Does Language Exist?

 

Noam Chomsky lectures on the existence of language

The University of Cologne has posted a couple of videos of lectures by Noam Chomsky, one of them on language. The language lecture is titled What Is Special About Language? The lecture runs a bit over an hour and a follow up Q and A adds another half hour, so it is not a quick distraction, but the video is well worth studying. Chomsky is a clear speaker and gets to the heart of his matter. I always come away from his work with a stronger sense of what makes my own thinking different. In this case I'm particularly struck by our contrasting views over what evolved..

Chomsky is clear, by the way, that he thinks language did emerge through a process of evolution, although he thinks piecing together events is likely too hard a question to be answered. More critical, however, is the discussion of just what emerged through that process.

Chomsky takes as his target a paper in Science magazine by N.J. Enfield ("Without Social Context?") that reviewed Tecumseh Fitch's book, Evolution of Language, and a collection of papers on the evolution of "human language."

Chomsky makes a pretzel of an argument because he accepts Enfield's conclusion that the studies of language origins have been a failure while disagreeing with Enfield's arguments supporting the conclusion. (I often find myself in exactly the same knot, agreeing with Chomsky's criticism of others while disagreeing with his reasoning.)

Language exists, Chomsky says, as a module in the brain that provides a universal grammar (UG) common to all cultures and peoples throughout the world. The UG is a computational system for generating sentences. It was likely selected because it enriched the cognitive-intentional system (thought and planning). This enriched system is the source of language's meaning.

The dominant objection to this definition is that it ignores languages social role. (I'm one of the people who makes that objection.) As evidence of the existence of language as he defines it, Chomsky considers the sentence He wondered whether the mechanics fixed the cars? and looks at ways it can be used to inspire other sentences. You can ask How many cars did he wonder whether the mechanics fixed? [As often is the case, I'm not persuaded this sentence counts as acceptable English, but I skip that.] You cannot ask How many mechanics did he wonder whether fixed the cars?

The point of this study is that the first sentence has subparts [He wondered whether [the mechanics fixed the cars]]. Mechanics is a plural so you can ask how many of them and cars is a plural so you can ask how many of them. How would you compute a follow-up question?

1. Insert the interrogative to the front and move the subject of the question to follow the interrogative: how many ____ (cars or mechanics).
2. Transform the first subpart of the original sentence, He wondered whether: did he wonder whether.
3. Merge the last subpart of the sentence: the mechanics fixed the cars.

This process generates two possible sentences:

• How many cars did he wonder whether the mechanics fixed the cars?
• How many mechanics did he wonder whether the mechanics fixed the cars?

Notice that neither sentence is what Chomsky gave us. That's because the first 3 rules generate the sentence's meaningful structure. There is a 4th rule that generates the externalized version that is actually given to the world: Simplify the sentence by removing moved subjects to end redundancy:

• How many cars did he wonder whether the mechanics fixed the cars?
• How many mechanics did he wonder whether the mechanics fixed the cars?

The second generated sentence is wrong, owing to a special rule and so, Chomsky tells us, we cannot ask the second question except through complex paraphrase, even though it is a perfectly valid question. [about 12  minutes into video]

I hope I got that derivation right. Chomsky does not take his audience through the generational process, but I want readers to grasp what Chomsky is up to. He is showing that he can generate sentences from other sentences through a fully automatic process that has no need to attend to social contex or even intelligence. This computational system exists and this system is the universal grammar, language itself..

But to investigate the workings of these three sentences Enfield says we have to consider their social context, and social context has nothing to say about these sentences. Ergo, there is nothing to investigate in these sentences. The social context formulation leaves us with nothing to offer as way of explanation about why one sentence is acceptable and one is not, and why the kind of mistake found in the unacceptable one is essentially unknown among native speakers.

There are many other important moments in the lecture, especially Chomsky's rejection of meaning as a reference to things. It comes an hour into the talk and is brief, but worth hearing. However, I'm stopping here to address the pretzel Chomsky has presented me.

First, I agree that we have to explain the puzzling sentences and I congratulate Chomsky and the movement he founded for its cleverness at identifying sentence after sentence that cannot be explained by simply referring to stimuli and conditioned reflexes. I note, however, that Chomsky's lecture makes no mention of the work by Morten Christiansen and his colleagues who have been looking into ways people can learn hierarchically structured sequences.

Second, Chomsky glosses much too hastily over social analyses of his sentence about mechanics and cars. He says, "Any inquiry [into social context] is very quickly finished; there is effectively none." [13 minutes into lecture] But is that true?

On this blog, sentences are assumed to work by directing a listener's attention, so any sentence is fundamentally social. The mechanics/cars sentence is an advanced one that is concrete as far as the nouns [he, mechanics, cars] go, but the verbs [wondered, fixed] are a bit more tricky and the pronoun/preposition whether is even trickier. This is not a sentence I would use early in a text book about piloting attention. So I'm just going to ask readers to imagine that perhaps an attentional (social) analysis is possible, at least possible enough to forbid Chomsky's quick brush aside.

Next, I wonder what data Chomsky has to offer to justify the process of producing those two questions about mechanics and cars. Chomsky would dismiss the question. The lecture includes a long discussion of scientific method—hypothesis, experiment, conclusion—and an argument that amassing a corpus of data is just useless and never done in real science. The last claim is an absurd exaggeration. Darwin's voyage around the world was not driven by hypothesis or experiment. He gathered masses of samples and examined everything he could find. His idea of natural selection came later, and then he went over his data to see if in fact it actually supported the idea. The same was true of Jane Goodall who had no training in field work, but observed wild chimpanzees and changed our understanding of what apes can do.

Language studies are much too underdeveloped to pretend the field is as theory-dependent as physics.

In particular, I'd like to know how we know that the secondary, how-many sentences are acceptable or unacceptable English. It cannot be just because they fit the rules of generation, as that would assume the correctness of what we want to test.

We could look in a corpus of data to see if similar sentences appear. In fact I Googled the phrase "did he wonder whether" and found only a few (69) examples (including at least one reference to the Chomsky usage). Many of the examples start off with the phrase, turning did into an  interrogative,  as in "Did he [Abraham] wonder whether he correctly understood the command of sacrifice?"   Did can forgo the interrogative role if the sentence is not a question, as in, "Only then did he wonder whether bats might be fertilizing the plants."Looking at this data has helped me understand just what it so confusing about the sentence Chomsky accepts. We are primed for a question and don't readily know how to deal with the sometimes interrogative did. That is a social confusion, not a syntactical one.

The Google corpus includes examples of usages in linguistic studies, e.g., "Some speakers seem to accept such forms as What did he wonder whether John saw?" ,

So we see two methods of determining usage acceptability. One is to check the corpus for information about  how phrases and words are used. Chomsky doesn't like that one. A second can be to ask people if they accept a sentence and have some arbitrary level of acceptance that determines the quality of a sentence.

A third method might be to ask people to provide their own sentences and see what they come up with. Show them He wondered whether the mechanics fixed the cars, and get them to generate interrogatives. Here are two sentences I might generate:

• How many cars did he wonder about?
• How many mechanics did he wonder about?

What do you  know? I've come up with simpler sentences and I've managed to ask what Chomsky says I can only ask about via complex parahrase. So why do I have to take Chomsky's monstrosities and resulting argument seriously?

Let's be clear. I agree with Chomsky that language evolved, syntax must be taken seriously, and that language exists. Language can take a variety of forms—spoken, signed, written, maybe even drums and/or smoke—and there must be something common to them all. Chomsky would agree with that last point, citing all these forms as "externalizations" of the language module's output. There I am not so certain; it could also be that the phenomenon began using just one form which was then internalized,as verbal thought  permitting other forms to emerge.

(Note: That last sentence was about the evolution of language, not what happens today. Modern human brains can handle signing right from the start.)

Comments

[JanetK]

I think the missing ingredient is communication. Some want to treat language as something other than a form of communication. They seem to want language to be a form of abstract algebra without a purpose/function such as communication. Language is basically social because it is basically communication and communication is basically social.Can anyone explain why the idea of communication seems to be shunned?
------------------------------------------
BLOGGER: Chomsky is very clear in his lecture. The function of language is NOT communication, it is thought of an improved variety.

[Karthik Durvasula]

I have given you some flak before for misunderstanding/miquoting Chomsky's words. I have to say this review was very fair and balanced. And I really enjoyed reading it; and seeing exactly where you disagree.

I have read Mortensen et al's work (a couple papers, not all of it :)), and from what I can see they make a few predictions which to me are most probably incorrect. If you ever do a review of their work (or perhaps you already have, and I missed it), I will definitely comment and let you know which parts I disagree with. :)

@Janet: About "They seem to want language to be a form of abstract algebra without a purpose/function such as communication" - I think the simple point one needs to see is at some level scientists are looking for systematicity. If X doesn't have a system...there is nothing to be studied. And something that is systematic can always be reduced to "abstract algebra". If you cannot reduce it to that, then you cannot study it. The curious truth is that there are things we cannot study scientifically even if we know they exist.

[Outerhoard.wordpress.com]

This is not a sentence I would use early in a text book about piloting attention.

This made me wonder how you might use it late in a textbook about piloting attention. I decided to make this an exercise and came up with this.

At _HE_, attention is directed at the protagonist of the sentence.
At _WONDERED_, the direction of attention is refined from the protagonist in general to a specific aspect of him: his thoughts.
The listener is also directed to anticipate learning what the protagonist wondered about. Anticipation is really a form of future-directed attention.
At _WHETHER_, attention is refined to a more specific type of wondering: wondering whether as opposed to, say, wondering how.
At _THE MECHANICS_, attention is directed at mechanics, while remaining within the context of the protagonist's thoughts.
At _FIXED_, attention is directed at mechanics fixing something, and the listener anticipates learning what the protagonist wondered whether the mechanics fixed.
At _THE CARS_, the sentence is complete, and attention is directed at the whole scenario.

Would you make any additional remarks?
-------------------
BLOGGER: Wow.

[Karthik Durvasula]

I just happened to see this in SciAm. I thought it would be an interesting read, as a continuation to the discussion of "abstract algebra" in linguistic theorising -http://www.scientificamerican.com/article.cfm?id=why-math-works

Again, I am not saying there cannot be another way of doing things. A lot of people in the biological/psychological sciences have adopted non-formal approaches. However, as I pointed out in an earlier blogpost, the one scientific methodology that has given us incredible returns in the physical sciences is that of abstraction (from the absolute data), and formalisation of problems. If one wants to go against the established trend, one may do so, but they have to show convincing reasons for doing so.
==============================
BLOGGER: Physics envy has been a regular part of the history of science and has paid off well. Lavoisier used it to create modern chemistry, and Lyell brought its standards to geology. But they didn't try to pretend that they could just apply physics idealization wholesale. Every physics idealization can be justified by masses of experiments. They are not just assertions. Chomsky's physics envy is promising, but he shouldn't pretend we have masses of experimental data that justify all his idealizatiions.

[Karthik Durvasula]

"They are not just assertions. Chomsky's physics envy is promising, but he shouldn't pretend we have masses of experimental data that justify all his idealizatiions."

This is just false. There is a tonne of work that is published regularly in theoretical linguistics journals that supports the position. Furthermore, there is really interesting work from Univ. of Maryland and NYU amongst others that is very strongly theory-inspired experimental work (EEG, MEG, FMRI, behavioural experiments) that is adding a wonderful new dimension to the theoretical claims.

I am afraid you aren't aware of these developments, and arguing from a position of ignorance.

It is one thing to have a difference in opinion, and another thing all-together to discredit a whole methodology based on an incorrect "assumption" of a lack of evidence.

I repeat an earlier comment that you often succumb to common folklore about the generative linguistic paradigm, and unfortunately shrink the work of the whole enterprise to a single man. And this is the crux of the difficulty.

[Roberto Gonzalez-Plaza]

pardon for me for asking Mr or Ms Durvasula: could you say what is the wonderful new dimension?

[Raymond Weitzman]

I am very curious to know what "behavioural experiments" Karthik Durvasula is referring to that are adding a new dimension to the theoretical claims of Chomsky.

[Karthik Durvasula]

1) This is not Chomsky's theoretical claims we are talking about; It's an entire field's theoretical claims. ("...unfortunately shrink the work of the whole enterprise to a single man")

2) I am not sure if the last two commenters are showing sarcasm; if it is, that is unfortunate and shows the view point of an uninformed person. If it is not sarcasm and instead an genuine interest, my current comment is irrelevant, and I commend the interest.

3) The wonderful research work I refer to is the work of Colin Phillips (UMD), Alec Marantz (NYU), Pylkannen (NYU), David Poeppel (NYU) and their many past/present students; who have not only shows deep syntactic hierarchies as proposed in generative linguistics, but also shown evidence for "island constraints" which were first developed way back in Ross's dissertation in the late 1960's and have been a very productive venue of research in the generative syntax paradigm since then.

4) The work by Bill Idsardi (UMD), Aditi Lahiri (Oxford) and many others arguing for evidence of abstract phonological representations thru very sophisticated neurobiological techniques techniques.

5) I do want to point out that by defining language as broadly as many on this blog, and by making it almost synonymous with just any communication system, a lot of the discussion remains in a mushy, almost unscientific, even untestable, area. There is a tonne of evidence for both the viewpoint I referred to earlier on and theoretical claims developed since the mid 50's. If you are not aware of them but still think you are holding a nuanced theoretical position, then I am afraid you should be reading more journals and fewer blog posts.

[Karthik Durvasula]

6) I also refers the readers to the work of Bill Labov (UPenn), and Charles Yang (Upenn) who have shown that even the incredible sociolinguistic variation apparent in natural languages is amenable to a generative framework. As they argue, not only is it "amenable", it is in fact best accounted for in such a fashion.

[Karthik Durvasula]

All right, the unnecessarily antagonist tone of point (5) was probably not helping the discussion, and if anything would distract everyone from the meat of the matter. I apologise for it.

[Jerry Moore]

Regretively, Colin Phillips’s work is lacking the basic theoretical foundation except from few speculative remarks of N.Chomsky and is vague going around neuro and motor processing of various grammatically driven speech constructs. I wonder how they define Language: as some attribute of speech practice?, grammar processing? Any idea? His discovery of phonemic mismatch is not making for Linguistics more then graphemic mismatch in: “The rset can be a taotl mses and you can sitll raed it whotuit a pboerlm.”

[Karthik Durvasula]

@Jerry Moore: I won't argue with the first half of your comment because it appears to be your opinion. I can add this much though: this is NOT an opinion shared by modern theoretical linguists, who see the work as in fact very-well laid of theoretically.

Re your question: the Phillips camp, as far as I know, will be one of the last people to think of language "as some attribute of speech practice". Their view is more about how to study language - which is by taking theoretical entities seriously, and testing them (than the rather mundane task of defining what language is - which is mostly a pop-science issue to a professional theoretical linguist. (Disclaimer: I do acknowledge, it appears to be an important question for anthropologists).

The second comment about "graphemic mismatch" and how it relates to "phonemic mismatch" seems to be way off the mark. I am guessing you are referring to the Mismatch Negativity (MMN) experiments conducted by the Phillips et al. MMN as an MEG/EEG paradigm has a relatively long and respectable history in the brain-imaging literature, and what Phillips et al do is to say if you extend the logic of MMN, you should get to abstract phonological represntations, which they go ahead and show. Not only do their experiments have a sound theoretical (linguistic) footing, their experimental paradigm (MMN) and their linking hypothesis all flow very nicely out of what we think we know about MMN's- which is what makes their result especially awesome!

To be quite honest, I can't imagine that you have honestly read the papers carefully and come to the conclusion. I almost wanna believe there is some trolling here (in Jerry Moore's name). But, if you really are Jerry Moore, and you honestly believe that the second part of your comment is evidence-based, then I can tell you that you are neither with most (if not all) experimental neurobiologists nor with most (if not all) theoretical linguists on this matter - and this is from a person who has participated in both kinds of research, not a non-professional linguist.

[Jerry Moore]

Karthil Durvasula, did you ever tried to formulate what is research of Chomsky et al. about? Language? Speech?
If it is targeting Language (as the plural for all spoken languages), it is tragically lacking Universals, as the most of results of the study statistically circumcised and hardly able to substantiate a qualitative approach.
If it is targeting Language (as the “biological presets” or “organ” [N.C.] for the grammar acquisitions), it supposes to be brave enough to state principal role of the formal (grammatical) structure and clear explain the derived or secondary status of the semantic hub of such “presets”. Till then neurolingustic part of that study is dealing with the unclear types of stimuli.
Few equations. You wrote:
“modern theoretical linguists, who see the work as in fact very-well laid of theoretically”
Can you advance us to that theory? In brief. It not supposes to be a secret one.
You wrote:
“the Phillips camp (I like your word “camp” J.M.), as far as I know, will be one of the last people to think of language …(cut)... Their view is more about how to study language - which is by taking theoretical entities seriously”
Could you point as any of their developed methodology or somehow summarise their advances in study?
You wrote:
“the rather mundane task of defining what language is - which is mostly a pop-science issue to a professional theoretical linguist”
Are you serious? With philosophy like this it is no wonder that main achievements of this branch of science (2 billion dollars pull of mainly taxpayers’ money) is the development of all sort of talking vending machines and automated telephone services so far.
On this note, please, let me leave unchallenged the last part of your comment.
----------------------------
BLOGGER: i don't mind commenters yelling at each other, but this dspute seems to me to have run out of gas.