Is "Universal Grammar" really as wrong as it sounds? 

Me too

Me too!@@DoughBrain

Me three!

I'd lovee to see this too!

Me too!

"What's the singular for people?"

I need to remember to use this comment in the intro!

Obviously, no serious language translation company today allows Chomsky-like ideas into their computers.

I don’t know any German but I would expect a machine to look at a long German word and occasionally divide it into sub words incorrectly with results that are hilarious nonsense to a human. In fairness, I have done this with Hebrew many times, such as wondering why Betzalel was named “the onion of G-d” instead of realizing it was “in the shadow of G-d”

@@wafelsen The old MT systems used dictionaries. If a word was not in these dictionaries, the systems could not translate it and definitely would not attempt to break it up into meaningful snippets. This would have slowed down the translation process. Our system ( and I assume most if not all of the systems of other companies) used an embedded complex morphology analyzer that recognized all possible inflected word forms on the source side and created the corresponding forms on the target side as final step. The dictionaries contained only the uninflected base forms such as singular forms, infinitives and uncontracted forms - plus codes (morphology) and lots of tags (grammatical, semantic and syntactic information) used by the engine and the rule system.

@@wafelsenflashbacks to me thinking that the "Gamali-" part of "Gamaliel" meant "my camel"...

I wonder why you didn't stick to languages within ONE language family. You went for two. I think because they are European?

I have LOADS of those

Computer science! Or at least computer pay the bills.

Yes, this is me. I like languages. I love computers. I love math. If I hadn't luckily discovered my incompatibility with academia as an undergrad, I might've ended up somewhere in the confluence of those studying generative linguistics and/or NLP right around the time that LLMs began blowing up... Eh. For the best.

yay for the computer science and linguistic theoretical crossovers

I was surprised, too. Language enthusiast, but also programmer and math nerd.

Also a cantab, had the same experience here (though I came to these conclusions during undergrad). I've ended up working on paradigmatic morphology which doesn't really fit such a compositional theory of language easily (contra Distributed Morphology and morphemic theory more generally), so I don't even really interact with Chomsky's work much these days, but my view is that Chomsky's work features 1 a confusion between description and explanation and 2 successive rounds of having less substantial things to say (e.g. minimalism has a much-reduced language facultt relative to P&P and G&B).

The problem with UG is that it's totally unfalsifiable. It's not a scientific theory. Also Chomsky is a hack, but that's not strictly related. There is no evidence in any field that the universal elements of human language are distinct from the definitional elements of language as a concept.

I find your criticism of null subjects to be very weak. UG never needed to have mandatory subjects; in fact the EPP was only proposed in the 1980s, as I assume you know. Even if every sentence in all languages we know of only had overt subjects, the EPP/obligatory subjects wouldn't HAVE to be part of UG. More recently, Chomsky (2013, 2015) even did away with the Extended Projection Principle. Besides, I don't see how misidentifying the subject of a language has any bearing on UG. It doesn't even mean the language doesn't have subjects (whether overt or not).

@@amazingcaio4803 I did not know that EPP was a later development. I never even heard the term but I guess it's the same as obligatory subjects. My point is that all the universals Chomsky proposed are refuted or irrefutable. And my broader point being that UG has a lot more baggage than the simple statement "Language is innate". That said, I wasn't referring to pro drop (like Spanish "te quiero") or dummy subjects (like English "it is raining"). Here I see that the subject is not expressed or semantically empty. But what about German impersonal passive ("Hier wird nicht spielt", word for word: "Here is not played"). And let me rephrase my other point: If two Western linguists analyze a very not Western language and come to different conclusions what the subject is, maybe "subject" is a Western concept that doesn't fit with foreign languages. The syntax theory I worked with tried very hard not to be Western centered and here is an article why it is so cool: www.computerworld.com/article/2929085/blame-chomsky-for-non-speaking-ai.html The article didn't age too well and the wikipedia article isn't too long but it has some links.

@@twipameyer1210 I do get some of your points. I also dislike that some claims people make are so wild they're essentially unfalsifiable. Not only would this be unscientific, but it would also go against one of the central goals of UG - to restrict language so that it explains language acquisition. There is indeed a lot more to UG than innateness. I don't even think innateness itself is controversial; language acquisition must ultimately be encoded in our genes. What's controversial is whether there is a domain-specific language faculty. While hypothesing its existence can easily lead to speculation, I do believe it is a worthy line of inquiry. I don't think that nowadays 'subject' has a formal definition in Chomskyan generative grammar. At most it's generally used as a synonym for the 'specifier of T(ense)', which is a theory-internal concept. This is hardly a Western concept (the obligatoriness of subjects, OTOH, could be) and people doing generative grammar actually like studying other languages so that they can gather enough data to refine the theory of UG. Hungarian was partially used to support the DP (determiner phrase) hypothesis and data from Chinese, Japanese, and Serbo-Croatian helped refine proper government and the Empty Category Principle. Chomsky also once proposed the configurational parameter based on data from Japanese. Even if the concepy of subjects (and/or the idea that subjects are obligatory) were Western-specific and didn't apply to other languages, this would mean that our analysis is biased, but I still wouldn't think we should abandon the idea of a universal grammar, especially because that's incidental to UG (there's no a priori reason for our account of UG to include subjects). Regarding subjects specifically, the general idea is that T must have a specifier, so the focus of inquiry isn't on the subject (read, the specifier of T), but on T itself (why does it need a specifier?). Most accounts don't really restrict the subject to a specific kind of element (any restriction is usually derivable from other principles or is lexically determined), so the subject could be a null pronoun, an expletive, or any other type of constituent (e.g. a locative, as in your example in German). Sure, looks can be deceiving and one could argue 'hier' isn't even the subject of your example sentence, based on empirical evidence or theory-internal reasons. That's why syntactitians may have different analyses. In fact, isn't that expected or even desired? In the end, it doesn't mean UG is refuted or irrefutable. Many universals still seem to be true. One example is Lasnik and Saito's (1984) restrictions on syntactic Wh-movement. (Their account may be outdated, but the empirical phenomenon still holds true, as far as I know.) Edit: BTW I personally have nothing against other theories of grammar and I don't deny their usefulness.

@@amazingcaio4803 Let me first thank you for this conversion! I enjoy it very much and think it's very fruitful, especially compared to other conversations on the internet. Let me now reiterate my claim: All models are wrong but some models are useful. UG is good at explaining some things, other theories explain or visualize other things better. Some Chomskyans will claim that UG is literally objectively what happens in our brains and strawmaning non-UG-linguists in saying they don't believe language is innate. That's what I felt the video did and that was what I was attacking. I don't think you think so, so we are good. And to lay my biases open: I'm more a computer scientist than a linguist. I did some basic linguistic lectures with Chomskyan professors and wrote my master thesis in the computer linguistics department with a professor who hated Chomsky so much that he unironically claimed that language is just a cognitive ability like any other. My (Chomskyan) syntax professor was so fond of the idea that the subject is universal that I had the impression that it's central to UG. I'm glad that isn't the case apparently. I find myself somewhere between these 2 worlds.

I’ve been wary of it, but I think I should

@@languagejones6784 Address Everett's claim that recursion is not the base on which language is constructed, including his claim that Piraha has no past tense

Would watch. I have had to explain UG too many times because of this controversy.@@languagejones6784

@@gordonbgrahamDon’t get me wrong, Everett might still be proved one day to be a hack or a bigot, but at this point no one can prove him wrong. And unlike Chomsky, he is making claims that can be refuted, meanwhile Chomsky just argues that UG is about biology, makes no predictions and is not a theory but a field of study, all which can’t be proven wrong.

@@impendio The lack of a concept of the past and number (other than a little or a lot) no words for colour and the apparent absence of recursion in Piraha (something that would counter Chomsky's claim vis a vis recursion) seems to suggest language is not fixed in its conceptual properties. That language is biological can't be refuted is certain, but that doesn't address the issue of recursive grammar which Chomsky is a proponent of.

One could also start to study Language/languages, not only as they developed during millions of years amonge primates, but also as how a mother tongue develops from the bond between mother and child, from the baby's perspective. The last months in the womb the unborn child hears both music and speach and other sounds, thus it is familiarized with the world of sound surrounding it. I have heard babies use sentence melodies, cadences, both declamatory and questions mode, when they could not talk and walk yet. They also recognize music and imitate melodies and entertain themselves . And I have heard babies take turns in interaction with their mother, using syllables of sounds and minimal pauses, as if they were words. Babies were in such situations very pleased with their sounds and the interaction they could bring about, ready to repeat the situation again and again. Patient mothers gently mold these sounds into (let's call it) a prelanguage, and then into the baby's mother tongue. Dads can do this, too, if they are around enough. Often parents understand the baby, and later the toddler, when nobody else does. Many cute situations come up, of entertaining value for the adult, too, but also misunderstandings - which are *meaningful* experiences, as part of learning how rules regulate language use. These foundations could contribute to explaining how come our languages are not so logical as binary sequences, instead there is a haphazardness to words and their forms within the sound system of the mother tongue (and the anatomy of toddlers).

But Chomskian grammar (assuming we're talking about generative grammar, which is a grammatical theory, and not the pseudoscientific innate syntax hypothesis which is frequently referred to as "universal grammar") has nothing to do either with how languages work in our brains nor with how they historically developed. Syntax is mainly a pathway to linearizing expressions that initially mirror our thinking that is primarily associative in nature: thoughts don't have hierarchy, they don't have constituents, they don't even have any linear order (even though they are experienced linearly due to the nature of time); they only have concepts and relations between them, relations between sets of related concepts - those can be active, patientive, causative, instrumentative, spatial, temporal, etc., - i.e. meaningful relations, not abstract-structural. Generative grammar is an abstract model that is supposed to be able to predict, mostly based on the structure of linguistic expressions, whether those linguistic expressions would be considered structurally possible within a given language - i.e. it is supposed to predict whether a person *perceiving* the text would consider it *structurally* inappropriate, and if they would, it is supposed to tell us the structural principle that was violated. Generative grammar does not and was never meant to predict nor explain the way syntactic structure is arrived at either in use or in historical development. Some structural patterns are obviously there, just like, for example, every walking animal moves its feet the same way as the other animals of the same species and the same age, even though they are free to move any other way they are physically capable of - they just won't be doing that most of the time, as long as the one of the millions of biomechanical alternatives is noticeably more comfortable in the long run; animals just have places they are aiming to get to, and all they do is step, step, step, etc., without having any "overall way of moving feet" in mind. With languages it also seems that this structure of speech isn't something that is ever the goal, but something that emerges on its own, out of the systematicity characteristic to the expressive means of a language - to the lexical units and their various counterparts within the lexicon, to the means for expressing predicative relations, semantic modification, communicative relations, rhetorical relations an so on - the systematicity that is quite necessary to ensure the possibility of unambiguously reconstructing the initial conceptual content that had led to this very structure that is observed (i.e. the systematicity, loss of which would lead to the impossibility to consistently "interpret" the resulting linguistic expressions). It's not the structural principles that lead the development of speech complexity - those principles aren't there yet, when speech never reaches any complexity: the language stays incapable of expressing that as long as the society that generates that language stays culturally incapable of the necessary verbal coherence - there are human languages that don't have any means of precisely expressing most kinds of relations between concepts, not even the predicate-subject-object relations. It is the cognitive development and cultural "fossilization" (roughly speaking, concept → myth → habit → tradition) that gradually lead to the development of this structure that we observe and that we can describe using the notion of generative grammar.

The linguist Guy Deutscher wrote a popular linguistics book all about how language development could have happened in prehistory: it is titled _The Unfolding of Language: The Evolution of Mankind`s Greatest Invention_

Per Goethe: “Wer fremde Sprachen nicht kennt, weiß nichts von seiner eigenen” - "Those who don't know a foreign language know nothing of their own."

When I studied Russian in school one of the best books I had was called English Grammar for Students of Russian. I think there are versions for several other languages as well.

​@@joshrotenberg5872--- YES!!! I recommend that series for everyone who wants to learn a language covered by that series. There's even 'Spanish Grammar for Students of English' written in Spanish.

As someone old enough to have learned to read through phonics and had English grammar explicitly taught in grade school and later got a master's in teaching ESL, I can tell you that 2/3 to 3/4 of English/ESL teachers actually have a very poor grasp of both English grammar and English spelling.

Your research sounds promising. But I'm sure I'm not alone wondering what kind of PhD program trained biologists AND linguists. Not my field. Just curious.

But Chomsky doesn't claim it evolved to be language specific. His claim is actually very similar to yours. There was some specific evolutionary change at some point that allowed other neural functions to click together and express this latent capacity as language. As more and more people spoke and languages evolved and changed, statistically, more and more and more different possibile arrangements of the computationally restricted possibilities of language emerged. Since language is a layer cake of different computacional systems of various constraints-pragmatics on top of semantics on top of syntax on top of morphology on top of phonology (itself constrained by the physical limits of phonetics)-you have an almost infite possibility of configutions. It's quite an elegant and simple solution to the problem of the vaaaast diversity of languages that exists.

@@ems4884you have to study them separately :-) although, i think it would be quite beneficial for linguists to combine their training with the biological one, especially in evolutionary biology, as there are plenty of shared fields of interest - e.g. population genetics. i started with biochemistry through molecular and system biology, and after some time, in fact, almost 20 yrs later, i went for general linguistics. my major interests in lingo are cognitive linguistics and sociolinguistics.

@@Eruntano42 from the biological point of view, the evolution of the language and the respective neural structures has to be more stepwise. e.g., arbib et al. hypothesise that there were at least seven distinguished steps (synergies of new neural networks) and, gradually, certain forms of language communication appeared, which were much simpler than contemporary human language but still helped to survive in more and more complicated human societies. the thesis of "pragmatics on top of semantics on top of syntax on top of morphology on top of phonology" is, according to this kind of study, not necessarily valid, e.g., pragmatics does not require syntax or morphology in the earliest assumed forms of language. it is still in the early development stage in all these directions, but what seems to be a sign of good science is that they can explain more than previous hypotheses.

The language acquisition literature is very much based on specific universal grammar mechanisms, and they are more productive than statistical learning.

Needless to say: yes to the intro in linguistics

You can’t escape!

'Gnome Chomsky' made me laugh, but familiarity does tend to make names sound more like plausible native words. Not a lot of native English vocab with an o-a transition :)

​@@rasmusn.e.m1064--- I don't think I've heard his name pronounced any other way than 'gnome'. Then again we're talking grad school linguistics in the early '90's.

@@ak5659 I'm not a native English speaker, but I typically pronounce his first name in English something like [ˈnɔ͡ʊ̆ˌəm], which I perceive as my realisation of /ˈnɔ͡ʊ̆ˌæm/ but I can see why the final schwa might easily get reduced in this particular constellation of //First name_Surname//, where the surname tends to get a lot of stress.

Zing!

Conversation with a child language acquisition lecturer, paraphrased: "young children are bombarded with language 24/7 and are sensitively attuned to it because their attention is always focused on their parents" "so why is UG premised on the paucity of the stimulus?" "my guess is that Chomsky never spent much time raising children". 😵🔥

@@johnnye87 Of course there is poverty of input. Universal Grammar (UG) approach claims that there is a universal set of principles and parameters that control the shape of human language. But the entire corpus of studied language is so small that it is impossible to derive what this universal set of principles and parameters actually is from using the entirety of human language as input. If the entirety of human language does not contain enough input to enable the principles and parameters of UG to be extracted, then how could a child receive enough input to work them out, unless they are already built in?

@@stevencarr4002 My issue isn't so much with the hypothesis that the evolved human language faculty preferences certain patterns over others, so that you can get a "complete" language without having to expose the child to enough data to unambiguously derive every rule. That seems pretty obvious and unremarkable. My issue is that they tend to treat the formulae they come up with as though they represent actual mental processes, when they bear no resemblance at all to how any other part of the human mind works. We don't operate on strict mathematical rules. A neural network like the human brain runs on weights and fuzzy logic, and the languagey bits of our brain aren't isolated from the non-languagey bits. Chomskyans produce formalised descriptions of a phenomenon and then mistake the map they drew for the territory.

@@stevencarr4002The argument question beggingly assumes there are principles of UG to be worked out. The usual anti-UG line is that language evolved to be learnable by human brains not the other way around. Cognitive biases were around long before before language, so any language that arose would of course be shaped by those biases. We don’t work out the laws of physics to learn to walk either, so is the correct formalization of physics innate also?

Or reduce Chomsky to the trivial conclusion that human beings can speak languages because they have brains. That is quite uncontroversial.

LLMs dont have the same language acquisition or generation as humans even at all. When was the last time you practiced backpropagation?

@@Finnnicus No wonder LLMs don't work and can't process human language. They don't use UG, which is a fundamental feature of human language.

Is it in the captions? I didn’t recall making an overlay…

@@languagejones6784 D'Oh! That must be it. Yet another example of "AI" proving to be 50% ill-yclept🙂

He seemed genuinely kind hearted and interested in talking to everybody

@@languagejones6784 i had to say it because i also remember the arguelles and laoshu times and he was such a chill dude. Nostalgia is a beautiful word

Chomsky on Cambodia

For example, the reason we see common patterns in how human languages deal with “nouns” is because of commonalities in how our brains model ideas about physical objects. Our shared rules about how we expect physical objects to behave cause common patterns when we describe those objects.

I am not sure, but I think it is because Esperanto is only spoken as a first language by people who grew up bilingual, speaking Esperanto and , for example, Russian. Modern Hebrew is an example of a language that went from no native speakers to several million (according to Wikipedia). I wonder if the same might happen to Esperanto.

​@@stevencarr4002It already has to some degree, there are thousands of native Esperanto speakers with different degrees of fluency and actual use of the language. Modern colloquial Esperanto is quite different to Zamenhof's Esperanto or modern formal Esperanto, with things like dropping the accusative marker -n being common.

That’s coming! It’s tricky because the strong form is used for prejudice, the weak form was thrown out with it, before research on unconscious bias, and that’s also bad. But also some of the research justifying the weak form of it is actually a bit shaky too. I’m definitely making a video on it but it’s gonna take some more thinking and writing!

LAD

Referring to the LAD, do you mean that the older a kid is when first exposed to language the less likely he is to acquire native speaker fluency? If yes, I taught people in that situation years before Language Deprivation Syndrome became a thing. Plehse excuse the awkward wording. Grad school was a looooong time ago.