Extended Data Fig. 1: Global optic-flow patterns and H2 neuron results.

Related to Fig. 1. a, Left: two optic-flow fields on a hemisphere (representing the fly’s right eye) induced by yaw (red) and roll (blue) rotations. Right: flow fields induced by thrust (red) and lift (blue) translations. The local red and blue flow vectors are orthogonal near the centre of the eye, but form skewed yet different angles near the boundary. b, EM reconstruction (in the FAFB⁶ volume) and light-microscopy image of a single H2 neuron (morphology consistent across 3 independent images of this driver line). c, H2 responses to dark and bright moving edges along 16 directions and to moving gratings along eight directions, at location 4 in Fig. 1f. d, Local H2 PDs for 7 individual flies measured in response to moving bright and dark edges at 6 retinal locations (left, individual cell responses to all 16 directions of bright and dark moving edges in Supplementary Data 1), and mean local H2 PDs for moving grating stimuli at 9 retinal locations (stimulus protocols detailed in Methods). Locations 2–6 correspond to those in Fig. 1f and e, and include 7 flies from experiment set #2 (8 directions); locations 7–9 include 5 flies from experiment set #1 (16 directions); location 1 includes flies from both experiment set #1 and experiment set #2. The retinal positions are combined across flies based on our procedure for aligning each fly’s head (uncertainty in the stimulus positions across all cells and positions, standard deviation <3°). Inset: key for plotting conventions of spatial data. e, Angular distribution of H2 local PDs with respect to the forward direction at six visual locations, represented as median +/- quartiles, from recordings of 7 cells at position 1, 2, and 6, and 6 cells at position 3, 4, and 5. Colour encodes stimulus types. Two-way ANONA without interaction found a statistically significant difference in the average angle by location (F(5) = 3.43, p < 0.01) but not by stimulus type (F(2) = 2.63, p = 0.08). Combining dark and bright moving edge data, one-sample Two-sided Wilcoxon test found significant differences (from 0) at locations 2 (p < 0.013), 3 (p < 0.0034), and 4 (p < 0.034). Three data points outside the plotting range are not shown. f, We simulate how 8 points regularly spaced on a circle, representing the distance travelled by a moving edge stimulus, will appear to the fly at 6 locations. The apparent expansion of these points (e.g., location 4) is due to the Mercator projection. Comparing distant positions, the vectors at position 4 (average ~11.6°) are ~7% smaller than those at position 1 (average ~12.7°). The angular distortions are accounted for in determining the local PDs of H2, but we do not correct the minimal changes in the speed of the stimulus. During the experiment, the display is repositioned, which further minimizes the distortions of the stimuli at higher elevations (see Methods).