Visual similarity gave no clear gain in a fixed ridge test of 5 to 30 ms voltage shapes in 88 mouse V1 connections

We tested whether visual response similarity helps predict recorded voltage shape after measured strength and recording controls. We used paired visual and electrical data from mouse V1. Source rules selected 92 detected directions. A ten-trial rule retained 88 directions from 20 mice. We tested mean voltage from 5 to 30 ms after an aligned spike and divided each curve by its RMS. Each test mouse stayed outside ridge model fitting. Adding visual cosine changed mean squared error from 0.039065 to 0.039879. The error decrease was -2.08%, with a central 95% range of -15.93% to +3.75% from 2,000 refitted mouse resamples. The inhibitory-only point gain was +5.29%, with a range of -18.12% to +22.64%. The known similarity-strength relation reproduced. Trial halves gave high raw curve agreement, but direction-centered shape alignment was only 0.366. The result gives no clear added shape prediction gain under the tested rules. It does not show equal biological dynamics or no role for tuned wiring. The package preserves compact trial data, code, source hashes, all predictions, model states, figures and repeat steps. A full source rebuild and independent numerical checks passed.

visual cortex · functional connectivity · parvalbumin · postsynaptic potential · waveform prediction · structure and function · reproducibility