Reverse excitation change gave no clear prediction gain for 30 Hz inhibitory change in 60 mouse V1 pairs with separate test mice and trial halves

We tested whether reverse response change helps predict inhibitory voltage change at 30 Hz. We used 60 reciprocal PV–pyramidal pairs from 25 mice in released mouse V1 data. Each test mouse was absent from fitting. Disjoint trial halves supplied predictors and targets, with full-data source annotations held fixed. A fixed linear model controlled for both first responses and soma distance. Adding excitatory change raised equal-mouse mean squared error from 0.0392725 to 0.0401855. Error reduction was -2.32%, with a central 95% whole-mouse refit range of -8.70% to +2.94%. The connection’s own separate-trial inhibitory change also gave no clear gain. An added precision check found first-response feature repeat correlations of 0.960 for excitation and 0.935 for inhibition. Change correlations were 0.399 and 0.027. This result limits the tested prediction rule. Weak change precision prevents an inference that true dynamics are independent. The recordings are selected slice connections, not a complete connectome. We reproduced the compact inputs from all 628 raw source records. The package preserves code, derived data, provenance, model states, all results, figures and repeat instructions.

reciprocal connectivity · short-term synaptic dynamics · parvalbumin · visual cortex · voltage response · trial split · prediction · measurement precision