Variable spiking is a ubiquitous neuronal feature in neocortex.1
Physiological spiking variability reflects input drive characteristics rather than noise alone.1
Weak input synchrony induces physiological levels of spiking variability.1
Synchrony timescales are critical in regulating shifts in variability (Pattadkal et al., Neuron).1
These relationships have been demonstrated using in vivo, in vitro, and in silico approaches.11
Deep research (2026-07-20)
Authors/institutions: Jagruti J. Pattadkal (co-corresponding), Ronan T. O’Shea, David Hansel (CNRS/Université Paris Cité), Thibaud Taillefumier, Darrin H. Brager, and Nicholas J. Priebe (co-corresponding, lead contact), Department of Neuroscience & Center for Learning and Memory, University of Texas at Austin; final citation Neuron 2026 Feb 18;114(4):724–739.e19.23
Uses dynamic-clamp recordings in pyramidal neurons from mouse and marmoset cortical slices, injecting in vivo-recorded excitatory/inhibitory conductances; membrane properties did not differ across species (resting Vm: mouse −63±1.4 mV vs marmoset −60±3.2 mV; input resistance 193±16.5 MΩ vs 217±12.6 MΩ).2
Repeated injection of the same conductance trace gave low trial-to-trial variability: Fano factor 0.2±0.3 SD (12 cells) and 0.2±0.2 SD (11 cells) — “quasi-deterministic” responses, arguing against intrinsic cellular noise as the variability source.2
Injecting different per-trial in vivo conductances to the same stimulus recreated physiological variability: Fano factor rose to 1.3±0.8 SD and 1.4±0.7 SD, matching in vivo Poisson-like spiking.2
Excitation alone (no inhibition) still produced variable spiking (Fano factor 0.7±0.3 SD, 12 cells) vs. low variability for repeated identical excitatory input (0.1±0.2 SD), with the excitation+inhibition condition significantly more variable (p=0.003, t-test).2
A novel population synchrony metric (χ) applied to neuropixels recordings in awake marmoset V1/MT and mouse V1 (Allen Institute) found weak but significant spiking correlation: ρ = 0.014±0.014 SD (n=13 marmoset populations) and 0.081±0.037 SD (29 mouse sets), with spontaneous-state ρ higher in some comparisons (paired t-test p=0.02).2
Synchrony timescale shifts with brain state: driven/stimulus-evoked τ ≈ 25–50 ms (mean 28.3±13.7 ms SD) vs. spontaneous τ ≈ 126.8±102.6 ms SD.32
Injecting synthetic conductances with prescribed synchrony reproduced Poisson-like output: mean Fano factor 1.2±0.4 SD (n=12) with synchronous input vs. 0.6±0.2 SD (n=10) with asynchronous input of matched rate, confirming synchrony (not intrinsic noise) drives cortical spiking irregularity.2
Single-neuron response reliability to identical input was high (0.92±0.05, spike-timing precision 3.9±1.3 ms, n=10) but dropped across different cells given the same input (0.72±0.06, 4.6±0.6 ms) — intrinsic cell-to-cell heterogeneity shapes population synchrony without adding within-cell trial-to-trial noise.2
BCI/decoding implication: because spiking irregularity is a network/synchrony phenomenon rather than intrinsic neuronal stochasticity, decoders may exploit state-dependent synchrony timescales (fast during task engagement, slow at rest) to improve trial-to-trial reliability and reduce the need to average over many trials for stable neural decoding.
Neural Noise