Fig. 2: Geometric relationship between 3D eye orientation and spatial models.
From: Integration of landmark and saccade target signals in macaque frontal cortex visual responses
a Scatter plot of the initial 3D eye orientation (in space) at the fixation (F) position. Data from every trial used in the analysis of FEF (top panel) and SEF (lower panel) neurons are shown for both monkeys. Note that in natural head-unrestrained conditions, eye-in-space torsion (vertical axis) is much more variable than in head-restrained conditions, causing the retina to rotate relative to space. b Schematic of the different egocentric coordinate systems available in this task. Curving arrows indicate rotation about the torsional axes of the eye and head. Note that in head-unrestrained conditions, eye orientation in space is determined by both the eye position relative to head, and the head relative to body. c Schematic of how eye torsion in space influences projection of the screen stimuli onto the retina for an example Target-Landmark Configuration (TLC2) relative to fixation (F). The top row represents the vectors between these stimuli (TF, TL, LF) as they appear on the screen, the middle row shows a schematic of two initial eye orientations (primary position vs. torsionally tilted), and the bottom row shows the resulting retinal projections of those vectors in eye coordinates [TF(e), TL(e), LF(e)], but without optic reversal, e.g., vectors on the left correspond to leftward stimuli. At primary (straight ahead) eye position, the retinal projections are relatively simple, but torsion results in both tilts and distortions of various stimulus vectors, depending on the different T-L-F configurations. These configurations could help remember where targets are, and the systematic distortions could help decode their locations in space in the presence of a faulty or noisy internal estimate of eye torsion in space. (Note that the amount of distortion has been slightly exaggerated in the figure for demonstrative purposes).
