Deciphering the Words of a Seething Vat

Cover Image: The titular seething vat, containing the fungus Flammulina velupites

What does a language need? Does it need a voice or tongue? No; sign language, which uses neither, is used by deaf and non-verbal people all the time. What about muscles? Or a brain?

Can there be a language without any sort of nervous system at all?

Language, stripped of all the externalities upon which we rely to identify it, is a string of transmitted data. It is by no means exclusive to humanity, to organisms with brains, or even to animals. Its traces have been found in an unlikely candidate, and a non-motile one at that: fungi.

This is not easy to believe. We might assume that this is just an exaggeration about another one of the thousands of involuntary signalling mechanisms of nature – surely nothing as rich as language. But we know that colossal datasets of human language can be analysed fairly easily and replicated. Silicon, in the form of unthinking circuitry, has proven itself capable of simulating language by predicting words. If patterns in the data of our own tongues can be found and predicted, can we not analyse similar patterns in other sets of data? And if these predictions can be made for another set of communicated information, is that not language?

Electrical mycolinguistics

The main question of this article is raised mainly thanks to the results of a small group of studies, by the same group of dedicated scientists. They are centred on an extremely obscure field, which, if the authors were pompous enough, may rightly be called “electrical mycolinguistics.”

In 2022, a group of scientists analysed spikes in electrical potential continuously in a few different fungi over multiple days. This was measured using electrodes connected to a sensitive voltmeter. The spikes had different amplitudes (the number of volts), durations, and gaps between them. Most produced close to a thousand spikes over the sampling period. One of the fungi studied, Schizophyllum commune, had multiple fruiting bodies at the time of measurement, and synchronised electrical activity was seen in all of them, showing that these are not random fluctuations. Presumably, the authors had reason to believe that the synchronisation was not just a result of electrical conductivity. We have known that we can measure this for some time: the musician Cosmo Sheldrake, brother of the (relatively) well-known fungus expert Merlin Sheldrake, has recorded the electrical fluctuations of fungi and converted  them to sounds for his songs multiple times. They are often quite rhythmic.

It is the statistical analysis done to the information, which operates under the assumption that it shows signs of language, that truly requires a leap of the imagination. Firstly, a language is made of some root words, or, more correctly, morphemes – indivisible units of meaning. To separate these long strings of spikes in electrical potential into groups, they tried two definitions of the boundary between two words. First, they designated all the gaps longer than the average interval between spikes to be word boundaries, and then they tried the same thing with gaps more than twice as long as the average interval. Basically, if the former assumption were correct, we would expect the time lag between spikes within a word to be consistently shorter than the time lag between words. They found that when they separated the sequence by the former assumption, the average “word” lengths that resulted were strikingly similar to those in languages like English and Greek, if the number of spikes were compared to the number of letters. This similarity was not noted for the latter, which produced longer “words”.

This comparison to human languages, though interesting on the face of it, breaks down very quickly. Making this link suggests some sort of universal statistical trend, or generalisation, for all languages. The sample of languages they chose for this, however, is skewed because English and Greek are related, being Indo-European languages. Furthermore, the method of deciding word length is wholly based on the writing system – the alphabet – and cannot be applied to languages with other systems. Do you count by syllable? Or do you count each sound separately? What about long vowels, or doubled consonants? Since the inclusion of this comparison in the article is such a thorny issue, it is not convincing at all.

Next, they analysed the sequences the “words” were allowed to occur in. Each word was represented by a node, and the sequences were mapped between them. The resulting graphs (state-transition graphs) strongly reminded me of something.

Fig 1: Left: the state transition graphs. (a) to (d) show the graphs formed when the word boundaries are designated as anything longer than the average interval, while (e) to (h) show the graphs formed when the word boundaries are placed where the interval is twice as long as the average interval. Right: Word graph as part of the solution to a semantics problem, which is what the state transition graphs reminded me of.

Even on the surface, without knowing anything about the specific kinds of esoterica I store in my notebooks, there are visible similarities between the two kinds of graphs. Each number representing a state (or “word”) on the state-transition graphs has an unknown meaning, and the same is true for the yet untranslated words on the graph I had made. The graphs map out specific rules governing what kinds of transitions can be made. My graph is part of the solution for a linguistics problem I had attempted in preparation for the Singapore Linguistics Olympiad. This kind of problem is called a semantics problem, and usually, you are given a set of compound words in a language and a mixed-up group of their translations in English. The goal is to match each of the compound words to its English translation.

This kind of problem is designed to be solvable. You need to separate the compound words into indivisible units of meaning and create a graph showing which connections between the units are allowed. Then, by making a similar graph in English, sometimes separating a word into two theoretical units of meaning (from the example, “dwarf = small person”) you can map one graph onto another and find the meanings of all the words and compound words.

Hours and hours of data are therefore able to give us information which may prove useful in deciphering this possible language. The graphs, however, have one other interesting property: they have certain states which cause the flow to be trapped in a small number of states with no way of getting out. How would we reconcile this with the hypothesis that there are signs of a language? This actually makes a lot of sense. All languages need points that can be used to end a sentence. For an even closer statistical parallel, we can look to Markov chains, which are algorithms that underlie predictive text, such as word suggestions in some documents: they imitate language, and they do often have states which allow for the sequence to end.

In an earlier article by the same authors, statistical analysis was done to sequences to objectively analyse their complexity. There are formulae that can calculate the complexity of binary sequences. Unsurprisingly, most of the methods of calculation showed that human languages (the study used a sample of a podcast in English and one in Mandarin) are more complex than the fungal electrical fluctuations. One algorithm that measured something called Lempel-Ziv complexity seemed to show, however, that fungal languages are more complex than human languages. That is, it is less repetitive. Therefore, the pulses must have some high-level regularity to produce the rhythmic pulses recorded by Cosmo Sheldrake, but have great complexity underlying them.

Despite all the mathematics that went into this study, it cannot conclusively prove that fungi have a language. Nevertheless, it raises some very interesting questions.

Infectious utterances

This particular analysis may appear to conspicuously lack any mention of a characteristic of languages which some of you may consider essential: its use in communication. These scientists have identified some of the symptoms of language in fungi, so to speak, but they have not explicitly proven that they are the result of something transmissible.

Many, many plants’ roots are colonised by fungi that live within the root cells in a mutually beneficial relationship. It was found that the electric currents produced by the plant itself are, on the whole, not changed at all despite colonisation by fungi, which themselves produce electric currents, although this result was measured when we couldn’t reach the same levels as precision that we do today. This is not to be expected from plant roots, whose currents are highly sensitive to all sorts of interference. It suggests a very intimate relationship between the fungus and the plant which involves a regulation of electric currents produced by both parties. It was also found that the fungi produce electric currents all on their own, and this is in most cases not reliant on the uptake of ions or nutrients from the surroundings. Therefore, it makes sense to consider the possibility of these electric currents being used in communication.

There are other examples of unexpectedly sophisticated communication in beings considered not to be sentient. At short ranges, plants can communicate with each other. Certain receptors for some volatile chemicals have been identified that would allow plants to perceive the chemical signals of other plants. This would prompt responses through systems that are poorly understood but manifestly complex. A number of experts have even thought it appropriate to call certain chemicals “words.” Furthermore, each plant produces a unique set of chemical compounds both in the air and underground, which can be detected by nearby plants. If a plant is damaged, it releases certain chemical compounds, and other plants can respond to this in different ways, but plants that are most closely related to the damaged individual respond most strongly. Even though these can be explained away as simple chemical interactions, is the human brain not made of the sum of equally simple relationships?

Even more sophisticated systems have been found in animals. The rate of information transfer in ants has been objectively identified, and it has been found that they can count small quantities, add and subtract small numbers, and pass this information to others.

Several experts have claimed that non-human communication systems cannot be considered language because, for instance, they are not flexible enough, they are not abstract enough, or they show too little regional variation. These are objections based on matters of scale, rather than absolute quality, and we may look at them differently as our technology allows us to analyse data in a more discerning way. Some have excluded animal communication systems from the category of languages on the basis that they serve a purely biological function and are not purely for communication. What if, however, language were a phenomenon coming not from culture, from society, or from psychology, but from complexity itself? If so, it would be something that inevitably comes from complex life and serves its purposes.

A brief digression on thoughts, words, and deeds

It might seem strange to consider the fungi to have language if we cannot make any sense of what it is saying; these patterns currently have no meaning to us. But we are not likely to stay ignorant  about this. The field of semantics in linguistics means that we can understand any language if we have a large enough body of content in the correct context. Although the example I provided earlier is a problem designed to be solvable, which is not the case when studying such vastly different organisms, different contexts can be simulated in the lab to arrive at a set of possible meanings.

After all, how would one group of people begin to understand another upon first making contact? We do not face this situation much in the modern world, but it has happened countless times in history, and people have managed to learn unfamiliar languages. Also, babies learn languages without any prior knowledge. With our tools and theory, there is literally no limit to what we can understand. Crossing the language barrier is not as easy as science fiction may have us believe, but it is also not impossible. We do not even have to travel far to encounter alien languages. They are here, inundating the soil and the trees, and potentially a thousand other places we have not thought to look.

From this vantage point we gain the ability to make a speculative leap.

Thoughts are manifested as measurable processes: neurons fire, so tongues wag and limbs move. But the thought itself is inaccessible through scientific means. A thought (just like all the sensations of life) is an experience that is a part of consciousness, one that no one experiences but yourself. Therefore, consciousness is something we know to be true but cannot verify.

For example, I do not need to be conscious to think about and write this article. My senses provide information in the form of words that I have read about the theories surrounding language and consciousness. Then, the information goes along well-travelled but extremely complex pathways in the brain’s language centre.  More neurons fire to produce some connections that create new  content in the form of language, which is distributed along some pathways to make my fingers type out words. All this only shows the physical workings of the brain, and not the subjective experience of the world that makes up my consciousness. I am conscious of my own thinking and typing, but my consciousness does not cause these actions.

Contemplating this might be disturbing: is everything deterministic and predictable? But determinism is not necessary in this explanation. Quantum mechanics shows that the transmission of signals along synapses between neurons is totally based on probability, because of the inherently unpredictable behaviour of tiny particles and not because of the participation of a truly free consciousness. (This is only one interpretation of quantum effects on free will; for another that affirms free will, see [12].)

All in all, despite it seeming otherwise, you have no evidence that I am conscious because everything you have seen is a consequence of purely physical matters. But you assume I am conscious, simply because you are – at least, that is what I assume.

Consciousness thus seems to be purely a spectator sport, and a property that must be assumed. Generally, we consider consciousness to be an emergent property of complex biological systems. That is, if you have a nervous system that is complex enough, a conscious mind would naturally emerge from it, like how language emerges naturally from a society of intelligent organisms. If this is dependent upon complexity, which varies in extent, it seems logical that the extent of consciousness must also vary based on the systems they emerge from. We should not then exclude slightly less complex systems from consciousness just because of a difference in scale. Now that we have analysed the spikes in electric potential in fungi and found them to be very complex patterns, does it make sense for us to assume that fungi are not conscious?

References

1. Potential fungal language:

   1. Adamatzky, Andrew. “Language of Fungi Derived from Their Electrical Spiking Activity.” Royal Society Open Science, vol. 9, no. 4, Apr. 2022, https://doi.org/10.1098/rsos.211926. https://royalsocietypublishing.org/rsos/article/9/4/211926/96736/Language-of-fungi-derived-from-their-electrical

2. Complexity analysis of potential fungal language:

   1. Dehshibi, Mohammad Mahdi, and Andrew Adamatzky. “Electrical Activity of Fungi: Spikes Detection and Complexity Analysis.” Biosystems, vol. 203, May 2021, p. 104373, arxiv.org/pdf/2008.10276.pdf, https://doi.org/10.1016/j.biosystems.2021.104373.

3. More information on Cosmo Sheldrake:

   1. https://www.toa.st/blogs/magazine/sounds-of-a-healthy-ecosystem-with-musician-cosmo-sheldrake

4. Solving problems in linguistics:

   1. Neacșu, Vlad A. Linguistics Olympiad. Language Science Press, 14 May 2024. https://books.googleusercontent.com/books/content?req=AKW5Qad8wmHPQKPVdcjy_gjY_SjUM5Fw01eWn2rjrJH1vErRYHbdi4M0_G_n1yQMT4fVigNfGZTz14lK9tQBY_EYWaiTav4ir5s1XRCBhriyTLTBGo1SqGN0hTB8lSW7N_3_TTunjqdkdtaAkEGhNz3nalxqbMvtmklp3SKWLJkpGnPJN5FphwePnGfKGMs8q-1NGTo7gZenJwnXhn9O4_jPndd4pAuteL184gvdAkooPZgtYvb8wdOabvAUkpIV6cEEHdq6FN6AQuwsXOZVwlaZJDz2W5eIIA

5. Application of absorbing Markov Chains in text generation (see section 2, “Preliminaries”)

   1. “Rainbow Padding: Mitigating Early Termination in Instruction-Tuned Diffusion LLMs.” Arxiv.org, 2025, arxiv.org/html/2510.03680v1. Accessed 16 June 2026.

6. Signs of intelligence in plants

   1. Trewavas, Anthony. “Intelligence, Cognition, and Language of Green Plants.” Frontiers in Psychology, vol. 7, 26 Apr. 2016, https://doi.org/10.3389/fpsyg.2016.00588.

7. Plant “words”

   1. Šimpraga, Maja, et al. “Language of Plants: Where Is the Word?” Journal of Integrative Plant Biology, vol. 58, no. 4, 22 Feb. 2016, pp. 343–349, https://doi.org/10.1111/jipb.12447. Accessed 11 Apr. 2021.

8. Electric interactions between arbuscular fungi and plant roots

   1. BERBARA, R. L. L., et al. “Electrical Currents Associated with Arbuscular Mycorrhizal Interactions.” New Phytologist, vol. 129, no. 3, Mar. 1995, pp. 433–438, https://doi.org/10.1111/j.1469-8137.1995.tb04314.x.

9. Mathematics in ants

   1. Ryabko, Boris, and Zhanna Reznikova. “The Use of Ideas of Information Theory for Studying “Language” and Intelligence in Ants.” Entropy, vol. 11, no. 4, 10 Nov. 2009, pp. 836–853, https://doi.org/10.3390/e11040836.

10. Reasons why language might be exclusive to humans

    1. Beecher, Michael D. “Why Are No Animal Communication Systems Simple Languages?” Frontiers in Psychology, vol. 12, 19 Mar. 2021, https://doi.org/10.3389/fpsyg.2021.602635.

11. Quantum neurological investigation on uncertainty in the brain’s processes

    1. Georgiev, Danko D. “Quantum Propensities in the Brain Cortex and Free Will.” Biosystems, vol. 208, 1 Oct. 2021, pp. 104474–104474, https://doi.org/10.1016/j.biosystems.2021.104474.

12. Quantum neurological discussion on how “causal efficacy” (free will) is caused by consciousness

    1. Schwartz, Jeffrey M, et al. “Quantum Physics in Neuroscience and Psychology: A Neurophysical Model of Mind–Brain Interaction.” Philosophical Transactions of the Royal Society B: Biological Sciences, vol. 360, no. 1458, 29 June 2005, pp. 1309–1327, www.ncbi.nlm.nih.gov/pmc/articles/PMC1569494/, https://doi.org/10.1098/rstb.2004.1598.

13. Why only humans have language

    1. https://pmc.ncbi.nlm.nih.gov/articles/PMC8018278/

14. Convoluted article on quantum effects and causal efficacy

    1. https://pmc.ncbi.nlm.nih.gov/articles/PMC1569494/

15. Quantum effects on decision-making/free will

    1. https://doi.org/10.1016/j.biosystems.2021.104474