Scientific American says Universal Grammar is dead: a response

squirrel sitting on wooden pallet

The following is a small break in sewing/knitting fodder. It is of interest to no one but myself. I’ve written it as though someone might actually read it. Pah!

In a previous life, I was a linguist. And once a linguist, always a linguist. This means that when people say silly things about language, I immediately revert into uppity, arrogant linguist mode. It’s an affliction for which there is no cure. It is the herpes of academia.

The trouble is, of course, that people say A LOT of silly things about language. This is through no fault of their own however. Most people aren’t scientists and most scientists aren’t linguists. Most people have no real understanding of the mechanisms for studying language through the lens of modern science. So I am prepared to accept a certain level of nonsense from the layman, even when it makes me twitch.

The Inuit have hundreds of words for snow. Uh no, they don’t.

Everybody uses language but very few people have spent any time thinking about what “language” really means. I have a lymphatic system, for example, but haven’t a clue as to what it actually looks like. I defer to scientists, experts in the field of lymph (Lymphists? Lymphicians?) for all my lymphatic info.

So when I read this article (Ibbotson and Tomasello), published in Sept 2016 in Scientific American, my spidey sense got to tingling at maximum level. You’d think an article on linguistics published in a reputable SCIENCE magazine might provide some real insight into the state of the art. It should be, at the very least, well researched. It should espouse arguments that are coherent. Right? Yeah, I think so.

The SA article claims that recent research has discredited Noam Chomsky’s theory of Universal Grammar and that scientists are happily throwing it away. It’s a flashy claim, sure to excite those who view Chomsky as a hoity toity, anarchist, intellectual cray-cray. (To be clear, I know virtually nothing about Chomsky’s politics. And to be extra clear, his politics have nothing to do with his research in linguistics. Like, really, actually nothing. I swear.)

The main problem I have with the article, much like the problem with the claim that the Inuit have hundreds of words for snow, is that the people behind it don’t seem to actually know what they’re talking about.

Now I should, at this point, admit that I have been out of the linguist game for quite some time. My syntax is pretty rusty, my memory of terms like “recursion” and “merge” is pretty fuzzy. But while my technical skills might be a little outdated, my academic herpes precludes me from staying mum on such important subjects. It is an unfortunate side effect. And yes, I’d be hard pressed to draw any form of syntactic tree these days, but rest assured that my understanding of the basic tenets of modern linguistics is still, in my opinion, quite sound.

Before I started typing this response, I made a point of relistening to this lecture by Chomsky. I attended a similar lecture at UQAM a few years ago and had been meaning to revisit it.

So, what exactly does the SA article claim? Well, to be honest, I’m not really sure I know. And I’ve read it through a bunch of times. I’ll do my best to summarize a few keys points and I’ll try to rebut them with some interesting quotes taken from the Chomsky lecture.

From what I can gather, the article claims that instead of a language specific part of the brain (Universal Grammar or UG), we are endowed with general learning capabilities that allow us to learn language but are not specific to language. These include:

the ability to classify the world into categories (people or objects, for instance) and to understand the relations among things. These capabilities, coupled with a unique human ability to grasp what others intend to communicate, allow language to happen.

In order to argue this, the authors would therefore have to show how general learning capabilities, applicable to other known learning behaviours, can also account for the way we learn languages. Do they do this? Unless I missed something… uh, no.

One strange example they give is the following:

For example, a recent study co-authored by one of us (Ibbotson) showed that children’s ability to produce a correct irregular past tense verb—such as “Every day I fly, yesterday I flew” (not “flyed”)—was associated with their ability to inhibit a tempting response that was unrelated to grammar. (For example, to say the word “moon” while looking at a picture of the sun.)

Uh. Hmm. Okay, I’m not sure I get the link here. I haven’t looked at the original study, but I fail to see how this proves their point. Sure, it’s about lexical items. Okay, so at a certain developmental stage a child gets good at remembering particular lexical items (including exceptions to past tense forms). I wonder if this developmental phase might also be linked to other interesting things. For instance, the child starts to say “flew” around the same time they get good at eating spaghetti without getting sauce on their forehead. Or at the same time they get good at going potty.

Ugh, I don’t know. I see the basic link between the two things, but I don’t understand how this disproves that we are innately endowed with a language specific grammar. What is this meant to show? And also, can someone tell me why children are “tempted” to say moon when they see a picture of the sun? (Are all children predictably devious in this way?)

The main problem here is that the authors of the SA article fail to show that they have any understanding of the object of study in this case: “language”. At least, they don’t seem to understand what Chomsky (and LINGUISTS) mean when they say language. Or maybe – I’ll give them the benefit of the doubt here- they disagree with the definition. But if that’s the case, how can there even be an informed debate? If I study Volleyball and you study Basketball, how will we ever agree on the rules? Noam saying?

So, what then is UG? What is language? What does Chomsky actually say about this?

It’s perfectly obvious that there is some genetic factor that distinguishes humans from other animals and that it is language-specific. The theory of that genetic component, whatever it turns out to be, is what is called universal grammar.    (This quote isn’t from the lecture.)

The SA article does seem to agree that language is unique to humans. They say that “if you understand what language is, you comprehend a little bit more about human nature”. Yes, I agree. Chomsky agrees. Linguists agree. Rejoice.

Chomsky also says:

The “basic property” is such that “each language provides an unbounded array of hierarchically structured expressions […] and each language incorporates a computational procedure satisfying the basic property”, in other words, a generative grammar.

It’s this basic property that makes language so unique. It turns out to be unlike any other mental function like “categorisation” or “understanding the relations among things”. Well, so say Chomsky et al.

So the language faculty uses some kind of computational procedures to produce structured outputs. The output is infinite in nature (you can potentially create an infinite number of sentences) and is therefore generative. The fact that the output is “structured” is a slightly less obvious conclusion but one that is pretty easily demonstrated. I won’t do that here, so you’ll just have to trust me.

It’s the details of these computational procedures that are being worked on by linguists.

Oof, ok. So those are the basics. What then are the specifics of the so-called evidence against UG that Ibbotson and Tomasello invoke?

The authors discuss how Chomsky has most recently suggested that what is unique about UG (the language specific component of the human brain) is its use of something called computational recursion. The authors claim that Pirahã, an Amazonian language, “gets by without Chomskyan recursion”. Again, the article claims that this is a checkmate for UG.

I won’t detail it here, but as it turns out, Pirahã does actually make use of recursion. BOOM. Because, well… the authors don’t seem to know what recursion really means. BOOM. Those interested can read this and this to learn about recursion and Pirahã.

But even if it were the case, that some detail of the grammar, as posited by linguists, turned out to be falsified by a newly discovered language, that does nothing to disprove UG. We used to think bloodletting was a good idea. It wasn’t. We revised our theory. What we didn’t do was throw out the notion of a circulatory system. (Not that I think recursion is akin to bloodletting. Just trying to use some understandable analogies.) A newly discovered language may (and often does!) disprove some mechanism linguists had originally come up with, but it does nothing to disprove that innateness of the basic property.

The authors of the article really don’t get this fact. They argue that because languages studied in the last decades have revealed that the earliest notions of syntax, grammatical categories, etc. were insufficient, this somehow means the whole enterprise is moot.

They say:

These so-called outliers were difficult to reconcile with the universal grammar that was built on examples from the European languages. Other exceptions to Chomsky’s theory came from the study of “ergative” language, such as Basque or Urdu, in which the way a sentence subject is used is very different from that in many European languages, again challenging the idea of a universal grammar.

Huh?

Chomsky doesn’t see syntax the way he saw it in the mid 20th century. Linguists have come up with new incarnations of the grammar to account for the widening breadth of evidence. They haven’t decided that because two languages are different we must not actually be predisposed to language. No, instead, they’ve modified their theory of what that innate component looks like. Because, ooh look, a new piece of evidence challenges the way we understood that amazing genetic endowment that is the language faculty!

Below the comment about ergative languages, the authors have included a wholly nonsensical figure comparing “Chomsky’s Universal Grammar” with “Usage-based learning”. I can’t make heads or tails of their argument here at all. Beside the “usage-based” illustration they say “this theory of building up knowledge of word meaning and grammar approximates the way the two- and three-year-olds actually learn language”. So, under their theory, children learn word meaning and grammar. How on earth does this contrast with Chomsky’s view? Through exposure to language around them, children learn lexical items and the particulars of the surface forms of the target language: word meaning + grammar.

The authors also introduce the figure by saying: “Chomsky’s theory has gradually been challenged by new theories asserting that language is acquired as children discern patterns in the language they hear around them.” Pardon the expletive, but NO SHIT. The key here, of course, is what they mean by “pattern”. They obviously don’t mean syntactic pattern in the linguistic sense (i.e. a big fancy many branched tree). But then, what do they mean?

They include an example using these two sentences: The dog wants the ball; The dog wants food. The child observes the dog wanting a ball, then observes a dog wanting the food and concludes that “the word “food” might replace the word “ball” after the phrase “the dog wants””. Interestingly, the way they illustrate these two example sentences is with a diagram that looks suspiciously like the beginning of a syntactic tree. Hmm.

Confused yet? I am. So let’s instead approach the subject from a different perspective: what ISN’T language according to Chomsky?

Externalization is an ancillary process. […] Particular uses of language that depend on externalization (among them communication) are even more peripheral [to the scientific study of language or UG]

And what he means here by “externalization” includes speech, as in, the sounds that come out of your mouth when you’re talking. So, the subject of inquiry of linguistics ISN’T really speech or what we use speech for (communication). Instead, it is the mental construction that precedes this externalization.

And in the case of the dog wanting the ball, Chomsky would assert that there is “no connection between an internal element and a mind independent entity [it always involves] some mental construction”. So, the word “ball” refers to a mental construction “ball” and is something different from the “ball” out there in the real world. Linguists aren’t interested in real balls. They aren’t even really interested in the word (read: mouth movements + sound waves) “ball”. These two other things are interesting to linguists only insofar as they give clues to the true nature of the mental construct “ball”. It’s a fine distinction, but an important one. (It probably takes a graduate degree to sort through this. I don’t claim my ball thoughts are the definitive thesis on the matter.)

I understand that this is not the easiest concept to grasp. I mean, I communicate with language all the time. I refer to things in the world. Isn’t that what it’s for? Aren’t language and communication one and the same?

Chomsky’s comments on the origins of modern science are useful to keep in mind here. As he points out, through science we discover that:

“Our beliefs are mostly senseless and our intuitions are all wrong.”

When we allow ourselves to be puzzled by seemingly simple and obvious phenomena, we can often be surprised by the conclusions we must draw. My intuition is that the sun goes round the earth. It really seems that way. That’s what I feel. (So it must be true, n’est-ce pas, Trump?).

Now, I’m sure Ibbotson and Tomasello are very smart guys. They probably have a lot of interesting and illuminating things to say about their research interests specific to their fields. What they are not, however, are linguists.

In the lecture, Chomsky talks about recent papers on language published in the journal Frontiers of Psychology that ignore in their definition of language the “basic property”. Instead, they characterize language only as “the full suite of abilities to map sound to meaning”. A pretty paltry definition. I can map the sound of my cat’s meow to the meaning “I’m hungry, feed me, or else”. I can map the sound of a fart to the meaning “poop approaching”.

No biologist would study the evolution of the visual system […] assuming no more about the phenotype than it provides the full suite of abilities to map stimuli to percepts. You couldn’t get anywhere with that. And you couldn’t publish a paper on it either. The editors in the audience might take that to heart.

I’m looking at you, editors of Scientific American.