Abstract
A saccade brings a retinal locus to a target in the visual field. For normally sighted individuals, this retinal locus is the fovea. Central field loss (CFL) caused by macular degeneration often leads to the adoption of a preferred retinal locus (PRL) in the peripheral retina for saccades and fixation. Factors underlying the development of a PRL are not known. Here we show that a conceptually simple computational model can account for the formation of a PRL and its idiosyncrasies. We assume that the visual system always intends to aim the retinal locus with the highest expected post-saccade acuity at the saccade target. The expected post-saccade acuity of a retinal locus is a function of the physiological acuity at and around the locus and the expected saccade error. The expected saccade error is a combination of motor error that is proportional to the saccade amplitude (vector error) and the spatial uncertainties associated with the retinal locus and the saccade target (endpoint errors). We assume that the motor error does not improve, but the spatial uncertainties associated with the neural representation of the endpoints are optimally re-estimated after each saccade from the observed saccade error. A generic forgetting function is assumed to prevent spatial uncertainty from vanishing. Simulations showed that immediately after CFL, the utilized retinal loci are close to the edge of the scotoma on the side nearest to saccade targets. After each saccade, spatial uncertainties associated with the pre-saccade target and utilized retinal locus decrease. Decrease in spatial uncertainty increases the expected post-saccade acuity of the retinal locus. The net effect is that a previously selected retinal locus is more likely to be selected for a future saccade, further reducing its spatial uncertainty, and forming the PRL. Idiosyncrasies at the early stages of CFL strongly influence PRL formation.
Meeting abstract presented at VSS 2015