A latent release model leaves 11 of 28 reliability orders unresolved for selected MICrONS contact counts

We tested whether a latent release probability removes ambiguity in reliability ranks based on structural contact counts. We used the count support and histogram from 8,128 connected MICrONS cell pairs in one mouse. Release parameters were assumptions, not measurements. At release probability 0.5 and pair correlation 0.2, sharp latent-model bounds left 11 of 28 distinct-count comparisons unresolved, versus 15 under two moments only. Three contacts permit success probability from 0.780 to 0.820; four permit 0.812 to 0.892. Explicit latent laws give both strict orders. A single identical latent law across connections instead makes every count order strict. Finite exchangeable pool models show how extension assumptions narrow bounds. Exact rational witnesses, polynomial certificates, linear-program dual checks and independent vertex checks support the result. The probability methods are established. Equal release moments, conditional independence, one selected animal and absent physiological release data limit biological interpretation. The package contains original code, aggregate evidence, source hashes and reproduction instructions.

MICrONS · Allen Institute · contact count · release reliability · latent probability · exchangeability · moment bounds