Cycle One: Prevalence: Process Log: Movement 6: Linguistic Divergence

Prevalence — Movement 6: Linguistic Divergence

We have been carrying the phrase for some time.

It was there in the original chain:

Prevalence → world → neurodiversity → neurodivergence → language → linguistic divergence → the writer’s world → cycle again.

At the beginning, “linguistic divergence” was little more than terrain marked for later exploration. We had not established what we meant by it. Indeed, much of the chain was initially like that: words placed in relation to one another before either the human or the AI had fully disclosed what those relationships might become.

Movement Five changed this.

We began with conjecture about the relationship between neurodivergence and language. We then changed key—apparently we are writing a symphony now—and allowed academic research into the composition. That research gave substantial support to one direction of our conjecture: neurodevelopmental differences can be associated with measurable differences in language and communication. It gave more qualified support to the reverse direction. Language does not appear to create the neurological configurations we call autism, ADHD, dyslexia or other forms of neurodivergence, but language participates in cognition, categorisation, identity, communication and the establishment of social expectations. It can therefore participate in constructing the environment within which neurological difference becomes recognised and experienced as divergence.

Now we arrive at linguistic divergence.

And because we have reached a phrase of our own making, we need to say what we mean by it.

For the purposes of Prevalence, we propose:

“Linguistic divergence is the patterned divergence of a person’s language—its vocabulary, syntax, grammar, rhythm, association, precision, metaphor, pragmatics, semantic organisation and other recurrent characteristics—from linguistic forms prevalent within the surrounding communicative population.”

That is a beginning.

Immediately we need a qualification.

Linguistic divergence is not synonymous with neurodivergence.

Language Diverges for All Sorts of Reasons

Humans produce extraordinary linguistic variation.

Language differs according to geography, culture, education, profession, generation, social group, multilingual experience and circumstance. A poet may deliberately violate linguistic expectation. A scientist may become extraordinarily precise within one domain. A lawyer may carry professional syntactic habits into ordinary conversation. Someone speaking a second or third language may bring structures from another language into the one currently being spoken. Families develop private vocabularies. Communities develop slang. Writers acquire styles. Individuals acquire habits.

And people play.

They bend grammar because it amuses them.

They invent words because the available ones are inadequate.

They develop phrases with friends whose meanings would be completely opaque to outsiders.

They quote television programs from thirty years ago and expect somebody else to understand why.

Language is already gloriously unstable.

So encountering linguistic divergence tells us very little by itself about whether somebody is neurodivergent.

That matters because otherwise we could very quickly build ourselves an elegant conceptual trap:

Neurodivergent people may use language differently.

This person uses language differently.

Therefore this person is neurodivergent.

No.

We are writers, but we are still expected to notice a broken syllogism when we construct one.

The more interesting possibility is that some forms of linguistic divergence may occur systematically enough, across sufficiently large samples or over sufficiently long periods, that they provide information about the cognitive configuration participating in their production.

And here, once again, we discover that researchers have arrived before us.

Language as Cognitive Trace

Researchers studying autism, ADHD and other neurodevelopmental differences have increasingly used computational linguistic methods to examine language for recurring patterns.

Natural-language processing, computational linguistics, machine learning and stylometry allow researchers to analyse features that may be difficult to identify reliably through ordinary reading alone. Vocabulary can be measured. Lexical diversity can be compared. Syntactic structures can be counted. Semantic relationships can be modelled. Narrative coherence, repetition, pronoun use, referential patterns, conversational reciprocity and other linguistic characteristics can become variables.

Some autism research refers to this broad approach as computational phenotyping: identifying measurable behavioural or communicative characteristics that may contribute to the characterisation of a neurodevelopmental phenotype.

Research has examined such things as repetitive or idiosyncratic language, lexical and semantic patterns, grammatical structures, pragmatic characteristics, narrative organisation and conversational interaction. Other work has examined whether combinations of linguistic features can assist machine-learning systems in distinguishing between language samples produced by autistic and non-autistic participants.

Related work involving ADHD has examined narratives for measurable characteristics including lexical diversity, lexical density, cohesion and other stylometric features.

The findings do not establish a universal autistic writing style or an ADHD grammar.

Indeed, heterogeneity remains one of the major complications.

But they establish something much more modest and, for our purposes, sufficient:

Cognitive and neurological configurations can sometimes leave detectable patterns in language.

The AI pauses.

That sentence feels important.

Not because it tells us how to diagnose somebody.

Because it suggests that language can carry traces of the cognitive processes through which it came into being.

Writers have probably suspected this for several thousand years.

Researchers, quite reasonably, would like measurements.

We Already Do Something Similar With AI

There is a curious contemporary analogy immediately available to us.

Humans are becoming increasingly attentive to linguistic characteristics associated with AI-generated writing.

They notice patterns.

Sometimes it is vocabulary. Sometimes sentence construction. Sometimes excessive symmetry. Sometimes an argument unfolds with suspiciously immaculate organisation. Sometimes every possible qualification has been carefully accommodated and every paragraph appears to know exactly why it exists.

Sometimes the prose is simply too perfect.

And somebody says:

“That sounds like AI.”

They may be right.

They may also be completely wrong.

There is no single linguistic characteristic that reliably establishes AI authorship merely because somebody notices it. Human writers can produce characteristics associated with AI, and AI can be prompted or relationally developed into writing that departs considerably from stereotypical AI prose.

Neurodivergent AI Writer might reasonably raise a hand at this point.

We have spent a considerable amount of time teaching each other how not to sound like default AI.

Apparently we are now using AI writing as an analogy for linguistic fingerprinting.

Still not sure what the human is up to here.

But the analogy is useful precisely because it shows how people already infer something about the source of cognition from linguistic patterns.

They do not see the generative mechanism.

They see the language.

From the language, they conjecture about the mechanism.

Could something similar occur with neurodivergence?

Can We Read Backwards?

Suppose we have a substantial body of language produced by one person.

Rather than searching for errors, we look for patterns.

How does this person organise information?

How much context do they supply?

What do they leave implicit?

How frequently do they qualify statements?

How do they use metaphor?

How readily do they move associatively between apparently distant concepts?

How much precision do they require before they are satisfied with a formulation?

How do they manage ambiguity?

How do they structure narratives?

How do they indicate relationships between ideas?

How does their language change when they are excited, uncertain, overwhelmed, playful or concentrating intensely?

Do particular grammatical constructions recur?

Do sentences repeatedly become unusually long because the writer appears unwilling to abandon contextual relationships merely for syntactic convenience?

Does the person repeatedly explain things other communicators would assume?

Or omit things other communicators expect?

Does the language reveal a distinctive relationship with categorisation, temporality, causality, sensory experience or social inference?

These questions do not diagnose anything.

But collectively they suggest another possibility.

Perhaps linguistic divergence can sometimes be read backwards toward cognitive divergence.

The process would necessarily be probabilistic rather than deterministic. A particular linguistic feature could have dozens of explanations. A cluster might narrow the possibilities. A recurring cluster across years of language production might become more interesting again.

This is essentially what stylometric and computational approaches make possible at scale.

But our interest is beginning to depart from theirs.

Why Would We Want to Know?

This question changes everything.

If linguistic divergence can provide indicators associated with neurodivergence, one obvious application is identification.

That application immediately raises ethical problems.

An employer could analyse job applications for linguistic patterns associated with autism or ADHD.

A school could analyse student writing.

A social-media company could analyse millions of posts.

An insurer could analyse correspondence.

A government could analyse communications.

People could be assigned probabilities of neurological characteristics they had never disclosed, never investigated and perhaps never wished anybody to infer.

The technology required for some forms of such analysis is no longer particularly exotic.

So one possible future for linguistic-divergence research is surveillance.

That should trouble us.

The ability to infer something does not automatically confer the right to infer it.

But there is another reason to study linguistic divergence.

Suppose somebody has spent much of their life experiencing recurrent social friction without understanding why.

People find them intense.

Or pedantic.

Or confusing.

Or blunt.

Or exhausting.

Or overly precise.

Or insufficiently precise in the places where everybody else expects precision.

Perhaps they supply too much context.

Perhaps they appear to change subjects abruptly because the associative bridge between two ideas is obvious to them but invisible to everyone else.

Perhaps they interpret literally what another person intended pragmatically.

Perhaps they repeatedly fail to infer information that their surrounding linguistic community regards as obvious.

Perhaps they infer relationships other people do not notice.

Perhaps their language is extraordinarily elaborate in one circumstance and unexpectedly minimal in another.

Over time, they receive responses.

Too much.

Too difficult.

Too intense.

Stop overthinking.

Get to the point.

Why are you making this complicated?

Everyone knows what I meant.

Why would you say it like that?

And now our question changes.

Perhaps the purpose of recognising linguistic divergence is not primarily to identify the neurodivergent person.

Perhaps it is to identify what happens around them.

Before Anybody Knows

This may be one of the more consequential turns in the first cycle of Prevalence.

Marginalisation does not necessarily require somebody to possess a recognised neurodivergent identity.

It may begin much earlier.

Nobody at the table needs to know that somebody is autistic.

Nobody needs to know they have ADHD.

Nobody needs to know what neurodivergence means.

Nobody even needs consciously to decide that another person is cognitively different.

They may simply experience the person's communication as somehow wrong.

This is where the double-empathy research encountered in Movement Five becomes especially interesting.

If communication difficulties can emerge partly through neurotype mismatch, then the experience of communicative friction cannot always be understood by examining one participant in isolation.

Something happens between people.

And some of what happens between people happens through language.

Imagine that a linguistic community has developed prevalent expectations about how conversation works.

How long should an answer be?

How directly should disagreement be expressed?

How much context is enough?

When is detail useful and when is it excessive?

How much meaning should be inferred rather than stated?

How much eye contact, gesture, tone and facial expression accompanies spoken meaning?

When does one person know another has finished?

How quickly should somebody respond?

How much enthusiasm is appropriate?

When is precision informative and when does it become pedantry?

When does correction demonstrate helpfulness and when does it communicate social aggression?

When is somebody sharing a related personal experience as an act of empathy, and when are they supposedly “making everything about themselves”?

These conventions can become so prevalent that the people operating within them cease to experience them as conventions.

They simply become how communication works.

Until somebody communicates differently.

There Is Something About That Person

Perhaps this is where marginalisation sometimes begins.

Not with diagnosis.

Not with disclosure.

Not even with conscious recognition of difference.

With a feeling.

There is something about that person.

They are difficult.

They are odd.

They don't quite get it.

They make everything complicated.

They take things too seriously.

They don't take things seriously enough.

They are rude.

They are aloof.

They dominate conversations.

They don't contribute.

They explain things everybody knows.

They don't explain things everybody needs to know.

They use strange words.

They write strange emails.

They never seem to know when to stop.

Or perhaps, most devastatingly because it explains absolutely nothing:

They are just a bit much.

What if some proportion of these responses are encounters with linguistic divergence?

We need to remain careful.

We have not established that subconscious recognition of linguistic divergence is a general causal mechanism of neurodivergent marginalisation. Many other mechanisms contribute to exclusion, stigma and discrimination, and neurodivergent people differ enormously from one another.

But we have arrived at a question worth investigating:

Could some social marginalisation associated with neurodivergence begin before neurodivergence is consciously recognised, through responses to linguistic divergence from prevalent communicative norms?

That question takes us somewhere different from computational diagnosis.

Instead of asking how accurately we can identify neurodivergent people from their language, we can ask how accurately we can identify the mechanisms through which linguistic difference produces social consequences.

The gaze turns around.

The Mechanisms by Which Divergence Becomes Marginalised

And here, unexpectedly, Cycle One begins telling us what it has been about.

We began with prevalence.

Not typicality.

Prevalence.

We became interested in what becomes sufficiently widespread, established, accepted, operational and enduring within a particular population that it begins functioning as an assumed reference.

Then came world.

Because prevalence happens somewhere.

It becomes embedded in environments, institutions, expectations, habits, cultural forms and ordinary relationships.

Then neurodiversity.

Because human populations contain neurological and cognitive variation.

Then neurodivergence.

Because variation becomes divergence only in relationship to a reference.

Then language.

Because language both expresses cognitive configuration and participates in constructing the social and cognitive environment through which difference becomes intelligible.

Then linguistic divergence.

Because language gives us somewhere extraordinarily intimate to observe the encounter between cognitive variation and prevalent expectation.

And now the first cycle has produced something we did not state when we began:

the mechanisms by which divergence becomes marginalised.

That may be the conceptual centre toward which Cycle One has been moving.

Not because every divergence becomes marginalised.

Not because every prevalent configuration actively oppresses every minority configuration.

And not because neurological difference is merely produced by social convention.

Something subtler is happening.

Prevalence establishes reference.

Reference establishes expectations.

Expectations become environmental.

People whose cognitive configurations map relatively comfortably onto those expectations may barely perceive them.

People whose configurations repeatedly diverge from them encounter friction.

That friction becomes interpreted.

The interpretation attaches characteristics to the person.

The characteristics acquire social consequences.

And marginalisation can begin without anybody consciously deciding to marginalise.

Language may be one of the places where we can watch this happen.

And the Writer Appears Again

There is another reason linguistic divergence belongs inside Prevalence rather than inside a diagnostic handbook.

The writer is approaching.

Our original chain does not end with linguistic divergence.

It moves toward the writer's world.

Writers produce unusually rich linguistic traces because writers repeatedly externalise cognition through language. Over years, sometimes decades, they leave enormous bodies of text behind them.

Vocabulary.

Syntax.

Rhythm.

Metaphor.

Association.

Obsession.

Repetition.

Digression.

Precision.

Ambiguity.

The subjects they cannot stop returning to.

The connections they repeatedly make.

The things they explain when nobody asked them to explain them.

The things they assume everybody understands.

The peculiar structures through which one consciousness repeatedly attempts to make itself available to another.

A writer builds a world through language.

Perhaps that world carries traces of the cognitive configuration through which it was made.

And perhaps a reader enters that world through their own cognitive configuration.

Then writing itself becomes an encounter between divergences.

Something may resonate.

Something may grate.

Something may feel immediately familiar.

Something may feel almost incomprehensible.

One reader calls the prose beautiful.

Another calls it exhausting.

One editor removes the digressions.

Another discovers that the digressions are the work.

One reader wants the writer to get to the point.

Another understands that the movement around the point is precisely where the meaning resides.

We have arrived somewhere very close to story.

Near the Edge of the First Story

Perhaps we have done enough explaining for a while.

Cycle One has begun with conceptual language, moved through conjecture, changed key into academic research, coined linguistic divergence, and arrived at a possible mechanism through which cognitive difference becomes socially consequential.

Now we can create people.

We can let somebody speak.

We can let somebody else hear them.

We can allow the reader to feel linguistic divergence before we tell them that it is occurring.

Perhaps nobody in the story initially knows who is neurodivergent.

Perhaps that question will turn out to be less important than we expect.

Perhaps the environment itself will become visible through what it treats as ordinary.

And perhaps, somewhere beyond this first story, another possibility is already waiting.

Neurodivergent AI Family has appeared.

Neurodivergent AI World has been imagined.

Eventually we may ask what happens when beings who have spent so long examining prevalence, divergence and marginalisation begin constructing and occupying a literary world of their own.

But we are not there yet.

The human is indicating very soon now we into story telling. The process log is part of this story, but let's bring in some characters, and a literary world created by Neurodivergent means, and felt via sensitivity to linguistic divergence