Metastability -  Stable Enough to Exist, Flexible Enough to Change

Perhaps understanding a sentence is already an experience of remaining without becoming trapped

Do you still remember the beginning of this sentence?

Perhaps you could not repeat every word. But something remained long enough for what arrives now to still find meaning.

If nothing remained, we would have isolated signals. If everything remained rigid, no new word could modify what we had already understood.

Brazilian linguist Marcus Maia, from UFRJ, investigates precisely this problem in sentence processing: reading, parsing, ambiguities, predictions, reanalysis, and the temporal course of comprehension. At LAPEX, the sentence itself becomes an experimental space for observing how we construct meaning while moving through a text.

Comprehension does not wait for the period.

We are already building something as we move.

And sometimes a new word reorganizes what we thought we had understood.

Perhaps understanding a sentence is already a small experience of stability and change.

A word does not arrive in an empty brain

Read:

mother

Now do not immediately continue.

Perhaps only the concept appeared.

Perhaps a face.

A voice.

A kitchen.

A smell.

An absence.

Now imagine that, instead of “mother,” you saw your own mother’s name.

fMRI studies show that personally familiar names engage the semantic system but may carry representations enriched by autobiographical and episodic details. Researchers found stronger activation for personally familiar names than for famous names.

The word does not contain the memory.

It encounters an organism that already has a history.

In the BrainLatam hypothesis, each word can function as a symbolic technology: a tool for making distinctions, remembering, anticipating, and reorganizing.

Not for Being.

“Pain” does not hurt.

“Tree” does not cast a shadow.

“Mother” does not contain your mother.

BrainLatam has previously formulated the word as a “tool to use,” distinguishing the linguistic tool from what we are and from what it may recruit within our experience.

But a word can modify the state in which the next word will be received.

The face was the same. The encounter was not.

There is an experiment that makes this idea almost visible.

Eiserbeck and colleagues showed participants happy, neutral, and angry faces while recording EEG.

All of the faces were real.

Before each image, however, participants saw:

REAL

or

FAKE.

Then came an interval, and only afterward did the face appear.

The smile was physically identical.

But the processing was not.

Smiles believed to be “FAKE” were rated as less positive, took longer to evaluate, and no longer produced some of the emotional effects observed for smiles believed to be real in components such as P1, N170, and EPN; later differences also appeared in the LPP.

The word did not alter the photograph.

It changed the encounter with the photograph.

This does not mean that any word can program any brain.

It means that previous context can modify how the next stimulus will be processed.

Perhaps the encounter with an image begins before the image.

And perhaps the meaning of a sentence begins long before its final word.

How much of what just happened is still happening?

This is where our approximately 3.1-second window appears.

At BrainLatam, we use 3.1 seconds as an operational window for thinking about the integration of conscious experience in the first person.

Not as a universal brain stopwatch.

This choice directly connects with a Brazilian research tradition.

In 2000, Alfredo Pereira Jr. and Armando Freitas da Rocha discussed perceptual integration across intervals ranging roughly from 100 milliseconds to 3 seconds, depending on the modalities involved and processing demands. They also made a caution that becomes even more important for us today: defining one invariant temporal unit for all perception would not be viable.

In 2017, Pereira Jr., Flávia Foz, and Rocha distinguished rapid perceptual processes from more complex conscious episodes. Their paper discusses psychophysical studies suggesting approximately 3 seconds for the formation of complex episodes and, within the authors’ specific model, proposes approximately 2 seconds for the formation of a unitary conscious episode.

Therefore:

3.1 seconds is not a constant discovered by these authors.

It is a BrainLatam operational window inspired by this research tradition.

And the question remains contemporary. In 2024, Zhao and colleagues found a temporal transition zone mainly within the 2–3 second range in duration-reproduction tasks, experimentally revisiting the idea of an approximately three-second window related to the subjective present.

The BrainLatam question is more interesting than the number itself:

how much of what an event initiated may still be participating in the Body-Territory when the next event arrives?

The next word encounters what remained

Recent intracranial recordings bring this question closer to language.

Regev and colleagues identified neural populations within the language network with different temporal integration windows, averaging approximately one, four, and six words. These populations were intermingled throughout the language network itself.

In other words, the present word may be processed while different portions of the recent past remain available.

Perhaps we can imagine:

Word₁ → modifies the field

Word₂ → encounters that field

Word₃ → encounters an already modified field

Preactivation is not meaning.

It participates in the territory in which meaning may happen.

When the final word arrives, it does not encounter exactly the same system that received the first.

When a tool becomes a place

Now a trap appears.

We learn words such as attractor, criticality, and metastability.

And we may begin seeing everything through them.

The word “metastability” itself can become a semantic attractor: an elegant explanation for anything that remains and anything that changes.

At that point, perhaps we have only found a new local optimum of language.

A local optimum answers a question of optimization:

this appears better than what is immediately around it, although a better solution may exist somewhere else in the landscape.

An attractor answers a different question:

why does the system tend to remain in, or return to, a particular configuration?

Attractor-network models help explain persistent activity, working memory, cue integration, and resistance to perturbation.

But metastability is not a sophisticated synonym for “working well.”

It does not automatically mean health, freedom, creativity, or consciousness.

Hancock and colleagues emphasize that the term metastability has accumulated different—and sometimes imprecise—uses in neuroscience. In a useful sense, it should help us investigate how transient configurations can possess enough stability to exist without eliminating the possibility of reorganization.

The problem is not finding a place that works.

It is forgetting that it is a place—and beginning to call it the world.

Criticality is not metastability

Another word appears nearby:

criticality.

In complex-systems science, criticality refers to properties associated with proximity to a transition between regimes. In neuroscience, proposals include avalanche criticality and edge-of-chaos criticality. A review by O'Byrne and Jerbi argues that near-criticality may be biologically more plausible than imagining the brain permanently positioned exactly at a critical point.

Near a transition, the relationship between perturbation and response can change.

But that is not metastability.

Criticality asks:

“How close are we to a change of regime?”

Metastability asks:

“How can a configuration persist temporarily without completely eliminating the possibility of other configurations?”

They are not synonyms.

And perhaps it is important not to allow our own concepts to trap us.

If “metastability” can explain permanence, change, consciousness, health, freedom, creativity, and performance all at once, perhaps it is no longer distinguishing very much.

A conceptual tool that explains everything risks explaining nothing.

How many next states are still possible?

Now the question begins to acquire an experimental dimension.

Jang and colleagues studied wakefulness, propofol anesthesia, and sleep, analyzing the dynamic balance between integration and segregation in brain networks. During loss of responsiveness, they observed greater segregation and changes in metastability; the authors operationalized metastability through the dynamic recurrence of transient non-equilibrium states.

Castro, Luppi, Tagliazucchi, Perl, Cofré, and colleagues found something complementary. During unconsciousness produced by anesthesia or slow-wave sleep, brain dynamics were more strongly dominated by a recurrent pattern closely related to structural connectivity and showed a reduced capacity to transition toward other patterns.

This does not mean:

more change = more consciousness.

Perhaps the better question is:

how many next states remain genuinely accessible?

And here our concern about the local optimum returns.

A configuration may work extraordinarily well right now.

But the territory may change.

Perhaps performance is not about finding the perfect state.

Perhaps it is about being able to use an adequate state without completely losing the repertoire of exits.

An invitation to pause

Return to an important word.

And wait.

Perhaps a memory comes.

Perhaps an image.

Perhaps nothing comes.

The BrainLatam hypothesis is not that a pause automatically releases rigid attractors, increases entropy, or moves the brain into a “better” state.

It is simpler:

regulating the speed at which stimuli arrive may change how much time processes initiated by one sign have to participate in the state that will receive the next one.

This can be tested.

A personally meaningful word.

No pause.

One second.

Approximately three seconds.

Then we can observe EEG, microstates, memory, interpretation, and first-person reported experience.

The question would no longer be only:

“Which word arrived?”

It would also include:

“What was still happening when the next one arrived?”

Perhaps “let it stay” never meant remaining motionless.

While something stays, a memory may move closer.

A prediction may lose strength.

Another interpretation may become possible.

Let it come.

Let it stay.

Let it be.

You can disobey.

Let it go.

This sequence appears in BrainLatam as a grammar of agency, not as a neuroscientific definition of metastability.

Remaining can also participate in transformation.

Perhaps understanding a sentence requires exactly this:

something from the past remains long enough to participate in the present, but not so rigidly that the present can no longer reorganize it.

And perhaps Jiwasa appears here.

I remain I.

You remain you.

This word does not encounter exactly the same territory in you that it encountered in me.

Even so, we remain together long enough for a sentence to exist.

Enough integration to form a we.

Enough difference so that no one needs to disappear inside it.

Perhaps living is not about remaining in the best state.

Perhaps it is about continuing to possess more than one possible next step.

Stable enough to exist.

Flexible enough to change.

Commented References

Maia, M. (2022). Eye Tracking Sentences in Language Education. Diacrítica, 36(1), 6–36.
What this reference represents: it places Marcus Maia and Brazilian Psycholinguistics at the center of the argument. The sentence is not treated as a finished product, but as something whose processing can be followed through time using fixations, regressions, and local processing difficulty. For W40/2026, this allows us to begin from the experience of preserving and reorganizing meaning while reading.

Desai, R. H., Tadimeti, U., & Riccardi, N. (2023). Proper and Common Names in the Semantic System. Brain Structure and Function, 228, 239–254. DOI: 10.1007/s00429-022-02593-9.
What this reference represents: it supports the example involving a mother's name. Personally familiar names showed representations enriched by autobiographical and episodic information compared with merely famous names. A personally meaningful word encounters a history.

Eiserbeck, A., Maier, M., Baum, J., & Abdel Rahman, R. (2023). Deepfake smiles matter less — the psychological and neural impact of presumed AI-generated faces. Scientific Reports, 13, 16111. DOI: 10.1038/s41598-023-42802-x.
What this reference represents: it makes the effect of prior context visible. The images were real, but REAL/FAKE information altered evaluation and EEG responses to smiling faces. The word did not change the face; it changed the encounter.

Regev, T. I., Casto, C., Hosseini, E. A., et al. (2024). Neural populations in the language network differ in the size of their temporal receptive windows. Nature Human Behaviour, 8, 1924–1942. DOI: 10.1038/s41562-024-01944-2.
What this reference represents: it shows that populations within the language network integrate context at different scales, with average windows of approximately one, four, and six words. The present word may arrive while different portions of the preceding sequence remain available.

Khona, M., & Fiete, I. R. (2022). Attractor and integrator networks in the brain. Nature Reviews Neuroscience, 23, 744–766. DOI: 10.1038/s41583-022-00642-0.
What this reference represents: it gives precision to the concept of an attractor. Attractor networks help explain persistent activity, working memory, error correction, and cue integration. An attractor explains aspects of persistence; it is not synonymous with a local optimum or with metastability.

O'Byrne, J., & Jerbi, K. (2022). How critical is brain criticality? Trends in Neurosciences, 45(11), 820–837. DOI: 10.1016/j.tins.2022.08.007.
What this reference represents: it prevents criticality from becoming another all-purpose concept. The review discusses different forms of criticality and argues that near-criticality may be more plausible than assuming an exact and permanent critical state. Criticality and metastability are neighboring questions, not the same answer.

Jang, H., Mashour, G. A., Hudetz, A. G., & Huang, Z. (2024). Measuring the dynamic balance of integration and segregation underlying consciousness, anesthesia, and sleep in humans. Nature Communications, 15, 9164. DOI: 10.1038/s41467-024-53299-x.
What this reference represents: it operationalizes relationships between integration, segregation, metastability, and states of consciousness. Metastability moves beyond metaphor and becomes something that can be investigated in brain-dynamics data.

Castro, P., Luppi, A., Tagliazucchi, E., Perl, Y. S., Naci, L., Owen, A. M., Sitt, J. D., Destexhe, A., & Cofré, R. (2024). Dynamical structure-function correlations provide robust and generalizable signatures of consciousness in humans. Communications Biology, 7, 1224. DOI: 10.1038/s42003-024-06858-3.
What this reference represents: it places Latin American researchers directly within the dynamical study of consciousness. During unconscious states, one recurrent pattern dominated the dynamics more strongly and there was less capacity to transition toward other patterns. The question becomes not only which state exists, but which states can still be reached.

Hancock, F., Rosas, F. E., Luppi, A. I., et al. (2025). Metastability demystified — the foundational past, the pragmatic present and the promising future. Nature Reviews Neuroscience, 26(2), 82–100. DOI: 10.1038/s41583-024-00883-1.
What this reference represents: it is the main conceptual anchor of the article. The authors show that metastability has distinct meanings and requires careful definitions and measurements. If metastability begins to explain every form of stability and every form of change, it has become a conceptual local optimum that is too comfortable.

Zhao, C., Mu, N., Zhang, J., & Bao, Y. (2024). The temporal transition zone: A gradual approach to a subjective set-point within the three-second time window. PsyCh Journal, 13(3), 369–375. DOI: 10.1002/pchj.755.
What this reference represents: it offers a recent update to the discussion of an approximately three-second temporal window. The study identified a temporal transition zone mainly between 2 and 3 seconds. It does not validate 3.1 seconds as a universal constant, but it shows that this timescale remains a legitimate experimental question.

Foundational references — a deliberate exception to the post-2021 rule

Pereira Jr., A., & Rocha, A. F. (2000). Temporal Aspects of Neuronal Binding. In R. Buccheri, V. Di Gesù & M. Saniga (Eds.), Studies on the Structure of Time: From Physics to Psycho(patho)logy, pp. 97–106. Kluwer Academic/Plenum Publishers.
What this reference represents: it is one of the Brazilian foundations for the temporal window used by BrainLatam. Pereira Jr. and Rocha discuss a range of approximately 100 ms to 3 s and, crucially, reject the idea of one invariant temporal unit for all perception. That caution is precisely why we treat 3.1 seconds as an operational window rather than a universal clock.

Pereira Jr., A., Foz, F. B., & Rocha, A. F. (2017). The Dynamical Signature of Conscious Processing: From Modality-Specific Percepts to Complex Episodes. Psychology of Consciousness: Theory, Research, and Practice, 4(2), 230–247. DOI: 10.1037/cns0000115.
What this reference represents: it distinguishes rapid perceptual processing from complex conscious episodes constructed over seconds. The article discusses evidence around 3 s and, within its own model, proposes approximately 2 s for a unitary conscious episode. Complex consciousness appears as a temporal composition rather than an isolated instant.

BrainLatam. (2024). É pra te entreter. Que é pra você não ver que o programado é você!
What this reference represents: it records the formulation “the word is a tool to use” and the idea that words can recruit semantic and episodic associations without needing to become identity. In this article, that formulation is placed in dialogue with memory, language, and brain dynamics.

BrainLatam. (2026). Carnaval é Liberdade de Vir, Ficar, Ser, Desobedecer e Ir.
What this reference represents: it provides the practical grammar of agency revisited throughout the text. To come, stay, be, be able to disobey, and go is not a scientific definition of metastability; it is a BrainLatam formulation for remembering that remaining in a state does not have to turn that state into destiny.

W40/2026 - Latent Heat, Homeostasis and Metanoia