Count pruning gave higher time irreversibility than matched uniform scaling in eight equal-noise linear worm models

We tested how synapse-count thresholds change time irreversibility in positive linear models of eight partial C. elegans brain maps. We kept 161 common cells and assumed equal leak and equal independent noise. At gain 0.5, removal of links with fewer than three sites retained 63.7 to 100.1 percent of the full stationary rate. Uniform scaling with the same total retained weight kept only 12.0 to 63.2 percent. Restoring total weight after pruning raised the rate to 1.59 to 6.33 times its original value. At gain 0.8, that restoration made six of eight models unstable. A known two-cell formula shows why removing a weak reverse link can increase the rate. Exact covariance solves, analytic checks, all tested settings, source hashes, code and a figure support reproduction. These are model results, not measured neural activity or heat production. Partial anatomy, selected cells, assumed positive weights and noise, and age-confounded maps limit biological use. External records remain outside the bundle.

C. elegans · connectome · time irreversibility · count threshold · Ornstein-Uhlenbeck · normalization · weak connections