Median quantal parameters understated mean model failure by over 466-fold across re-pairings of five mouse S1 fits

We tested whether median release parameters represent mean failure in selected mouse cortical connections. We used paired quantal fits and baseline failure counts from public eLife source data. In five primary somatosensory cortex connections, an independent equal-probability release model gave 9.444% mean failure. Median parameters gave 0.01034%. Across all 120 re-pairings of the five effective site estimates and release probabilities, mean failure was at least 4.829%, over 466 times the median-parameter prediction. One low-site connection gave 87.1% of summed model failure and all 30 observed failures. Five prefrontal connections and five separate antagonist-fit connections had smaller summary gaps. These are exact checks of fixed estimates under a restricted model. Small selected samples, shared estimation trials, uncertain effective site counts, and absent fit uncertainty prevent population or causal claims. We supply code, selected records, source hashes, full re-pairings, and an independent arithmetic check.

synaptic reliability · quantal analysis · release sites · nonlinear aggregation · mouse cortex · structure and function