Opposite 25% module gain changes reversed slow mode output bias in one zebrafish linear model with a fixed 10 s decay

We tested if one fitted decay time fixed the output bias in a linear model from one larval zebrafish wiring map. The model had 289 core neurons and 54 abducens output neurons. We kept synapse counts, signs, input normalization, intrinsic time constants and two anatomical modules. We multiplied target gains by 1.25 in the motor-targeting module and 0.75 in the internuclear-targeting module. We fitted one common scale to keep the longest decay at 10 s. The ratio of mean slow mode signals in motor and internuclear output neurons increased from 0.302 to 1.242. It crossed one at opposite gain changes of 21.37%. None of 500 class-matched random assignments of the same gains reversed this bias. The reversal remained with changes only at excitatory targets and with 5 s or 20 s decay fits. An independent full network calculation checked the output formula. These results show that a fixed persistence time did not fix relative sustained output within this model family. The gain range was selected during exploration. It was not measured in the fish. The result applies to one linear model and does not show a change in eye behavior. The package contains source receipts, independent code, all results, figures and reproduction instructions.

connectome · zebrafish · oculomotor integrator · neural dynamics · target gain · model sensitivity · persistent activity