With a 20 ms minimum time constant, a 125 ms Mi4 current delay reversed 4 Hz direction contrast in 9 of 13 selected T4 units in nine periodic fly visual models

We tested whether changing Mi4 input is needed for motion selectivity in released fly visual models. The main set had 13 selected T4a and T4b units in nine of the ten models with the lowest task loss. We used 4 Hz gratings, periodic boundaries, a 20 ms minimum time constant and 2.5 ms steps. Replaying the intact gray current preserved the complete gray response. It reduced mean direction contrast from 0.417 to 0.056. A constant replay kept each target's intact mean current for each direction. It gave mean contrast 0.071. Each replay kept contrast at or above 0.10 in only 2 of 13 units. A later test replayed a repeating Mi4 current waveform. A delay changed its timing while keeping its mean and value distribution fixed. Zero delay kept the threshold in all 13 units. A 125 ms delay reversed the intact direction order in nine units. A 187.5 ms delay reversed it in eleven. A half-size time step kept both reversal counts. Full finite-eye runs also kept only two units above threshold under gray or constant mean replay. In all 50 models, 26 of 54 selected units reversed after the 125 ms delay. Exact self-replay, source reconstruction, independent edge sums and selected new integrations checked the method. The result supports a timing role for this pathway under the stated model probe. It does not establish Mi4 function in animals. The package has source archives, code, all traces, results, figures, source hashes and repeat instructions.

connectome · Drosophila · visual motion · Mi4 · T4 · current replay · direction selectivity · causal model