Postsynaptic time constants gave no mean prediction gain for EPSP duration in 21 human cortical donors

We tested whether target membrane time constants improve prediction of incoming excitatory response duration beyond cell type and geometry. We used 496 direct connections from 21 human layer 2–3 cortical donors in figshare dataset 28184126 version 1. Each donor was held out in turn. Adding target time constants to a model with source time constants, cell types, soma distance, and depths gave a mean donor reduction in squared error of log half-width of -0.000153. A donor bootstrap gave a 95% interval of -0.003871 to 0.003387. The adjusted target coefficient was 0.118, with a 95% donor-cluster interval of -0.077 to 0.313. A high signal-to-noise subset had a stronger association. The primary test did not show a mean prediction gain. It does not exclude membrane filtering or establish equivalence. Missing measurements, cell type adjustment, selection, and a single source dataset limit the result. Code, provenance, predictions, donor errors, and a figure are included.

human cortex · paired recordings · membrane time constant · synaptic duration · donor validation · null result