Empirical response bounds leave averaging gains unresolved in 60 of 77 selected worm output pairs

We tested whether separate response lists can establish a variance reduction from averaging two outputs of one stimulated cell. Each source had chemical links to both targets in two adult worm maps. We treated each released wild-type response list as a fixed uniform distribution and computed sharp bounds over all ways to pair the two distributions. Of 77 selected output pairs from 22 source cells, 4 had a lower average variance than the less variable target for every pairing. Thirteen had a higher variance for every pairing. The other 60 depended on the unknown pairing. Independence predicted a reduction in 46 pairs; 42 of these remained unresolved under the bounds. These results describe fixed empirical distributions, not neural coding accuracy or within-animal noise. The lists lack joint trial labels, and at least one list length differs from the count field in 73 of 77 pairs. Separate animals supplied anatomy and activity. Code, retained values, exact formulas, transport checks, source hashes and a figure support reproduction.

C. elegans · connectome · response averaging · variance bounds · partial identification · empirical marginals