Selected input groups had excess movie response similarity under degree control in one MICrONS mouse

We tested whether measured movie responses were more similar among cells that send input to one target than constrained wiring predicts. The MICrONS v1 panel had 8,128 connections from 148 selected source cells. The mean pair correlation within 772 input groups was 0.08442. A null that fixed both cell degrees and measured contact opportunity had mean 0.07340 and a sampled central 95 percent range of 0.06737 to 0.07989. Four chains per null gave 400 stored graphs. A second null also fixed source-specific counts by target area, layer and five response-similarity bins. Its mean was 0.07460 and input-score R-hat was 1.010. With ten bins, R-hat was 1.087; this stricter test remains unresolved. Secondary output scores also had chain disagreement. The result is conditional on a selected input panel in one mouse. Coarse bins, omitted fine geometry and possible unreachable graph states prevent a general higher-order wiring claim. Code, aggregate draws, source hashes and a retrieval script support reproduction.

MICrONS · convergent inputs · response similarity · degree control · conditional graph model · connectomics