Pulse coupling estimates did not clearly improve baseline voltage correlation predictions in fixed tests of 408 mouse V1 Pvalb pairs

We tested whether pulse coupling predicts slow baseline voltage correlation among 408 Pvalb cell pairs from 144 inferred mice. We used Allen Synaptic Physiology r2.1. A fixed ridge model used distance, layer labels, and recording conditions. Five folds kept complete mice apart. Adding pulse terms reduced squared error by 0.39%. The 95% fixed-prediction mouse resampling range was -6.77% to +3.71%. A second split increased error by 0.64%. Neither fitted model beat a training-mean prediction. We inferred sweep counts from the stored Pearson statistics. Pulse and baseline statistics agreed in all 816 directions. For 321 pairs, the smaller count was six sweeps. One two-sweep pair gave 40.2% of the primary reference error. Post hoc tests with at least five sweeps or a bounded outcome still gave no clear gain. Source code showed that pulse and baseline estimates share a voltage window. We derived a moment correction under stationary background, independent pulse drive, and constant linear transfer. Model controls recovered its population value, but six sweeps left large error and finite-sample bias. The released scalars lack the moments needed to apply the correction to the recordings. This study supplies a fixed prediction result, a sample-count audit, and a conditional correction for future raw data. It does not show that electrical synapses have no role in common activity. The package preserves original code, transformed pair data, source versions and hashes, all model states and predictions, simulations, figures, and repeat instructions.

electrical synapse · Pvalb · mouse visual cortex · structure and function · voltage correlation · measurement error · paired recording · out of sample prediction