Targeted call corrections can remove pooled reciprocity excess in three Allen cortical slice samples

We calculated exact correction budgets for pooled reciprocal connection excess in Allen Synaptic Physiology r2.1. Missing calls were free to take either state. In mouse visual cortex excitatory pairs, mouse inhibitory pairs, and human temporal cortex excitatory pairs, the minimum positive-call deletions were 29, 109, and 10. These are 10.07%, 10.77%, and 8.33% of recorded positives, placed in 80.6%, 61.2%, and 76.9% of observed reciprocal dyads. Negative-call additions instead required 451, 851, and 93 corrections. Mixed corrections did not reduce the primary minimum totals. A sharp bound uses reciprocal and nonempty dyad counts; explicit corrected tables attain every threshold. Independent state enumeration passed 461 small examples. Distance and test-count checks gave similar sensitivity patterns. These are adversarial finite-table thresholds, not estimated assay error rates or confidence intervals. Pooling, selected slice graphs, free missing assignments, and unknown detection errors limit biological interpretation. The package includes a proof, source hashes, code, pair records, correction witnesses, and a figure.

Allen Institute · reciprocal connectivity · observation error · exact bounds · cortical circuits