yesterday i gave an impromptu explanation of deep structure to explain why Neal Stephenson's usage of the idea in Snow Crash was extremely shitty and bad and he didn't know what he was writing about

ok so basically here's my explanation of Deep Structure from yesterday:

if you start from the hypothesis that human languages can be best described with Context Free Grammars, and you try to model certain phenomena, such as English auxiliary verbs, you get some interesting tricky stuff

for example, consider the following example sentences with auxiliaries and negation:

John will run
John will not run
John has run
John has not run
John is running
John is not running

John will have run
John will not have run
John will be running
John will not be running

John will have been running
John will not have been running

as you can see, the negation comes after the first auxiliary, no matter how many auxiliaries there are

and you can explain this by say well, there's just lots of rules for how auxiliaries work. like, maybe it looks like this:

AuxComplex -> Will | Will Neg | Will Perf | Will Neg Perf | Will Prog | Will Neg Prog | Will Perf Prog | Will Neg Perf Prog

so you can certainly explain the pattern with a giant list of the possible arrangements

but thats an inelegant theory and prooobably that's not what people "know" about english, right

like, if a kid only ever saw senteces with either 2 auxiliaries, 2 auxes + negation, and 3 auxes, but NO sentences w/ 3 auxes + negation

eg if a kid only saw

Aux Aux
Aux Neg Aux
Aux Aux Aux

probably the kid would negate the last one as

Aux Neg Aux Aux
or
Aux Aux Neg Aux

so like, probably people are learning precisely what i described earlier:

Negation comes after the first auxiliary

so Chomsky proposed that the way you can model that is by starting from a tree generated by a Context Free Grammar

lets say something like

AuxComplex -> Neg? AuxV+

and then you apply some rewrites on that tree, some "transformations"

eg a rewrite such as

rewrite the tree (AuxComplex Neg AuxV AuxV*) to (AuxComplex AuxV Neg AuxV*)

so for example, the theory would say that the best way to model the sentence "John has not been running"

is with two steps:

1) Generate the tree for "John not has been running"

2) Rewrite the aux complex and produce "John has not been running"

within that approach to modelling grammars, called Transformational Generative Grammar, there are usually *many* transformations/rewrites that can apply, so the whole thing looks like

1) generate an initial tree with a CFG
2) apply rewrites
3) no more rewrites apply, youre done

within TGG, we give names to the (1) and (3)

namely, the initial tree you start with is called a Deep Structure, and the final tree you end up with is a Surface Structure

the analogy being you can "see" the trees in (3) more easily, like stuff on the surface of the ocean

while things "deeper" in the ocean are hard to see because the light is obscured as it travels through all that water

so that's what Deep Structure actually is:

the ASTs you generate before you do all your pretty printing to spit out the stuff you actually say

that's it. it's not magic or anything, it's just ASTs before any rewriting is done to them

but Neal Stephenson's write up of Deep Structure, which took place in the late 80s long after linguists stop talking about Deep Structure, totally misrepresents this, and instead gives it a kind of woo-y mystical quality

first there's this whole nonsense about Universals and Deep Structure having never been found etc etc

which is not even wrong

Universals and Deep Structure haven't been found in precisely the same way that electrons havent been found: they're *hypothesized*

we believe they exist to the extent that they are useful theoretical constructs. neither Deep Structure nor the Electron is something we ~see~

but he also goes on to say shit about how "Deep Structures" are the innate structures of the human brain that allow the brain to carry out formal operations on strings of text

and like, no, that's.. no

MAYBE you want to say that Deep Structure is *explained by* the presence of innate structures in the human brain

but the Deep Structures themselves are just the pre-transformation ASTs

Stephenson makes a looot of references to George Steiner, a french literary theorist who wrote about language and society, and also to Emmon Bach and Noam Chomsky

but he seems to reference Bach and Chomsky *via* Steiner

which leads me to believe that Stephenson never read either of Chomsky or Bach, both of whom are syntacticians not literary theorists, and that Steiner's interpretation of Chomsky and Bach is, in typical postmodern woo fashion, full of all sorts of nonsense misunderstanding

or maybe Stephenson misunderstood Steiner, which i think is less likely than Steiner getting the theory wrong

but Stephenson saw a good, fun sounding, Snow Crash compatible thing, and he ran with it, without looking into it more deeply, and that's unfortunate

at the end of the day, Deep Structure is not magically interesting. it's not some ~deep~ insight, and chomsky regretted using that name b/c people are dopes.

Deep Structure is something that is conceptually quite simple and boring

--- END DEEP STRUCTURE RANT ---

addendum: looking more into steiner and his connection to chomsky, i find this 1975 NYT article:

https://www.nytimes.com/1975/06/08/archives/george-steiner-on-language-and-languages.html

in it, there's explicit discussion of steiner critiquing chomsky and deep structure

but ALSO: it discusses the idea of the Story of Babel

ie the question of linguistic diversity

well, the part of Snow Crash that I'm talking about? it's embedded within a larger discussion of Babel

Stephenson talks about Steiner and Chomsky and Deep Structure in the context of Babel, and LO: a news article about precisely that!

and the news article is well older than Snow Crash

i bet you so much much money that Stephenson read this article and used it as the basis of that part of the book

or maybe Stephenson had read After Babel instead

who knows. i mean presumably he knows, but

also worth pointing out that Steiner was a fan of Whorf, and Whorf was a fool and an exoticizing racist

Whorf has this whole spiel about how Hopi speakers have different perceptions of time and don't even think about or understand time, b/c their language lacks time words/etc

some Hopi folx worked with actual linguists and produced a 700 page long book of hopi time words and grammar

whorf just never "noticed" it because he saw what he wanted to see: weird exotic ~natives~ with their ~strange~ languages and ~strange~ ways of thinking

and its like nah dude you're just racist. yeah they talk differently, language is fun like that, stop trying to make them out to be space aliens or something, jesus. they're just people

whorf also had this whole long thing about how like hopi or maybe navajo speakers, i forget which, were like.. perfectly suited for doing relativistic physics because their language makes "no distinction" between time and space, just like the theory of relativity! WOWWW :O

and like, what really happens is that their language makes extensive use of a time-space metaphor. it's really cool how it does it actually

except that metaphor isn't unique to their language. ENGLISH DOES IT TOO, just with different details

so you have this whole batshit racist theory of how weird and foreign and alien native americans are wowowow

and its actually just from stupid americans not looking at their own languages and cultures with the same perspective

its the racist anthropological gaze -- you can only see a thing in groups you consider Other, or Alien, and you can't see it in yourself because of course you're Normal and no one needs to study Normal People, only those Weirdos

fwiw i do think its interesting to READ whorf's discussions of these things

his descriptions are fascinating. but the conclusions he draws about the cultures and the people are just.. out there

native american languages are cool

there's a cool @LanguageLog post from back in the day on Anishinaabemowin (sometimes called Ojibwa, or more old-time-y and possibly more racistly?? called Chippewa)

really worth a read. Anishinaabemowin is neat

http://itre.cis.upenn.edu/~myl/languagelog/archives/002001.html