Back to Journal

40 Days of Continuous Fungal Electrical Recordings: What Failed, What I Corrected, and What I’m Doing Next

None
January 27, 2026

40 Days of Continuous Fungal Electrical Recordings: What Failed, What I Corrected, and What I’m Doing Next

Date: 2026-01-27

Context

Since writing my first hypothesis, a lot has happened.

I've been called out - more than once -on my ignorance, and pointed toward existing frameworks that already formalize state spaces: Markov models, Hidden Markov Models, and related tools. I fully acknowledge that these frameworks exist and are well established. Still, I'm glad I arrived at many of these ideas independently. The derivation mattered; the process shaped how I now think about the problem.

In the time between then and now, I learned a number of concepts that are basic for many, but were new to me:

All of these became useful largely because I had already spent time thinking carefully about what I meant by state in my earlier writing.

Experimental Context

Over the last month, I've been running a continuous experiment.

I have four fungal blocks in a dark room, divided into two sections:

Each block has 16 electrodes arranged in a 4×4 grid.

The light transitions followed alternating light-on / light-off intervals of 12h, 6h, and 3h.

At this point, I have roughly 40 days of continuous data.

What We Tried

The original goal was to identify electrical activity patterns following light stimuli in the LD blocks that did not appear in the always-dark blocks.

We ran a broad bank of analyses:

At one point, I suspected that fungal states might be encoded in frequency bands analogous to alpha or theta waves in the brain. That hypothesis did not appear to hold (though more analysis is required).

I also attempted unsupervised clustering (HDBSCAN). The result suggested either:

Something about that result didn't feel right.

A Major Mistake (and a Lucky One)

Eventually, we realized a serious issue.

I had mislabeled the blocks.

One block from the LD group and one block from the always-dark group were being treated as a single group. This meant that a significant portion of the analysis had been run on mixed conditions.

Fortunately, all analyses are implemented as Python scripts and can be rerun cleanly with corrected labels. Re-running the entire pipeline with correct grouping is one of the next steps.

Now that this error is resolved, it's possible that the corrected analyses will reveal much more structure. That remains to be seen.

A Key Conceptual Correction

Reading a paper by Adamatzky
https://arxiv.org/abs/2601.08099
marked a turning point.

It made clear that my comparison approach was fundamentally flawed.

Each electrode channel represents a spatially distinct region with:

The fungal network is heterogeneous. Yet much of our earlier analysis implicitly treated channels as equivalent and directly comparable.

This insight forced a major reframing.

Even for stimuli like light which is applied uniformly across the block surface - it is no longer justified to average responses across channels. Each channel must be treated as an independent zone, which may or may not correlate with other zones. Therefore, all comparisons going forward are intrachannel - each electrode serves as its own reference frame. This is consistent with Adamatzky's observations.

Returning to "State"

Despite these corrections, I kept returning to the idea of state space.

Since electrical activity is defined through deviations from baseline, I began asking more fundamental questions:

This led to a new working hypothesis.

Peace as a Stable Configuration

I now think of the baseline as the fungus being in a state of peace (metaphorically, not literally).

By peace, I do not mean inactivity. I mean a stable configuration - homeostasis - a region of state space in which the system is internally consistent and requires minimal corrective work.

This is conceptually similar to:

Peace is not necessarily a single state, but a set of states the system can occupy without needing to actively correct itself.

In this framing, peace corresponds to a match between prediction and reality.

Example

Assume (without claiming it is true) that memory is encoded in the physical configuration of the fungal network. Based on this configuration, the fungus has effectively "determined" that a food source lies in the northeast direction.

A Note on "Prediction"

I'm borrowing language from Michael Levin's framing, where all agentic life forms, across scales, optimize toward local goals, and optimization requires prediction.

In this view:

Fungi likely operate somewhere between a bacterium and a dog.

I find this framing useful as intuition: electrical deviations may correspond to gaps between what the fungus "expects" and what it encounters.

Why Peace Is Not Static

An organism in peace is not hungry; but peace does not last forever.

Even in a stable configuration, energy is continuously consumed. Over time, peace degrades into something functionally equivalent to hunger. At that point, the fungus begins to act:

In this framing:

Electrical activity like spikes, variance, fluctuations, are measurable traces of the work required to maintain or re-enter a stable region of state space.

Working Conclusion

Therefore, corrective work should manifest as increased fluctuation, often observed as spikes.

Operationalizing State

We do not yet know what truly represents state in fungi.

However, given the hypothesis of a base state (peace) and deviations requiring corrective work, it is reasonable to operationalize state in terms of:

I will test both approaches and report which proves more informative.

Most of the literature - particularly Adamatzky's work -treats spikes as the primary indicator of response. Spikes are typically defined as amplitude excursions beyond a threshold.

Two Insights

Insight 1: Global Baseline from Long-Term Variability

Take the full 40 days of data from a dark-only condition for a single channel. Compute mean or variability in fixed windows (e.g., 30-second windows). Construct a histogram of these values.

We tested for slow drift using linear and cubic regression across the 40-day recording; none was detected, supporting the use of a single global baseline per channel.

If the system occupies certain ranges most frequently, those ranges can reasonably be interpreted as the global peace state (or global baseline) for that channel.

Insight 2: State as Deviations per Unit Time

State can be expressed as the number of deviations from the global baseline (in raw voltage or variability) per unit time.

State as Spike Rate

Using this definition, the state of a channel over a longer interval (e.g., one hour) can be represented as:

State = number of spikes per unit time relative to the global baseline

Window size and threshold remain free parameters; sensitivity testing will be reported in the next entry.

This definition is operational rather than biological. It does not claim to identify the underlying physiological state, but provides a measurable proxy.

In this framing, circadian or long-timescale rhythms may emerge naturally as systematic changes in spike rate over time, rather than requiring explicit state labels.

Open Questions

Many unknowns remain. Validating any of this framework requires further experiments - most critically, a dead control (electrodes in inert material) to rule out equipment artifacts. Other open questions include optimal threshold and window parameters, whether amplitude or variability better captures state, and whether the framework holds across different fungal species or substrates.

This is my current working model. It is certainly incomplete - but it feels internally coherent enough to guide the next phase of analysis. The next entry will address whether these predictions held and what that implies for the framework.