The 50th Anniversary of Noam Chomsky's Aspects of the Theory of Syntax(chronicle.com)
chronicle.com
The 50th Anniversary of Noam Chomsky's Aspects of the Theory of Syntax
http://chronicle.com/blogs/linguafranca/2015/06/22/revolutionary-methodological-preliminaries/
26 comments
But meaning is compositional! At least in large part.
Most sentences we encounter are novel, and yet they're understood. So, somehow, we break up the meaning of the novel sentence, relating it to sentences we've seen.
And actually the process there isn't that mysterious. The basic facts of English syntax are pretty obvious. Any English speaker will agree that "The blahdies merikaled the beemers" implies "The beemers were merikaled by the blahdies", without knowing what these words mean.
I see from your Dad's comment below that he's fond of the systemicist school, following Halliday etc. I've done substantial work with this theory, as it's what I learned in my undergrad. The rest of my career has been spent working with others.[1]
If I had to identify a central objection to SFL, I might say this: SFL is very fond of proposing non-linear effects. The theory goes that it's all about interactions, and if you study anything in isolation, you won't find anything out --- because everything's non-linear.
This assertion is basically unsupported. It just sort of sits there as an assumption. But we should assume simplicity, not complexity. The simplest assumption is that there is no relationship between X and Y, but the next simplest is that there's a linear relationship. If you want to tell me there's a non linear relationship, you'd better come with evidence. And SFL never does.
[1] https://scholar.google.com/citations?user=FXwlnmAAAAAJ&hl=en
Most sentences we encounter are novel, and yet they're understood. So, somehow, we break up the meaning of the novel sentence, relating it to sentences we've seen.
And actually the process there isn't that mysterious. The basic facts of English syntax are pretty obvious. Any English speaker will agree that "The blahdies merikaled the beemers" implies "The beemers were merikaled by the blahdies", without knowing what these words mean.
I see from your Dad's comment below that he's fond of the systemicist school, following Halliday etc. I've done substantial work with this theory, as it's what I learned in my undergrad. The rest of my career has been spent working with others.[1]
If I had to identify a central objection to SFL, I might say this: SFL is very fond of proposing non-linear effects. The theory goes that it's all about interactions, and if you study anything in isolation, you won't find anything out --- because everything's non-linear.
This assertion is basically unsupported. It just sort of sits there as an assumption. But we should assume simplicity, not complexity. The simplest assumption is that there is no relationship between X and Y, but the next simplest is that there's a linear relationship. If you want to tell me there's a non linear relationship, you'd better come with evidence. And SFL never does.
[1] https://scholar.google.com/citations?user=FXwlnmAAAAAJ&hl=en
"Two comments: First, I'm not fond of the systemicists. I think Halliday took a wrong turning.
Second, you say that "somehow, we break up the meaning of the novel sentence, relating it to sentences we've seen." Quite correct. But I think the essential point of the Firthian approach is that there is not necessarily a single decomposition of the string. There can be multiple decompositions, each yielding components that can be related to other utterances. All of the standard approaches start by assuming single decompositions (or decomposition trees) on each level. That's where I think the whole structurallist (and systemicist) project went off the tracks. I did manage to crank out a few papers in this direction many years ago, but left it before I had gotten my act together, and I succumbed to the lure of Milton's "that bad eminence" (administration)."
-My dad
Second, you say that "somehow, we break up the meaning of the novel sentence, relating it to sentences we've seen." Quite correct. But I think the essential point of the Firthian approach is that there is not necessarily a single decomposition of the string. There can be multiple decompositions, each yielding components that can be related to other utterances. All of the standard approaches start by assuming single decompositions (or decomposition trees) on each level. That's where I think the whole structurallist (and systemicist) project went off the tracks. I did manage to crank out a few papers in this direction many years ago, but left it before I had gotten my act together, and I succumbed to the lure of Milton's "that bad eminence" (administration)."
-My dad
Thing is, Chomsky is (at least partly) responsible for the so-called Cognitive Revolution. Look up B.F Skinner and notice what happened to his career after Chomsky reviewed one of his papers. Skinner's brand of behaviorism was surely something we didn't need around anymore. He also changed the way everyone thinks about language. Before Chomsky, there really was no real theory of syntax that was even close to being scientific and formal.
Although probabilistic machine learning methods replaced the Chomskyan brand of formal theoretic methods in NLP, a lot is owed to is work. Obviously there is the Chomsky Hierarchy but even things like the Penn Treebank POS tags are taken straight out of work in the Generative tradition from circa 1970 (not sure why no one has tried the updated and more accurate tags currently being used in theoretical linguistics). Anyways I could go on and on but Im wondering what your Dad has to say about this ?
P.S There is an ongoing joke in the field of Linguistics about people who either think they understand Chomsky or just straight up don't understand him. Its difficult stuff.
Although probabilistic machine learning methods replaced the Chomskyan brand of formal theoretic methods in NLP, a lot is owed to is work. Obviously there is the Chomsky Hierarchy but even things like the Penn Treebank POS tags are taken straight out of work in the Generative tradition from circa 1970 (not sure why no one has tried the updated and more accurate tags currently being used in theoretical linguistics). Anyways I could go on and on but Im wondering what your Dad has to say about this ?
P.S There is an ongoing joke in the field of Linguistics about people who either think they understand Chomsky or just straight up don't understand him. Its difficult stuff.
"No question that a lot of the tools developed under generative grammar have proved useful, and one can certainly cite the Huxleyan principle that truth comes sooner from error than from a vacuum. I think I would say that the American structuralist tradition was headed in a wrong direction from at least 1933 (Bloomfield's Language) and that Chomsky (despite the claims of his followers that he made a revolutionary break from structuralism) simply gave structuralism a mathematical rigor that allowed it to be taken to its logical conclusion. The tools that were developed in the process have allowed the field (or the more brain-oriented parts of it) to move in a rather decisively different direction.
Full disclosure: I got my PhD at Yale under Syd Lamb, so naturally I would not be classed as a Chomskyan. It took me some years to realize that Lamb's StratGram itself suffered from the same reductionist flaws. The only theoretical school I know of that had a chance of moving in the right direction (until very recently) was Firth's. Unfortunately, neither he nor his followers developed the theoretical tools that were needed. Certainly Halliday's Systemics became as reductionist as any of the others. Hudson's word grammar made some good moves but never really developed well. So yeah, Chomsky set something in motion that produced some useful theoretical tools that could be used to help push off in new directions. But the TGG tradition, being a well developed extension of the Bloomfieldian tradition, is still basically a dead end."
-My dad
Full disclosure: I got my PhD at Yale under Syd Lamb, so naturally I would not be classed as a Chomskyan. It took me some years to realize that Lamb's StratGram itself suffered from the same reductionist flaws. The only theoretical school I know of that had a chance of moving in the right direction (until very recently) was Firth's. Unfortunately, neither he nor his followers developed the theoretical tools that were needed. Certainly Halliday's Systemics became as reductionist as any of the others. Hudson's word grammar made some good moves but never really developed well. So yeah, Chomsky set something in motion that produced some useful theoretical tools that could be used to help push off in new directions. But the TGG tradition, being a well developed extension of the Bloomfieldian tradition, is still basically a dead end."
-My dad
I mean, if you happen to have some data that Stratificational Grammar can explain better than Generative Grammar I would love to see it. There are a bunch of theories directly influenced by Chomsky which offer alternatives to generative grammar. I've yet to see someone show that one is superior. In many ways, they are all extensionally equivalent.
"I never said that StratGram can explain things better than TGG. I said it had the same reductionist flaws. It and all the theories influenced by Chomsky are really outgrowths of Bloomfieldian structuralism and share its underlying orientation. As you say, in many ways they are all extensionally equivalent. The field needs to move beyond all of them.
Don't know what assumptions you are referring to, but a forum like this really isn't the place for detailed arguments (even though I used to require my students to provide data, analysis, and argument in two-page, double spaced papers). Just check out J. R. Firth, and read him not for the outdated details but to see what his underlying assumptions are."
-My dad
-My dad
Your Dad really likes assumptions and not providing arguments but Ill look into some of the names Ive never heard of. Thanks !
Perhaps. It's kind of a trope these days to deride pre-Chomsky linguists as a bunch of wrong-headed behaviorists. But Skinner was advocating a theory of psychology. Most linguists prior to Chomsky were not really committed to behaviorist psychology, just happy to borrow it for use against the prevailing scientific racism.
Chomsky is a theoretical linguist, but not really a scientific one. Generative linguistics does not have a good tradition of basing conclusions about language on empirical study of language use, with the implication that there's hardly any point in studying the structure of diverse languages (since they're all the same deep structure). Take the speech establishing the Principles & Parameters theory, which likened the linguistic feature of pro-drop to a series of switches the brain could flip on or off to establish a language as pro-drop...when subsequent research immediately showed pro-drop to be more of a continuum than a binary feature. There is all the wasted effort, like generative grammars of endangered languages that relied on morphology rather than syntax to communicate the same information.
Then there are the bad assumptions, like the "poverty of the stimulus" argument, that had no real empirical basis when he made it. This supported ideas like linguistic nativism that got exported out of linguistics since Chomskyan linguistics was a dominant theory at the time. Whereas in truth linguistics has always been a multi-paradigmatic field.
Chomsky is a theoretical linguist, but not really a scientific one. Generative linguistics does not have a good tradition of basing conclusions about language on empirical study of language use, with the implication that there's hardly any point in studying the structure of diverse languages (since they're all the same deep structure). Take the speech establishing the Principles & Parameters theory, which likened the linguistic feature of pro-drop to a series of switches the brain could flip on or off to establish a language as pro-drop...when subsequent research immediately showed pro-drop to be more of a continuum than a binary feature. There is all the wasted effort, like generative grammars of endangered languages that relied on morphology rather than syntax to communicate the same information.
Then there are the bad assumptions, like the "poverty of the stimulus" argument, that had no real empirical basis when he made it. This supported ideas like linguistic nativism that got exported out of linguistics since Chomskyan linguistics was a dominant theory at the time. Whereas in truth linguistics has always been a multi-paradigmatic field.
Theoretical Linguistics in the Generative tradition is entirely based on empirical data. If a theoretical linguist proposed a theory without any empirical data to support it, they would (and do) get laughed at.
The poverty of the stimulus argument (which can be traced back to Plato) has plenty of evidence supporting it. Yes there is negative evidence provided to children in some cultures. In North American elementary schools and often at home, children are told when they are saying something that is considered 'wrong'. There are many cultures, families and situations where this does not happen at all and children will learn their native language just by being exposed to people in their community using language. In that case the child is only exposed to positive evidence. What is crucial here is the fact that children CAN learn a language with only positive evidence. They just listen.
In addition, the similarities between the languages of the world far outweigh the differences. There is no way to account for this fact without positing some innate knowledge of language.
Also, read up on the distinction between competence and performance. I noticed you mentioned 'language use'. Yes, it is a mistake to base a theory on language use. We are talking about an ideal speaker-hearer situation. Language in a vacuum. Sound familiar?
Also, read up on the distinction between competence and performance. I noticed you mentioned 'language use'. Yes, it is a mistake to base a theory on language use. We are talking about an ideal speaker-hearer situation. Language in a vacuum. Sound familiar?
Plato made plenty of bad arguments, too, and there are a lot of arguments that go back to his time. Anyone who is learning a language receives a constantly stream of feedback about how their communication is perceived, whether or not this feedback is explicit, modeled correct behavior, or delivered simply through incomprehension. Children are constantly deducing good and bad hypothetical "rules" all the time, and putting them to the grim and unforgiving test of trying to get people to understand them. Kids will happily overapply a general rule (like forming a past tense with "-ed" with a strong verb, "I readed it") even in the face of constant explicit negative input, but if they try a bad rule of syntax they'll get a string of gobbledygook that people don't understand and may even ask for clarification.
There are many, many fewer linguistic universals than some would have you believe, and many of these are best explained as similar tradeoffs between different needs, rather than a universal underlying pattern mysteriously encoded in the human mind.
The centrality of the distinction between linguistic competence and linguistic performance is exactly one of the bad premises of Chomskyan linguistics. The ideal speaker-hearer is as worthy a topic of scientific study as a unicorn: nothing more than a thought experiment. But then, Chomsky was not trying to describe the psychological reality of language use and language production, but rather articulate a formal system that could represent the structure of carefully curated and constructed phrases.
Humans are embodied, and they have things like working memories and perceptual metaphors. And human languages are not formal schemes of rules, but rather a complex and messy welter of social conventions.
There are many, many fewer linguistic universals than some would have you believe, and many of these are best explained as similar tradeoffs between different needs, rather than a universal underlying pattern mysteriously encoded in the human mind.
The centrality of the distinction between linguistic competence and linguistic performance is exactly one of the bad premises of Chomskyan linguistics. The ideal speaker-hearer is as worthy a topic of scientific study as a unicorn: nothing more than a thought experiment. But then, Chomsky was not trying to describe the psychological reality of language use and language production, but rather articulate a formal system that could represent the structure of carefully curated and constructed phrases.
Humans are embodied, and they have things like working memories and perceptual metaphors. And human languages are not formal schemes of rules, but rather a complex and messy welter of social conventions.
You were almost coherent until you get to your assumptions about 'what people would have me believe' about linguistic universals. There are many universals that can be explained by generative grammar. There are numerous examples. 'tradeoffs between different needs' is not an actual argument. It is ridiculous that you are replacing 50 years of work with something so unscientific and vague.
I think you know damn well that there are patterns in the mind. They are posited for many things. Generative Linguists want to determine which of those patterns can account for certain linguistic phenomenon.
Physicists and Mathematicians study idealistic events, matter and object in idealistic spaces and conditions all the time. They don't study unicorns.
You cannot account for the fact that humans can instantly parse novel sentences using 'complex and messy social conventions'.
I cant tell you have no idea what Generative Linguists do and have never really read any of the literature. You (like a lot of people here) are just repeating the predictable things people say when Chomsky comes up that they probably overheard once.
I think you know damn well that there are patterns in the mind. They are posited for many things. Generative Linguists want to determine which of those patterns can account for certain linguistic phenomenon.
Physicists and Mathematicians study idealistic events, matter and object in idealistic spaces and conditions all the time. They don't study unicorns.
You cannot account for the fact that humans can instantly parse novel sentences using 'complex and messy social conventions'.
I cant tell you have no idea what Generative Linguists do and have never really read any of the literature. You (like a lot of people here) are just repeating the predictable things people say when Chomsky comes up that they probably overheard once.
It's certainly true that I've read more of Chomsky himself than the immense literature he spawned; and most of my reading of the generativist literature is at its messy edges where those explanations tend to be pretty unconvincing. Focus your reading on dubious generative accounts of morphologically-rich languages and it's easy to become disillusioned with the paradigm. "Principles and parameters" is where he lost me, but Chomsky himself has retreated from so many positions over the decades, when theory meets evidence, that it's hard to tell in a coherent way what his theoretical framework actually is anymore.
I'm not really "replacing 50 years of work" so much as speaking from a different paradigm. Linguistics is a discipline of multiple competing paradigms, and suffers the resulting problem of people talking past each other because they do not share the same premises or assumptions about the field of study. It's not my fault that Chomsky led a generation of linguists down so many blind alleys for decades. But I don't know how many people would defend something like "The Sound Pattern of English" over say, optimality theory.
I don't even agree that there are many linguistic "universals". Most so-called "universals" are rather implicational tendencies, which do not require appeals to native grammatical knowledge to explain. You don't need a "head branching parameter" to explain many head directionality tendencies, just the general human tendency to pay more attention to elements at the edges of statements and the process of diachronic systematization. True linguistic universals are very few, and frequently seem to me to be more general cognitive universals in linguistic disguise. https://www.princeton.edu/~adele/LIN_106:_UCB_files/Evans-Le...
Generative grammar itself does not attempt to explain the psychological reality of how living humans parse and process novel sentences, but rather to describe an abstract system of knowledge of a WEIRD and mono-lectual standard language known by ideal speaker-hearers (who lack such prosaic things as "memory limitations"). Idealizing language is one thing, but this idealization seems quite different from language as the semiotic practices, social conventions, and shared cognitive tools used for communication by multi-lectual human animals as a core element of their social behavior.
I'm not really "replacing 50 years of work" so much as speaking from a different paradigm. Linguistics is a discipline of multiple competing paradigms, and suffers the resulting problem of people talking past each other because they do not share the same premises or assumptions about the field of study. It's not my fault that Chomsky led a generation of linguists down so many blind alleys for decades. But I don't know how many people would defend something like "The Sound Pattern of English" over say, optimality theory.
I don't even agree that there are many linguistic "universals". Most so-called "universals" are rather implicational tendencies, which do not require appeals to native grammatical knowledge to explain. You don't need a "head branching parameter" to explain many head directionality tendencies, just the general human tendency to pay more attention to elements at the edges of statements and the process of diachronic systematization. True linguistic universals are very few, and frequently seem to me to be more general cognitive universals in linguistic disguise. https://www.princeton.edu/~adele/LIN_106:_UCB_files/Evans-Le...
Generative grammar itself does not attempt to explain the psychological reality of how living humans parse and process novel sentences, but rather to describe an abstract system of knowledge of a WEIRD and mono-lectual standard language known by ideal speaker-hearers (who lack such prosaic things as "memory limitations"). Idealizing language is one thing, but this idealization seems quite different from language as the semiotic practices, social conventions, and shared cognitive tools used for communication by multi-lectual human animals as a core element of their social behavior.
Very accurate. Honestly, reading Chomsky in a lot of places he says things like there's no way a baby could learn this, it must be hard coded in their brain! Then in baby studies done later we actually find babies do have a natural understanding of things like probability and can do the things he says are impossible.
There are many other flaws in his work. I tend to think of him as just an old man who wrote his stuff before we knew as much as we do about genetics, so was unfortunately really off base. It's nice that people who don't know the science were excited by him, I guess.
There are many other flaws in his work. I tend to think of him as just an old man who wrote his stuff before we knew as much as we do about genetics, so was unfortunately really off base. It's nice that people who don't know the science were excited by him, I guess.
Could you link some of those 'baby studies' you are referring to ?
I suspect this relates to the debate over the 'poverty of the stimulus' argument. It's a pretty fundamental premise of Chomskyan innatist arguments that seems to me totally wrong.
https://en.wikipedia.org/wiki/Poverty_of_the_stimulus
> was the foundation for the ultimate mathematical elaboration of a structuralist theory that bears no relationship to anything in nature
So, like most theories?
Or do you really think of electrons as 'tiny things'?
You do the math and you get the results.
But yeah, Chomsky structures seems to be more useful when applying it to computer languages than to real ones.
So, like most theories?
Or do you really think of electrons as 'tiny things'?
You do the math and you get the results.
But yeah, Chomsky structures seems to be more useful when applying it to computer languages than to real ones.
I think I've missed the point of your response (electrons?).
Just to clarify, Chomsky's theories setup a framework for artificial reduction of language which did nothing to help elucidate the true nature of language (which is obviously a product of the evolution of the brain).
The problem seems to be that Chomsky's influence sent a lot of grad students down a path that didn't really advance linguistics in terms of the human element. I don't even know if his theories really helped with computer analysis of language.
Chomsky and people who study in the generative tradition have a lot to say about language as a biological organ (which obviously evolved from the language organs Chimps possess). In fact, Chimps have very similar systems to us. They can perceive voiced onset time (VOT) just as we do and they can reason about semantic issues. They can also utter and understand sound-meaning pairs however, they have nothing like our syntax for making novel and complex utterances out of the simple ones.
What is the thing with languages anyway. Previous generations were so fetishist about it. I found I've learned deeper things by toying with musical instruments and geometry, although encoding geometry (among other things) as visual symbols was a great foot in the abstraction door. But these intellectual aesthetics are far from human communication IMHO.
ps: sorry for that digression btw
ps: sorry for that digression btw
> It is rather surprising that more has not been done this year (thus far, anyway) to commemorate a significant semicentenary: the 50th anniversary of what could reasonably be called the most influential linguistics book of the 20th century.
Linguistics as a field is in a post-Chomsky hangover and none-too-happy to celebrate the man that put them there. Chomsky's many years as the dominant voice in the field were traumatic for linguistics. Once a decade or so he would promulgate The New Way of doing syntax, handed down from on high from this very influential man, which meant incredible disruption not only for syntacticians but also for those working in semantics and morphology/phonology.
Much of Chomsky's influence came not from his intellect (though that was surely prodigious) but from his forceful personality. His will to publicly ridicule not only opposing ideas but opposing personalities was a big part of his rise and continuing dominance. Linguistics was a nasty place to work for a long time. (See "The Linguistics Wars").
Now that he's not actively working to maintain it, Chomsky's academic reputation has suffered substantially and the field is largely trying to move on. Large-scale corpora and computational capabilities enable more empiricist rather than rationalist approaches to human language, which has made the argumentation less personal and more productive.
At least, that's my take.
Linguistics as a field is in a post-Chomsky hangover and none-too-happy to celebrate the man that put them there. Chomsky's many years as the dominant voice in the field were traumatic for linguistics. Once a decade or so he would promulgate The New Way of doing syntax, handed down from on high from this very influential man, which meant incredible disruption not only for syntacticians but also for those working in semantics and morphology/phonology.
Much of Chomsky's influence came not from his intellect (though that was surely prodigious) but from his forceful personality. His will to publicly ridicule not only opposing ideas but opposing personalities was a big part of his rise and continuing dominance. Linguistics was a nasty place to work for a long time. (See "The Linguistics Wars").
Now that he's not actively working to maintain it, Chomsky's academic reputation has suffered substantially and the field is largely trying to move on. Large-scale corpora and computational capabilities enable more empiricist rather than rationalist approaches to human language, which has made the argumentation less personal and more productive.
At least, that's my take.
This was actually not Chomsky's first book on this topic. That was Syntactic Structures (1957).
Neither the title here nor the article itself claim it to be his first book, only an incredibly influential one.
The article even agrees with you:
> §3 sets out a radically revised structure for transformational grammars, quietly abandoning the notion of “kernel sentence” from Chomsky’s first book, Syntactic Structures, and introducing the term “deep structure,” which in a sense replaces it.
The article even agrees with you:
> §3 sets out a radically revised structure for transformational grammars, quietly abandoning the notion of “kernel sentence” from Chomsky’s first book, Syntactic Structures, and introducing the term “deep structure,” which in a sense replaces it.
Correct. That book, over time, changed a lot of things.
-My dad (Ph.D. Linguistics, Yale - way back in the day)