Time constants gave no clear direction prediction gain beyond type and depth in 539 one-way human temporal cortex pairs

We tested whether membrane time constants help predict the direction of detected one-way excitatory connections. We used 539 cell pairs from 21 human temporal cortex donors in the Planert Figshare version 1 tables. Both directions had explicit tests. We excluded repeated directions and disynaptic flags. A fixed logistic model used the signed log2 time constant ratio, depth difference, and six signed cell-type pair terms. The fitted direction odds ratio was 1.448 per twofold time ratio. Its 95% interval from 5,000 whole-donor bootstrap refits was 0.963 to 2.222. All refits passed. Each donor was then excluded from regression fitting. Adding time constants to type and depth raised mean donor Brier error from 0.205328 to 0.205912, or 0.284%. Error fell in 11 of 21 donors. A descriptive range for error reduction was -0.003763 to +0.002310. This range keeps test predictions fixed and omits refitting uncertainty. The model gives no clear added prediction gain. The association interval still permits a substantial positive preference for longer target times. Selection, slice recording, measurement error, source labels, and the fixed model restrict the result. These are functional connection graphs, not complete anatomical connectomes. We preserve original code, exact source row selections, source hashes, every bootstrap slope, predictions, figures, and repeat instructions. Independent parsing and GLM checks passed.

human temporal cortex · functional connectivity · membrane time constant · connection direction · paired recordings · donor validation