Reverse decay gave no clear inhibitory decay prediction gain in fixed ridge tests of 114 reciprocal mouse V1 pairs

We tested whether the decay time of an excitatory response helps predict decay in the reverse inhibitory connection. We used Allen Synaptic Physiology r2.1 and two mouse primary visual cortex panels. The main panels had 54 reciprocal cell pairs from 35 inferred mice and 60 pairs from 51 other inferred mice. The base model used own response rise, both response strengths, both membrane times, distance, cell labels, age and temperature. Every test mouse was absent from fitting. A fixed ridge model added the log excitatory decay time. Mean squared log error fell by 12.20% in the coarse panel and rose by 1.47% in patch-seq. In 2,000 whole-mouse resamples with full refitting, the central 95% gain ranges were -6.01% to +34.39% and -10.65% to +7.63%. After pairs near documented decay fit limits were removed, error rose by 3.08% and 2.10%; both ranges still included zero. The coarse-panel fitted slope was negative, but that slope range crossed zero after the fit-limit restriction. Separate source queries recovered 456 selected kinetic values exactly. Independent solvers, multiplicity checks and a large artificial effect passed. A full source rebuild and numerical replay matched. The result supplies a reproducible timing benchmark for sparse measured connection graphs. It does not establish independent biological time constants, a causal rule or intact-circuit feedback speed. The package preserves code, source versions and hashes, all selections, predictions, refits, figures and repeat steps. Original measurements remain external under their source terms.

reciprocal connectivity · synaptic decay · visual cortex · inhibitory feedback · synaptic physiology · prediction · measurement limits · reproducibility