[Eeglablist] Open online discussion: How Do Cable Theory and AMPA/GABA Balance Compare in Their Contributions to 1/f?

Cedric Cannard ccannard at protonmail.com
Tue Apr 14 17:22:24 PDT 2026 

Hi Makoto and everyone,

I won't be able to attend unfortunately. One thing I am curious about is why we mainly focus on AMPA receptors as the excitatory model. I am no expert here, but from what I recall from undergrad classes, NMDA receptors are the dominant glutamate receptors in the human brain, so it feels like they should probably be included if we want to better understand the aperiodic background at the neuronal level?

According to an AI-assisted exploration (to be fact-checked), AMPA receptors are simpler to model because they are fast and Ca2+-impermeable, with stereotypical decay kinetics that map cleanly onto cable theory. NMDA, on the other hand, has a voltage-dependent Mg2+ block that introduces nonlinearities, which is why it tends to appear in more sophisticated compartmental models combining Hodgkin-Huxley conductances with multi-compartment cable structure (https://urldefense.com/v3/__https://pubmed.ncbi.nlm.nih.gov/8684467/__;!!Mih3wA!FIzLKzvJ2yvgcESdkJKVn6WMFR279-nFGjfNcMr1mO_rEcMhkKcRqWkebEG4KRMHmf4jQgYzj18GHSUE8URMHPkFbA$ ; https://urldefense.com/v3/__https://pmc.ncbi.nlm.nih.gov/articles/PMC10600871/__;!!Mih3wA!FIzLKzvJ2yvgcESdkJKVn6WMFR279-nFGjfNcMr1mO_rEcMhkKcRqWkebEG4KRMHmf4jQgYzj18GHSUE8UQ7a9S8hw$ ).
I believe Eugen and Makoto (and others) have already discussed on Github the possibility that one of the mechanisms underlying 1/f-ness is electrodiffusive, related to ionic diffusion in the extracellular medium and how voltage signals filter through brain tissue (https://urldefense.com/v3/__https://pubmed.ncbi.nlm.nih.gov/17025932/__;!!Mih3wA!FIzLKzvJ2yvgcESdkJKVn6WMFR279-nFGjfNcMr1mO_rEcMhkKcRqWkebEG4KRMHmf4jQgYzj18GHSUE8UQL2GLiYA$ ; https://urldefense.com/v3/__https://pubmed.ncbi.nlm.nih.gov/19348744/__;!!Mih3wA!FIzLKzvJ2yvgcESdkJKVn6WMFR279-nFGjfNcMr1mO_rEcMhkKcRqWkebEG4KRMHmf4jQgYzj18GHSUE8URLWVnFGQ$ ). 
Including NMDA feels relevant here because its slow, voltage-dependent conductance would extend the source signal spectrum, while electrodiffusion shapes it on the propagation side. 

Specifically:
- Multi-timescale kinetics. NMDA decay constants (~50-500 ms for NR2B) combined with AMPA (~5 ms) and GABA-B (~200 ms) naturally produce power-law-like spectra over a wider frequency range via superposition of Lorentzian processes, making the 1/f approximation more robust (https://urldefense.com/v3/__https://pubmed.ncbi.nlm.nih.gov/15483395/__;!!Mih3wA!FIzLKzvJ2yvgcESdkJKVn6WMFR279-nFGjfNcMr1mO_rEcMhkKcRqWkebEG4KRMHmf4jQgYzj18GHSUE8UQnALGmEQ$ ).
- Distal dendritic enrichment. NMDA receptors are concentrated on apical dendrites where cable filtering is strongest, potentially producing location-specific spectral signatures.
- Tonic NMDA activity. In high-conductance background states realistic in vivo, membrane fluctuations may regularly push voltage past the Mg2+ unblock threshold, contributing a slowly fluctuating conductance noise critical for the low-frequency end of the 1/f spectrum.
- Interaction with HCN (Ih) channels. Both are voltage-dependent and co-localized in distal apical dendrites (https://urldefense.com/v3/__https://pubmed.ncbi.nlm.nih.gov/25609619/__;!!Mih3wA!FIzLKzvJ2yvgcESdkJKVn6WMFR279-nFGjfNcMr1mO_rEcMhkKcRqWkebEG4KRMHmf4jQgYzj18GHSUE8USX3x02ZA$ ), so their interaction would probably be unavoidable in any realistic model.

Anyways, that's AI-heavy so to judge with caution, but it feels like there may be some useful food for thought in here.


Cedric



Sent with Proton Mail secure email.

On Tuesday, April 14th, 2026 at 9:59 AM, Gin Estrella Cruz via eeglablist <eeglablist at sccn.ucsd.edu> wrote:

> Dear all,
> 
> Thank you both -- these are very helpful perspectives.
> 
> I should say that I'm still quite new to 1/f analysis itself but very
> interested in it because of my broader interest in multiscale brain
> analysis, how synaptic /circuit-level mechanisms may relate to larger-scale
> signals.
> 
> I think the shift away from asking which mechanism is the single "correct"
> one toward asking whether different mechanisms may matter differently
> depending on recording scale, brain state, or experimental context is quite
> useful in this discussion.
> 
> Experimental ideas: These make me wonder whether a good near-term step
> might be to start with existing datasets that compare scales or states,
> e.g. scalp EEG vs more local recordings, or wake/sleep/anesthesia
> transitions, just to see which effects seem large enough to pursue more
> directly.
> 
> Longitudinal rTMS: Very interesting result. A large exponential shift over
> only a few hours does seem to suggest that at least part of the 1/f
> structure is dynamically changing vs being explained only by fixed
> structural properties. At the same time, could such a shift still reflect
> several factors at once, e.g. inhibition, arousal, oscillatory changes, or
> broader network reorganization, instead of one mechanism alone?
> 
> Would it make sense to think of this in two stages: (1) use available
> longitudinal or cross-scale data to see where the strongest effects seem to
> be, and then (2) build a more focused modeling or experimental comparison
> around those candidate mechanisms?
> 
> Warm regards,
> 
> GIn
> 
> On Mon, Apr 13, 2026 at 9:11 PM Ching-Ming Lee via eeglablist <
> eeglablist at sccn.ucsd.edu> wrote:
> 
> > Hello Makoto and the 1/f community,
> >
> > Inspired by this discussion, I would like to share some empirical data that
> > might help quantify the relative contributions of cable theory versus E/I
> > balance.
> >
> > As a physicist working on materials science and EEG, I conducted an N-of-1
> > longitudinal study on a 65-year-old subject (a senior physicist). We
> > delivered four consecutive sessions of alpha-tuned rTMS (totaling 8,600
> > pulses) within 3.5 hours and monitored the state evolution.
> >
> > *Key Findings:*
> >
> >    1.
> >
> >    *Dramatic Exponent Shift:* We observed the 1/f exponent (FOOOF) drifting
> >    from ~1.2 (baseline) up to ~1.8 at Stage 4.
> >    2.
> >
> >    *Structural Stability vs. Dynamic Change:* Since the subject's neuronal
> >    morphology (cable theory properties) cannot change by 60% within 3
> > hours,
> >    this massive shift provides strong evidence that *E/I balance
> >    (GABA-mediated inhibition)* is the primary driver of 1/f *dynamics*,
> >    even if cable theory sets the *baseline*.
> >    3.
> >
> >    *The "Collapse" Threshold:* Our WPLI network analysis shows that while
> >    the "Small-World" topology optimizes initially, it undergoes a *"Network
> >    Collapse"* beyond a certain exponent threshold (Stage 5), where global
> >    efficiency drops and local clustering disintegrates.
> >
> > *Implications for the Discussion:* This data supports the idea that
> > 1/f-ness is not just a passive physical property but a dynamic biomarker of
> > functional boundaries. It also challenges the clinical "more is better"
> > approach in TMS therapy, suggesting that we can push the brain into an
> > over-inhibited disordered phase.
> >
> > These findings are part of a manuscript currently in preparation.
> >
> > I look forward to discussing how we might use such longitudinal
> > "perturbation" data to weigh the factors Makoto and DeepSeek mentioned.
> >
> > Best regards,
> >
> > Ching-Ming Lee
> >
> > Graduate School of Materials Science
> >
> > National Yunlin University of Science and Technology
> >
> > Евгений Машеров via eeglablist <eeglablist at sccn.ucsd.edu> 於 2026年4月13日週一
> > 上午9:38寫道：
> >
> > > This is wonderful.
> > > I tried to devise an experiment that would allow me to choose between the
> > > hypotheses. Due to my laziness, I asked DeepSeek about it, but he
> > suggested
> > > what was probably a wonderful plan, but it required methods unavailable
> > to
> > > me (perhaps I could interest colleagues from another institute, but the
> > > chances are slim).
> > > Here's his advice:
> > > "Excellent question. This moves the discussion from theory to
> > experimental
> > > testing.
> > >
> > > The main problem is that in real EEG/LFP, all four mechanisms act
> > > simultaneously. The task is not to choose "the one true one" but to
> > > estimate the relative contribution of each in a specific context (type of
> > > activity, brain region, state, species, age).
> > >
> > > Below is one "clean" experiment (or type of analysis) for testing the
> > > primacy of each hypothesis. The gold standard is to compare model
> > > predictions with data, manipulating only one parameter at a time.
> > >
> > > 1. Testing hypothesis 2.1 (spatial averaging / source size)
> > > What experiment: Simultaneous scalp EEG + intracortical LFP (high-density
> > > linear array) in the same animal/human (intraoperatively).
> > >
> > > Logic:
> > >
> > > LFP registers local sources with a radius of ~0.5–2 mm — here, the mutual
> > > cancellation between spatially separated generators is minimal.
> > >
> > > Scalp EEG is the result of averaging over areas of several cm².
> > >
> > > Prediction: If spatial averaging is the main contributor to the 1/f
> > slope:
> > >
> > > In LFP, the spectral slope will be significantly flatter (closer to
> > > 1/f⁰…¹) compared to EEG.
> > >
> > > The difference in slope (EEG minus LFP) will be positive and correlate
> > > with the size of the coherent region, estimated via cross-coherence on
> > LFP.
> > >
> > > Control: If there is no difference, then the 1/f on the scalp is mainly
> > > shaped by local membrane/synaptic mechanisms (2.2, 2.3).
> > >
> > > 2. Testing hypothesis 2.2 (cable theory / dendritic filtering)
> > > What experiment: Patch-clamp from the soma + simultaneous activation of
> > > distal and proximal dendritic synapses (optogenetics or focal glutamate
> > > stimulation) in a single neuron in slice / in vivo.
> > >
> > > Logic:
> > >
> > > Stimulate proximal synapses (near the soma) — record EPSP at the soma.
> > >
> > > Stimulate distal synapses (on the same neuron) — record EPSP at the soma.
> > >
> > > Compare the spectra of these two response types.
> > >
> > > Prediction: If cable filtering is significant:
> > >
> > > EPSPs from distal inputs will have a steeper 1/f roll-off (less
> > > high-frequency power) compared to proximal ones.
> > >
> > > The peak amplitude (20–50 Hz) will be significantly reduced.
> > >
> > > Key experiment: Do the same in the presence of separate AMPA and GABA-A
> > > blockers to eliminate their kinetics. The remaining difference is purely
> > > passive filtering.
> > >
> > > Quantitative check: Compare with a cable model (Rall, NEURON) — match the
> > > spectral slope at different distances.
> > >
> > > 3. Testing hypothesis 2.3 (AMPA vs GABA-A kinetics)
> > > What experiment: Dynamic pharmacological manipulation (and optogenetics
> > > with different kinetics) at the same LFP site.
> > >
> > > Logic:
> > >
> > > Block GABA-A (bicuculline) — get predominance of fast AMPA-EPSPs.
> > >
> > > Block AMPA (CNQX) — leave slow GABA-A-IPSPs.
> > >
> > > Then restore the original balance.
> > >
> > > Compare the 1/f slope (2–100 Hz range).
> > >
> > > Prediction (per Gao & Donoghue 2016):
> > >
> > > AMPA-dominant → flatter spectrum (weak 1/f dependence).
> > >
> > > GABA-A-dominant → steeper spectrum (strong 1/f dependence).
> > >
> > > Critical control: Simultaneously monitor spatial coherence (to rule out
> > > mechanism 2.1) and vary stimulation distance from soma (to rule out 2.2).
> > >
> > > If effect exists — receptor kinetics make an independent contribution.
> > > If no effect — the 1/f slope is not primarily determined by kinetics
> > > (unlikely but possible).
> > >
> > > 4. Testing hypothesis 2.4 (slow ion concentration / open-loop current)
> > > What experiment: Long-duration (>10–20 min) recording of LFP +
> > > ion-selective microelectrodes (K⁺, Ca²⁺) in cortex without external
> > > stimulation (spontaneous activity).
> > >
> > > Logic:
> > >
> > > Standard microelectrodes record high-frequency LFP (0.5–200 Hz).
> > >
> > > Ion-selective electrodes give extracellular K⁺ concentration with ~0.1–1
> > > Hz temporal resolution.
> > >
> > > Prediction: If slow ionic shifts affect 1/f:
> > >
> > > The spectral slope in the 0.05–1 Hz range will correlate with K⁺
> > > concentration.
> > >
> > > When the "closed-loop" current is disrupted (e.g., via Na⁺-K⁺-ATPase
> > > blockade with ouabain, or during hypoxia), the 1/f slope will first
> > change,
> > > then return with a time constant of tens of seconds to minutes.
> > >
> > > Key test: Look at 40 Hz ASSR in two groups (red vs blue, as in your
> > > example) — if the difference in 1/f is at frequencies <<1 Hz but not at
> > >10
> > > Hz, that supports 2.4. If the difference is at high frequencies, that
> > > points to 2.1–2.3.
> > >
> > > Summary: which single experiment is the most powerful for disentangling
> > > these?
> > > There isn't one. But there is a combinatorial protocol that could yield a
> > > weighted contribution:
> > >
> > > Step    Manipulation    Tests
> > > 1       Scalp EEG vs intracortical LFP  2.1
> > > 2       Proximal vs distal responses (optogenetics)     2.2
> > > 3       Pharmacology (CNQX + bicuculline) at same LFP site      2.3
> > > 4       Long-term recording + ion-selective electrodes + ouabain
> > 2.4
> > > If after step 2 (distance control) the pharmacological effect remains —
> > > 2.3 is real.
> > > If after step 1 (accounting for source size) the 1/f almost disappears —
> > > 2.1 is the main contributor.
> > > If slope changes appear only in the very low-frequency range (<<1 Hz) —
> > > that's 2.4.
> > >
> > > This is a complex experiment, but it is already being partially performed
> > > in neurophysiology using optogenetics and high-density probe arrays."
> > >
> > > In the meantime, I'll try to come up with a way to differentiate between
> > > the two using existing EEG recordings of people in different states. I'm
> > > not sure it will work, but maybe it will be informative.
> > >
> > > Eugen Masherov
> > >
> > >
> > >
> > > > Hello 1/f people,
> > > >
> > > > Gin Estrella Cruz and I will meet online to discuss whether it is
> > > possible
> > > > to quantitatively compare the contributions of cable theory and
> > AMPA/GABA
> > > > (i.e., E/I) balance to EEG's 1/f power distribution. At this point, I
> > do
> > > > not know whether such a comparison is feasible. Anyone is welcome to
> > join
> > > > if the timing works, I'd love to hear your opinions and advice. Since
> > > there
> > > > is a 12-hour time difference between us, the meeting time is a bit
> > > awkward.
> > > >
> > > > *Time*: Apr 15, 2026 08:00 AM Eastern Time (US and Canada, EDT)
> > > > *Place*:
> > > >
> > >
> > https://urldefense.com/v3/__https://ucsd.zoom.us/j/3026035468?pwd=bQg61iUIe0AHfDQ5QSipOSEXi4FzCs.1&omn=97531862557__;!!Mih3wA!FZmTaWWKaDH-9ElW5sUmquTyS4E1TAy3KcNOWuur9efjaocNW7F7lNBCzhmS1d1FL9tA6L-gHGo2EbexRI-rf5m6dcE$
> > > > *Meeting ID*: 302 603 5468
> > > > *Password*: 1overF
> > > >
> > > > To provide a bit of background, since the publication of the original
> > > FOOOF
> > > > paper, the idea that a flatter or steeper 1/f slope in the EEG power
> > > > spectral density reflects excitatory/inhibitory states appears to have
> > > been
> > > > overgeneralized. However, the 1/f-like behavior of EEG is an intrinsic
> > > > property predicted by cable theory. If there are two contributors (at
> > > > least) to 1/f-ness, they should be compared quantitatively to determine
> > > > which contributes more.
> > > >
> > > > Makoto
> > > > _______________________________________________
> > > > To unsubscribe, send an empty email to
> > > eeglablist-unsubscribe at sccn.ucsd.edu or visit
> > > https://sccn.ucsd.edu/mailman/listinfo/eeglablist    .
> > > _______________________________________________
> > > To unsubscribe, send an empty email to
> > > eeglablist-unsubscribe at sccn.ucsd.edu or visit
> > > https://sccn.ucsd.edu/mailman/listinfo/eeglablist   .
> >
> >
> >
> > --
> > 李景明
> > _______________________________________________
> > To unsubscribe, send an empty email to
> > eeglablist-unsubscribe at sccn.ucsd.edu or visit
> > https://sccn.ucsd.edu/mailman/listinfo/eeglablist  .
> _______________________________________________
> To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu or visit https://sccn.ucsd.edu/mailman/listinfo/eeglablist .

More information about the eeglablist mailing list