Wing stiffness interacts with neural threshold to determine spike timing.

A: Accuracy as a function of wing stiffness. Chance accuracy is 50%, and stiffness factor 1 corresponds to 3 GPa, comparable to experimentally measured stiffness in the hawkmoth [28]. Green and purple dots indicate local peaks in accuracy. Shaded area indicates ±1 standard deviation. Top: Center schematic indicates the direction of rotation. Left and right schematics show the location of the single best sensor for each stiffness value. B: Filtered strain over a single wingbeat for flapping only (“no rotation”) and flapping with rotation (“rotation”) conditions at a single sensor location. Green and purple sensor locations in A correspond to left/green and right/purple filtered strain. Dashed horizontal line indicates the threshold (value at half-max) of the neural encoding nonlinearity. C: Filtered strain passed through the nonlinear function gives probability of spiking over time, P(fire). D: Spiking responses for 10 wingbeats each of the no rotation and rotation conditions. Spikes are generated probabilistically from P(fire), resulting in variable spike timing from wingbeat to wingbeat. Inset: Magnified view of the first firing event, surrounded by gray box, between 7.5 and 8.5 ms. E: Histogram of spike times (PSTHs) for each condition, summarizing spike timing over hundreds of wingbeats.