Fixational eye movements and retinal adaptation: optimizing motion to maximize information acquisition.

Fixational eye movements (FEMs) are small, fluctuating eye motions when fixating on a target. A possible reason for FEMs is overcoming retinal adaptation (fading perception of fixed image). We present a model for FEM influence on learning about an external stimulus, incorporating temporal stimulus modulation, retinal image motion due to FEMs, blurring from optics and receptor size, uniform sampling by the receptor array, adaptation via a temporal filter, and added noise. We investigate the information transmitted, via: i) mutual information between visual system response and external stimulus, ii) direct estimation of stimulus from the system response, iii) contrast threshold for signal detection, and iv) determination of target location. We find a common quantity that must be maximized: the power transmitted due to temporal modulation and phase shifts from FEMs, passed through the temporal filter. We demonstrate there is an advantage from adding local persistence to a diffusive FEM. We quantify the contribution of FEMs to signal detection for targets of different size and duration, providing a qualitative account of human psychophysical performance. We show that optimal motion for determination of target location depends on the blur lengthscale and adaptation time.