. The circuit equation is interpreted as a basis for quantitative predictions from the model that can be compared to psychophysical measurements of the elevation of VPEL. The interpretation provides the predicted relation for the visual influence on VPEL, V, by n inducing lines each with length l: thus, V= a +[ k iΣ θ i/ n+ ( k 2/ l )] , where θ i is the orientation of line i, a is the effect of the body-referenced mechanism, and k 1 and k 2 are constants. The model's output is fitted to the results of five sets of experiments in which the elevation of VPEL measured with a small target in the median plane is systematically influenced by distantly located 1-line or 2-line inducing stimuli varying in orientation and length and viewed in otherwise total darkness with gaze restricted to the median plane; each line is located at either 25° eccentricity to the left or right of the median plane. The model predicts the negatively accelerated growth of VPEL with line length for each orientation and the change of slope constant of the linear combination rule among lines from 1.00 (linear summation; short lines) to 0.61 (near-averaging; long lines). Fits to the data are obtained over a range of orientations from −30° to +30° of pitch for 1-line visual fields from lengths of 3° to 64°, for parallel 2-line visual fields over the same range of lengths and orientations, for short and long 2-line combinations in which each of the two members may have any orientation (parallel or nonparallel pairs), and for the well-illuminated and fully structured pitchroom. In addition, similar experiments with 2-line stimuli of equivalent pitch in the frontoparallel plane were also fitted to the model. The model accounts for more than 98% of the variance of the results in each case.