In 27 sources to one leech motor neuron, 55% of site pairs within 20 µm shared a source; distinct nearby sources showed no clear gain in swim coherence similarity

We tested whether distinct input neurons with nearby synapses had similar swim activity in one leech DE-3R motor neuron circuit. We used public electron microscopy and somatic voltage-dye coherence data. A declared selection gave 27 sources and 232 input synapse IDs. At a path distance of 20 micrometres, 290 of 528 nearby site pairs (54.9%) shared a source. Such pairs have zero assigned coherence difference when every site inherits its source value. Including them gave a descriptive site-pair swim similarity score of +32.1%; distinct-source site pairs gave -38.4%. The primary test gave each distinct neuron pair equal weight. For 37 nearby pairs among 351 source pairs, mean squared coherence difference was 18.1% larger than the all-pair reference across two swim trials. Shuffling complete response profiles within three synapse-count bands gave one-sided p = 0.774 with 19,999 shuffles. No individual trial or scale passed a shared maximum-statistic correction across 49 tests. The swim profiles repeated well (correlation 0.920). Count-normalized weights, Euclidean distance, confidence filters, and a wider mapped-source set gave no clear positive gain. Removing the largest source changed the effect sign, but the gain remained uncertain. We provide an exact decomposition of the effect of shared sources, a derived pair dataset, code, source hashes, figures, and repeat instructions. This is a test of a local pair rule in one target and one animal. It does not refute the source study's distinct cluster analysis or test local synapse activity. Crawl coherence exports were absent from the pinned release.

connectome · leech · motor neuron · structure and function · synapse clustering · coherence · neuron-level randomization