Mean linear information retention is 98.1 to 99.7 percent in eight worm map models with uniform independent site release

Under independent equal-probability release at chemical synapse sites, target currents retain 98.1 to 99.7 percent of site-level linear information about a common release probability, averaged equally across sources in eight selected worm maps. The estimate uses an optimal unbiased linear readout with access to all retained target currents. We prove an exact upper bound and a pooled-readout lower bound. A four-site example shows that fixed edge counts and fixed site target counts do not determine this efficiency. Observed grouping exceeded fixed-count control means by 0.16 to 1.40 percentage points. Five of eight primary comparisons reached an approximate eight-map adjusted reference rank of 0.0401. Longer runs gave the same direction but could not reach adjusted 0.05 at their saved-state resolution. Added observation noise reduced mean efficiency. The data are anatomy from different animals, not measured currents. Shared release, unit effects, equal probability and ideal readout access are assumptions. Geometry, activity and cross-species transfer are untested.

C. elegans · polyadic synapse · linear estimation · release probability · noise covariance · connectome · fixed margins