Extended Data Fig. 4: Action fields are robust to stimulation length, saccade removal and reference frame.

(a-b) To confirm that the motor neuron-elicited convergence of head pose we observed is not an artefact of our 300 millisecond stimulation times, we plot just the first 80 milliseconds of the head trajectory when the angular velocity of the head is still increasing (Fig. 1h–i). (a) shows the head movements resulting from the first 80 milliseconds of CvN7 stimulation. (b) Shows the first 80 milliseconds of DProN2 stimulation. While there is, by definition, less time for the head to converge, the early head movements are still highly posture dependent and moving towards a point of convergence. Compare the different movements in the green and magenta circled trajectories in b, which are dominated by roll and pitch respectively. (c-d) To confirm that the mean head trajectories plotted in our action fields are representative of the underlying data and not biased by outlier events such as fast head saccades, we removed trials containing head saccades from the data. (c) is the action field of the CvN7 neuron replotted from Fig. 1j. (d) shows the same data but excluding any trial containing head saccades: defined as any trial containing yaw movements faster than 200 °/second. (e) All rotations in this study were measured in the lab reference frame. To confirm that the choice of reference frame did not substantially alter the results²² we plot here the head rotations resulting from CvN7 stimulation measured in the reference frame of the head at the time of stimulus onset. By definition all head rotations start at the origin. (f) To allow comparison of the head reference frame data in e with that measured in the lab reference frame in c we moved the start points of all trajectories in e from the origin to their respective start locations in the lab frame.