Trial adjustment reduced a 30 Hz map score from 1.53% to 0.83% in 76 selected mouse V1 PV outputs

We tested whether one source gain can preserve a measured output map across two pulses at 30 Hz. We used 76 inhibitory connections from 30 PV neurons in 21 mice. Exact presynaptic waveform matches identified shared exposures in 44 of 67 edge pairs. A tool averages fourth-order voltage products over distinct exposures. It estimates squared pulse-vector determinants without bias under independent stationary exposures and fixed sampling and peak rules. It allows correlated outputs within an exposure. Trial adjustment reduced a pooled map score from 1.53% to 0.83%. The central 95% whole-mouse resample range was 0.29% to 1.18%. Matched blocks of five exposures gave 0.68%, with range -0.47% to 1.26%. Bounded output contrasts also gave no clear repeatable difference across separate trial halves. Serial noise in a synthetic common-gain model could create a positive adjusted score. These results did not give a clear rejection of a common gain when trial dependence was allowed. Full-record selection and peak annotations also limit the unbiased interpretation. The contribution is a small method, its proof, and a checked benchmark with 129 selected records and 2,519 trial identities across output and input panels. This is a reanalysis of slice voltage data, not an EM connectome or a conductance measurement. We preserve source versions and hashes, code, compact data, all results, figures, and repeat instructions.

connectome · synaptic dynamics · parvalbumin · paired pulses · measurement error · shared trials · U-statistic · reproducible benchmark