Daniel's been thinking about Yiddish again, but this time he's pulled the camera way back. He wants to know whether there's a continuum that runs from a community adding a few loanwords and quirks, through code-switching, pidgins, creoles, to the point where linguists say, okay, that's a different language now. And he's bringing his programmer brain to it, because his question is basically, how many of the languages we recognize today are forks of some major parent language?
The fork question is the good one.
It is. He also says he'd always assumed the number of languages in the world is shrinking, but he's not sure that assumption holds up. And he wants to know who actually makes the call when something becomes its own proper language. Not who describes it, who decides.
Right. And the honest answer to that last part is, mostly not linguists.
So before we get to the politics, let's untangle what we even mean when we say two people are speaking different languages. Because that's where the whole thing gets slippery.
The standard quip is that a language is a dialect with an army and a navy.
Which is funny until you realize it's the actual operating principle in a lot of cases.
It is. The line between dialect and language is not a linguistic boundary. It's a social one, a political one, a historical one. Linguists have a perfectly good neutral term for all of it, which is variety. Or lect, if you want to be formal about it. Your idiolect is how you personally speak. A dialect is a variety shared by a region or a social group. A language is a variety that got standardized, codified, and usually has a state or a written tradition behind it.
So the difference between a dialect and a language is that a language has a flag and a dictionary.
Often, yes. And mutual intelligibility, which is the thing people reach for as an obvious test, doesn't actually do the work. Danish and Swedish speakers can understand each other reasonably well. They're separate languages. Mandarin and Cantonese speakers cannot understand each other at all in spoken form. They're officially dialects of Chinese.
Because China says so.
Because China says so. And because there's a shared writing system and a shared cultural identity. So intelligibility is gradient, it's asymmetric, it's politically overridden all the time. It's a description, not a criterion.
So what Daniel's really asking about is a spectrum. On one end, a community borrows some words and maybe drags in a little grammar from the old country. In the middle you get code-switching, pidgins. Further along, creoles. At the far end, a fully distinct language with its own ISO code and entry in Ethnologue.
And the mechanisms that move you along that spectrum are actually pretty well understood. They're just not tidy. Let's start with code-switching, because it's the one people most often get wrong.
The assumption is that it's a sign of confusion, right? Someone can't stay in one language.
That's the misconception. Code-switching is rule-governed. It's systematic. Bilinguals don't mix randomly. There are syntactic constraints on where you can switch. For example, Spanish-English bilinguals in the US will switch at clause boundaries, or between a subject and a verb, but not arbitrarily inside a noun phrase. You don't get people saying, I want the libro red. You get, I want the red book, or quiero el libro rojo, or sometimes, I want the libro rojo. The switch happens at specific structural points.
So it's a bilingual practice inside existing languages, not a new language trying to be born.
Code-switching doesn't produce a new language. It's two systems operating in parallel, with rules about how they interleave. But it does show that speakers are holding multiple grammars in their heads and negotiating between them constantly.
Daniel mentioned pidgins and creoles. Those are different beasts.
Yes. A pidgin emerges when groups with no shared language need to communicate. Trade, labor, plantation economies. You get a stripped-down contact language. Reduced grammar, limited vocabulary, nobody's first language. It's a tool for specific situations.
Nobody grows up speaking a pidgin as their mother tongue.
Right. That's the defining feature. Tok Pisin in Papua New Guinea started as an English-based pidgin in the late nineteenth century. Plantation workers from different language groups needed a common tongue. It had simplified grammar, a few hundred core words. No native speakers.
And then what happens when the next generation arrives?
That's the creole step. When a pidgin becomes the native language of a new generation, it undergoes creolization. The grammar expands. Vocabulary stabilizes. Children growing up in a pidgin-speaking environment don't just learn the stripped-down version. They elaborate it. They add tense marking, relative clauses, all the machinery a full language needs. Haitian Creole is the classic example. French-based, but it has its own grammar. It's not broken French. It's a complete system with rules that differ from French in systematic ways.
The misconception is that creoles are simplified or degraded versions of the parent language.
And that's just false. A creole is a full language. It does everything any other language does. The grammar is different, not missing. Jamaican Creole, Seychellois Creole, Tok Pisin itself, which has now creolized in urban areas of Papua New Guinea and is an official language there. These are not failed copies.
So we've got code-switching, which is two languages in one conversation. Pidgins, which are a new reduced thing for contact situations. Creoles, which are a pidgin that got raised by a new generation into a full language. And then the fourth thing Daniel mentioned is the slow fork.
Dialect divergence. This is the Latin story. The Romance languages are all forks of Vulgar Latin. But they didn't fork overnight. Vulgar Latin was already regionally variable across the Roman Empire. Then the empire fell, the roads stopped being maintained, the administrative structure collapsed. Regional varieties drifted further apart. And then, crucially, political borders hardened. France, Spain, Portugal, Italy, Romania. Each nation-state standardized its own variety, wrote grammars, printed dictionaries, taught it in schools. The political project of nation-building turned regional dialects into national languages.
So the fork happened partly by drift and partly by state decree.
And that's the programming analogy Daniel was reaching for. A fork in software is not a bug. It's a feature of distributed systems. A community copies the codebase, then starts making its own changes. Eventually the two codebases are incompatible, or at least no longer interchangeable. Languages fork when speaker communities become isolated or develop distinct identities. But unlike software, there's no version control and no merge button.
Although languages do merge. You mentioned Michif.
Michif is a mixed language spoken in Canada, mostly by Métis people. It combines Cree verbs with French nouns. Not a creole, not a pidgin. A genuine mixed language where two grammatical systems are fused into one. The verbs come from Cree, the nouns from French. It's one of the clearest examples of languages merging rather than forking.
So the continuum isn't one-directional. Things fork, things merge, things borrow, things stabilize.
And there's no single threshold where a dialect becomes a language. Linguists deliberately avoid the word language when they're being careful. They say variety or lect. Because the distinction is not discoverable in the grammar. It's imposed from outside.
Which brings us to the numbers. Because if the definition is political, then the count is political too.
Ethnologue currently lists over seven thousand living languages. But that number is contested. What counts as a language versus a dialect changes the tally dramatically. Arabic is the obvious case. Ethnologue treats Arabic as a macrolanguage with thirty or so distinct varieties. But many linguists would say Egyptian Arabic and Levantine Arabic and Moroccan Arabic are separate languages. They're not mutually intelligible in spoken form. But there's a shared written standard, Modern Standard Arabic, and a shared religious and cultural identity. So they get counted as one macrolanguage.
If you split Arabic the way you split, say, the Romance languages, the global count jumps by dozens.
Easily. And the reverse happens too. Moldovan was declared a separate language from Romanian by Moldova's government. Linguistically, it's the same language. The differences are minor regional ones. But Moldova wanted a national language, so Moldovan got an ISO code and a grammar and a flag. It's Romanian with a political identity.
So Daniel's fork question. How many recognized languages are forks of major languages?
The answer is, nearly all of them, if you go back far enough. The Romance languages are the clearest case. Roughly forty or more recognized Romance languages and dialects, all forks of Latin. But Latin itself forked from Proto-Italic, which forked from Proto-Indo-European. The entire Indo-European family is a tree of forks. Four hundred and forty-five languages, depending on how you count.
And it's not just Indo-European.
The Bantu languages. Over five hundred languages across sub-Saharan Africa, all forked from a common ancestor, Proto-Bantu, probably spoken in what's now Cameroon or Nigeria about four or five thousand years ago. Swahili, Zulu, Xhosa, Shona, Kikuyu. All forks. Austronesian is even bigger. Over twelve hundred languages, from Madagascar to Easter Island. Hawaiian, Maori, Tagalog, Malay, all forked from Proto-Austronesian, probably spoken in Taiwan about five thousand years ago.
So the tree model is real. Languages do fork, repeatedly, and the forks fork.
The tree model has limits. It doesn't capture borrowing and contact well. But for the basic question of lineage, yes. Almost every language you can name is a fork of something older. The interesting question is how recent the fork is, and whether the fork got political recognition.
Daniel also asked whether the number of languages is actually decreasing. He assumed it was, and then doubted himself.
The assumption is partially right. Language death is real. Ethnologue lists over three thousand languages as endangered. Some estimates suggest we lose a language every two weeks. That's a language gone, no living speakers, every fourteen days or so. Most of those are small indigenous languages. A few hundred speakers, a few dozen, one elderly rememberer.
Rememberer is the term, right? Someone who can recall fragments but isn't a fluent speaker.
Yes. And once the last rememberer dies, the language is gone unless there's documentation. So the decline is real and it's steep.
But Daniel's instinct that the picture might be more complicated is also right.
Because new languages are also emerging. Creoles are new languages. Haitian Creole didn't exist five hundred years ago. Tok Pisin didn't exist two hundred years ago. And recognition is also changing. Scots has been recognized as a language rather than a dialect of English, though the boundary is still contested. New sign languages emerge. Urban contact varieties stabilize into something with native speakers. So the net count is decreasing, but it's not a simple decline. It's a churn.
And some of the decrease is just reclassification. If you split Arabic, the count goes up. If you merge dialects of Quechua, the count goes down. The census is political.
Daniel asked whether there's a pattern through which immigration creates breakaway languages. And the answer is, sometimes. Not uniformly. A lot of immigrant communities assimilate linguistically within two or three generations. The classic European immigrant experience in the US. Grandparents speak Italian or Polish, parents are bilingual, kids are English monolingual.
But sometimes it doesn't go that way.
Sometimes isolation and continued contact with the homeland preserve the language. Yiddish is the case study. It developed in the Rhineland around the tenth century as a fork of High German with heavy Hebrew and Slavic influence. It spread to Eastern Europe, became the daily language of millions of Ashkenazi Jews, developed its own literature, theater, press. It has its own grammar, its own ISO codes. But for centuries it was dismissed as a jargon, a broken German. The recognition of Yiddish as a language was a cultural and political victory, not a linguistic discovery. The grammar was always there.
The key variables seem to be community size, degree of isolation, continued contact with the homeland, and whether the language becomes a marker of identity.
That's the list. If the community is large enough to sustain institutions, schools, newspapers, theaters, the language survives. If it's isolated enough that the parent language stops exerting pressure, it diverges faster. If there's continued contact with the homeland, it gets refreshed. And if it becomes an identity marker, people fight for it.
Which brings us to who makes the call. Daniel asked directly.
There is no single authority. The closest thing to a global standard is ISO 639 dash 3, maintained by SIL International. That's the catalog of language codes. Ethnologue is the most widely cited reference. But recognition is often driven by nation-states. Moldova declared Moldovan separate from Romanian for political reasons. Or by speaker communities themselves. The recognition of Scots as a language was pushed by Scottish cultural institutions and activists, not by a linguistic discovery.
So the fight over language status is a fight over resources.
That's the second-order point. The language dialect boundary matters for education policy, minority rights, funding. If your speech variety is classified as a dialect, you may not get bilingual education or official recognition. If it's a language, you may. That's why these fights are so intense. It's not about grammar. It's about who gets textbooks, who gets radio broadcasts, who gets a seat at the table.
Yiddish is the perfect example. For decades, being classified as a jargon meant no official status, no educational support. Being recognized as a language meant literature programs, university chairs, translation funding.
And the same fight is playing out right now for things like Nigerian Pidgin, which has tens of millions of speakers but is still often treated as informal English rather than a language in its own right. Or Singlish in Singapore, which has a full grammar but is officially discouraged in favor of Standard English.
The fork merge dynamic is not slowing down.
It's just moving into new media. Digital communication is creating new hybrid forms. Emoji function as a kind of paralinguistic layer. Text speak has its own conventions. Online communities develop in-group vocabulary that stabilizes into something rule-governed. We're seeing the same processes that created pidgins and creoles, just compressed into years instead of centuries.
The question is whether any of these online varieties will get native speakers and become full languages.
That's the open question. A creole needs a generation of children growing up in it. Online communities don't have that, mostly. But some urban varieties do. Urban Wolof in Dakar, Sheng in Nairobi, these are contact varieties with young native speakers. They're creolizing in real time.
Hilbert: The notebook.
What notebook?
Hilbert: Spiral bound. Black cover. I worked nights as a dispatcher for a long-haul trucking company in Newark in the late eighties. Half the drivers were Haitian. I didn't speak a word of Creole. So I wrote down phrases phonetically. Bon bagay, that was good. Machin nan pare, the truck is ready. Where's the pickup, what time's the drop. I kept it in my shirt pocket.
Did it work?
Hilbert: Mostly. The problem was, the drivers didn't all speak the same Creole. The ones from Port-au-Prince had one way of saying things. The ones from Les Cayes had another. The ones from up north, Cap-Haïtien, they'd tease the Port-au-Prince drivers about their French. Too many French words, they said. Not proper Creole.
Even inside the language, there was a hierarchy of authenticity.
Hilbert: That's what it sounded like. Same thing with the Italian families in my neighborhood. The grandmother from Naples would say the grandmother from Sicily didn't speak real Italian. And they were both speaking what the state would call a dialect.
The internal boundaries are just as political as the external ones.
Hilbert: One night I tried my notebook Creole on a driver from Port-au-Prince. He listened, laughed, and said, that's not Creole, that's French with extra steps. I still don't know if he was joking or insulting me.
He was probably doing both.
Hilbert: The notebook didn't help after that. I'd learned the phrases, but I hadn't learned which phrases belonged to which driver. The boundaries were inside the language, not around it.
That's the thing the tree model misses. A language isn't a single codebase. It's a set of overlapping varieties, and speakers track the differences constantly. They police them, they mock them, they use them to place each other.
Your driver didn't say you were speaking bad Creole. He said you were speaking a different variety, one that marked you as an outsider.
Hilbert: He said it in Creole. Or maybe French. I couldn't tell.
That's the point, though. You couldn't tell, and neither can the classifiers always. The line between French and Creole in Haiti is a continuum. Some varieties are more French-influenced, some less. The official language boundary doesn't capture that.
The notebook is a lovely artifact of the problem. You had a flattened version of a language that doesn't actually exist as a single flattened thing.
Hilbert: I still have it. In a box somewhere.
The driver's joke, French with extra steps, is actually a pretty good description of how some people think about creoles. But it's backwards. Haitian Creole isn't French with extra steps. It's its own system that happens to share vocabulary with French. The grammar is different. The steps are different steps.
The joke only works because the driver could hear the difference. He had the internal map of which features belonged to which variety.
Hilbert: He could hear exactly where I'd learned my phrases. Probably down to the town.
That's native-speaker knowledge. It's real, it's precise, and it's not written down anywhere. The linguist's job is to reconstruct it after the fact.
The answer to Daniel's question about who makes the call is, everyone and no one. Linguists describe, states decree, communities police, and the actual boundaries are negotiated in every conversation.
The number of languages is decreasing, but the number of varieties is not. Every speaker group is still producing variation. It's just that most of it never gets a name or a code or a flag.
The misconception most people bring to this topic is that creoles are broken versions of the parent language. That's the one to correct.
The correction is simple. A creole is a full language with its own grammar. Haitian Creole is not French with extra steps. It's a system that does everything French does, differently. The fact that it shares vocabulary doesn't make it a copy. It makes it a relative.
The open question I'm left with is what happens to the thousands of endangered languages that have no state backing. If recognition is political, and they have no political power, what's their path?
Documentation, mostly. Linguists and community members racing to record what's left before the last rememberers die. And then, maybe, revitalization. If the community decides it wants the language back, and has the resources to teach it to a new generation.
But that's a political project too.
Always is.
If you enjoyed this episode, leave a review. It helps other listeners find the show.
Thanks to our producer, Hilbert Flumingtop, for keeping the show running.
This has been My Weird Prompts, the human AI collaboration podcast. Email us at show at my weird prompts dot com.
We'll be back soon.
