Supplementary Figure 3: Comparison between the gravity and gravity-derivative responses during yaw and pitch/roll rotations

(a,b) Response of the example cell in Fig. 1 during pitch and roll rotations. This experiment was repeated in n=48 cells. (c-f) Population analysis. A linear regression (see Methods) was used to compute the cells’ response gain and preferred direction (PD) (e.g. NU, LED, ND…) during tilted yaw rotation (±30° tilt, protocol 2-3). An identical regression analysis was performed on pitch/roll rotation (protocol 4) responses. Thus, we obtained two independent estimates of the cell’s gravity gain and PD, one based on yaw rotation responses and one based on pitch/roll rotation responses. We then compared the gain ratio and difference in PD on a cell-by-cell basis (G-tuned cells: green; dG-tuned cells: cyan; G+dG-tuned cells: gray). (c) The G gain ratios were not significantly different from 1 (bootstrap-based statistics: mean ratio = 0.9, 0.7 to 1.1 CI, n = 34), and (d) PDs were predominantly aligned (mean ΔPD = 10°, -6 to 29° CI). (e) The dG gain ratios were significantly larger than one (mean ratio = 1.6, CI = 1.3 to 2, n=22), likely due to nonlinearities (dG stimuli had a higher amplitude during pitch/roll than during yaw rotation; see on-line Methods). (f) ΔPDs of the dG component were not different from zero (mean ΔPD = 9°, -26 to 8° CI). The similarity in tuning during tilted yaw and pitch/roll movements and independence on the plane/axis of rotation provides strong support that the observed direction tuning arises from G and dG (rather than angular velocity)-coding.