Skip to main content
Sandbox Physics

Stochastic Ion Channels: Why Small Neurons Fire Unreliably

When a patch holds only a few thousand channels, does the same shock always fire a spike? Not in a small patch. Each channel opens and closes at random, and with only a few thousand of them the noise decides the outcome near threshold: a patch of 100 square micrometres fires on about half of identical shocks. Larger patches return to the reliable Hodgkin–Huxley spike, and tiny ones fire with no stimulus at all.

Shrink a membrane patch and repeat the same shock. With fewer channels the spike becomes a coin flip, its timing jitters and noise alone starts to fire it.

Interactive modelStochastic Ion Channels: Why Small Neurons Fire Unreliably
Channels in the patchNNa=6000, NK=1800N_{\mathrm{Na}}=6000,\ N_{\mathrm{K}}=1800
Trials that fire a spikep^spike=0.527±0.041 (n=150)\hat p_{\mathrm{spike}}=0.527\pm0.041\ (n=150)
Spread of first-spike timesσt=1.57±0.13 ms\sigma_t=1.57\pm0.13\ \mathrm{ms}
Deterministic Hodgkin–Huxley latencytHH=6.63 mst_{\mathrm{HH}}=6.63\ \mathrm{ms}
Recent stochastic trials and the deterministic membrane
First-spike latency over all trials
Open sodium channels in the trial shown, against the binomial expectation
More numbers and how they are computed
Patch area
00
Mean first-spike latency
00
All spikes per second of trial time
00
Trials run
00
Deterministic shock threshold
00
Channel densities and conductance
00
Time step of the channel chains
00

Each trial uses its own seed and settles for 10 ms before the record starts. A trial counts as firing when the depolarization crosses 50 mV upward after the shock. Standard errors are binomial for the fraction and large-sample for the spread. Estimates stop changing after 1500 trials.

Fewer channels, less reliable spikes

BackgroundThe Hodgkin–Huxley gate variables are averages over many channels. Each real channel is a protein that snaps open and shut at random, so a patch with a finite number of them carries a current that fluctuates even when nothing outside changes.

Why it mattersThin axons, dendrites and small neurons have far fewer channels than a squid giant axon. Whether their spikes are reliable is a question about counting statistics, and the answer changes near threshold first.

One-sentence intuition
At a fixed voltage the number of open channels is binomial, so its relative spread falls as N−1/2N^{-1/2}. Near threshold the deterministic membrane balances on a knife edge, and that small spread decides the outcome: a 100 μm2100\ \mu\mathrm{m^2} patch fires on about half of identical shocks.

Core mathematical model

Open channels at a fixed voltage

Nopen∼Bin(NNa, m3h),Var⁡Nopen⟨Nopen⟩=1−m3hNNa m3hN_{\mathrm{open}}\sim\mathrm{Bin}(N_{\mathrm{Na}},\,m^3h),\qquad \frac{\sqrt{\operatorname{Var}N_{\mathrm{open}}}}{\langle N_{\mathrm{open}}\rangle}=\sqrt{\frac{1-m^3h}{N_{\mathrm{Na}}\,m^3h}}

Channels are independent and each is open with the Hodgkin–Huxley probability, so the count is binomial. The benchmark checks the mean and variance at six times after a voltage step.

Channel numbers in the patch

NNa=60 A,NK=18 A,A in μm2N_{\mathrm{Na}}=60\,A,\qquad N_{\mathrm K}=18\,A,\qquad A\ \text{in}\ \mu\mathrm{m^2}

With 20 pS per channel these densities give the 1952 maximum conductances, 120120 and 36 mS cm−236\ \mathrm{mS\,cm^{-2}}.

Jitter in the small-noise limit

σt∝N−1/2(N→∞)\sigma_t\propto N^{-1/2}\qquad (N\to\infty)

Valid only when the noise is small. Between 100100 and 1000 μm21000\ \mu\mathrm{m^2} the simulated jitter falls faster than this, so do not extrapolate it to small patches.

Common difficulties

The noise is in the membrane, not the stimulus

Typical misconceptionTrials differ because the shock is not exactly repeated.

Better mental modelEvery trial gets an identical shock. The only random numbers in the model decide which gates open and close.

Large patches are not noise-free

Typical misconceptionAbove some size the patch behaves exactly like the equations.

Better mental modelThe spread shrinks but never vanishes. At 1000 μm21000\ \mu\mathrm{m^2}, sixty thousand sodium channels still give a latency spread of about 0.06 ms.

Run the experiment

  1. 01

    Shock at threshold

    Leave the coin-flip preset running for twenty seconds.

    What to observe: About half the trials fire. The deterministic membrane with the same shock fires only once, after about 6.6 ms, because 6.5 nC cm−26.5\ \mathrm{nC\,cm^{-2}} is just above its threshold.
  2. 02

    Shock hard and watch the timing

    Choose Spike times jitter.

    What to observe: Nearly every trial now fires, but the first spike times have a spread of about 0.28 ms. The deterministic membrane fires after 0.87 ms every time.
  3. 03

    Add channels

    Choose Many channels: back to Hodgkin–Huxley, then drag the rim inward.

    What to observe: With ten times the channels the spread falls to about 0.06 ms. Shrinking the patch brings the jitter back. Small patches also fire with no shock at all, about 38 times per second at ten square micrometres.

Related experiments and articles

What this page links to directly

See every article series on the content map →

Continue exploring

Chance and thresholds in living systems

How do random events, thresholds and neighbour coupling shape populations, neurons and tissue patterns?

  1. LabGenetic Drift and Fixation: Who Takes Over a Small Population?Leave Half and half: a coin flip running and watch chance alone let one type take over.
  2. LabRange Expansion Sectors: Why a Mixed Colony Grows in StripesChoose A straight front loses colours and watch the number of sectors fall as it grows.
  3. LabPulled and Pushed Waves: Where an Invading Population Comes FromCompare the Pulled and Fully pushed presets and see where the ancestors of the front sit.
  4. LabHodgkin–Huxley Model: How a Squid Axon Fires an Action PotentialCompare One all-or-none spike with Just below threshold and watch the membrane voltage.
  5. You are hereStochastic Ion Channels: Why Small Neurons Fire UnreliablyLeave Threshold becomes a coin flip running and count how often the same shock fires.
  6. LabExcitable waves and spiralsChoose A half-recovered sheet and drag a short stroke on each side to see where a wave starts.
  7. LabNeural Avalanches and Criticality: When a Power Law MisleadsCompare Cascades near the critical point with Shared slow input, no coupling on the size chart.
  8. LabIsing Critical FluctuationsCompare Order survives with Heat breaks the patches and watch how far aligned patches reach.
  9. LabHopfield Associative MemoryChoose Recall lightning, drag over part of the picture and watch the network restore it.
  10. LabTuring patternsChoose Spots emerge, then Diffusion smooths, and compare whether a pattern grows.
  11. LabActive Nematic FlowCompare Defects drive the flow with Reverse the stress and watch the flow turn around.

Statistical physics of neurons · Reading path

From a charged membrane to avalanches in a network of neurons

Articles on this path that use this experiment, followed by the other experiments they suggest.

  1. ArticleSplitting the gating variables into single channels
  2. ArticleChannel noise blurs the threshold and the spike time
Something broken, confusing, or missing on this page?Tell us
What kind of feedback?

Sent with this page’s address, its Lab ID, the site language and your device type. Please leave out personal details. Feedback is kept for up to 12 months.Privacy policy