How would you diagram this English sentence?

[Matt]

That we left should impress Phoebe.

A friend asked me how I would draw a syntax tree for this sentence and I'm stumped. Is that we left behaving as an NP here?

[zompist]

Yes, of course it is. It's the subject of the sentence.

What exactly the tree looks like depends on your favorite syntactic theory. However, it should be exactly parallel to "It should impress Phoebe".

The simplest explanation is that "that" is a subordinator used to turn a clause into an NP, exactly as in "I said that she was crazy".

A fancier explanation is that the subordinator turns the clause into a modifier ("the fact that she was crazy", "the idea that we left") and that the head noun is omitted. But I'm not sure that actually gains us anything.

[Matt]

A fancier explanation is that the subordinator turns the clause into a modifier ("the fact that she was crazy", "the idea that we left") and that the head noun is omitted. But I'm not sure that actually gains us anything.

That was my first thought but I wasn't sure how to fit that into a syntax tree.

Thanks!

[RTLPM]

That we left should impress Phoebe.

A friend asked me how I would draw a syntax tree for this sentence and I'm stumped. Is that we left behaving as an NP here?

It sounds like a vestige of English subjunctive to me: in French or Spanish, it would use it.

Does this sound right? -> That I were here should impress her.

I know we say: I wish I were... If I were to go...

[Dewrad]

That we left should impress Phoebe.

A friend asked me how I would draw a syntax tree for this sentence and I'm stumped. Is that we left behaving as an NP here?

How complicated do you want it? As Zomp says, how the tree looks depends on the theory you use, and how detailed you want to get. However, a fairly theory-neutral tree might look something like the following:

[Rory]

Using CPs and strict binary branching is hardly theory-neutral. Here's a more neutral tree, in my opinion:



The only contentious thing here is my treatment of "that" as a Determiner; I was considering just leaving the whole initial NP as an unexpanded triangle, but "vague" is not the same as "theory-neutral".

EDIT: While we're at it, here's a (bad) attempt at a tree (technically a proof) in Categorial Grammar:

[Dewrad]

Using CPs and strict binary branching is hardly theory-neutral. Here's a more neutral tree, in my opinion:

I did say "fairly", not "wholly" There's no raising, movement, vP-shells or IP's in there (primarily because none of that's relevant to the issue at hand). Can't fault a man for how he was raised (or taught.)

[zompist]

FWIW I'd probably diagram the tree as Dewrad did, though I'm agnostic about the labels for the higher nodes. It's interesting that the actual structure is identical to Rory's second tree.

By the conjunction test, "we left" and "please Chloe" are both constituents, as there are variants like

That we left and Johnny arrived should please Chloe.
That we left should please Chloe and annoy Mila.

[Rory]

I think I saw the combination of "theory-neutral" and "CP" and felt like I had to do something. Lambek calculus was the logical solution. Anyway, looks like we're all agreed about the structure, it's just putting labels on things that is controversial - which is a summary of most debates in theoretical linguistics, imho...

[Radius Solis]

A fancier explanation is that the subordinator turns the clause into a modifier ("the fact that she was crazy", "the idea that we left") and that the head noun is omitted. But I'm not sure that actually gains us anything.

I don't know that it does either, but the NP treatment is not flawless. If "that"-clauses really are NPs you'd expect to be able to make a PP out of one, but as far as I can think of, no preposition permits such a thing: *with [that we left].

But this certainly doesn't help the "omitted head noun" approach either::

because of [the fact [that we left]]
*because of [0 [that we left]]

[ná'oolkiłí]

Yes, of course it is. It's the subject of the sentence.

What exactly the tree looks like depends on your favorite syntactic theory. However, it should be exactly parallel to "It should impress Phoebe".

The simplest explanation is that "that" is a subordinator used to turn a clause into an NP, exactly as in "I said that she was crazy".

A fancier explanation is that the subordinator turns the clause into a modifier ("the fact that she was crazy", "the idea that we left") and that the head noun is omitted. But I'm not sure that actually gains us anything.

If I may be a little pedantic, [That we left] is not a NP (or a DP). It is a CP; a complementizer, not a noun, heads it. The that doesn't turn anything into a NP, it's simply the case that often verbs can select CP instead of a DP as an argument. As for "the fact that she was crazy", fact is a noun that can take a CP complement; ie [_DP the [_NP fact [_CP that ... ]]]—the clause isn't a modifier any more that fish is a modifier in "she eats fish".

[Radius Solis]

I don't know enough about the theory to debate the general merits of that analysis, but it does appear to address the problems I brought up.

That said, you're coming on a bit strong with all of your "is" and "simply the case", as though that approach were inarguable (as Rory makes clear it isn't).

[zompist]

Yah, I've never heard of a CP. McCawley simply labels it an S.

[merijn]

The use of theory neutral trees is in itself a contradictio in terminis, since not all frameworks use trees. Usually when somebody says they are drawing a theory-neutral tree, they are usually drawing a tree how Mainstream (Chomskyan) Generative Grammar did it at some point in the past, minus the movement. That is the case for Dewrad's tree and I think that it is the case Rory's tree, his just goes back to an earlier stage. That said, I think that amongst those theories that do use trees they would have the same binary branching tree as Dewrad's tree, without necessarily saying that all trees are always binary branching. I know that HDPG has adopted the idea that there are CP's, and that the complementizer is the head of that CP, and I *think* that LFG did the same. It is perhaps telling that Rory's tree of Categorial Grammar is exactly the same as Dewrrad's tree, even the heads of the phrases of Dewrad's tree correspond to what corresponds with heads in Categorial grammar, (that is where the the formula right to the slash is the same as the whole formula of to the dominating knot). So all in all i'd say that Dewrad's tree is more theory-neutral than Rory's tree.

And I think that all frameworks worth their salt make the distinction between what a CP is in Generative Grammar and what is a DP or NP is in Generative Grammar, their distributions are just too different to treat them the same.

[Radius Solis]

The use of theory neutral trees is in itself a contradictio in terminis

Only if you want to take the term much too literally for it to remain useful as a description of anything. Though I've never seen the definition of "theory-neutral" explicitly discussed, from context I have gathered that it's more about whether something is controversial between competing theories than whether it arose in any particular one.

[merijn]

There are quite a few linguists who believe that a tree structure is not an adequate way of describing grammars of natural languages, for instance lots of people who follow various types of Dependency Grammars (and many more who believe that tree structures alone are inadequate). Of course you are right that it is a useless term if used absolutely, because no analysis can be theory-neutral that way, and people (including me) tend to use it relatively rather than absolutely. My point was that Rory's tree wasn't theory-neutral in the absolute sense either, and that in the relative sense nowadays Dewrad's tree may be more theory neutral (saying things that are less controversial between different frameworks) than Rory's.

[Curlyjimsam]

It acts very much like an NP, but I don't think you'd want to write it as such - you'd probably want to call it a CP, as Dewrad does. However, it still occupies the same position in the tree as a subject NP would.

[linguofreak]

Yah, I've never heard of a CP. McCawley simply labels it an S.

In my syntax course last fall CP (complementizer phrase) was used to replace S pretty much everywhere, including main clauses, the theory being that subordinate and main clauses should have the same structure, and that since subordinate clauses are headed by a complementizer such as "that", and this complementizer can sometimes be omitted (eg, "I think that I'll go to the store today" vs. "I think ∅ I'll go to the store today"), therefore, main clauses must contain a complementizer too, and, since a main clause of the form *"That I'll go to the store today" is ungrammatical, the complementizer in main clauses must always be ∅.

I personally think the theory is too clever by half (at least in the form it was taught in in my syntax course). I hate theories that treat less common cases (like subordinate clauses) as the rule, with a deep structure and surface structure that more or less match, and less common cases (like main clauses) as the exception, with lots of hidden goodies in the deep structure that don't match the surface structure. (For some reason this type of exception-as-the-rule-and-rule-as-the-exception thinking seems to be really common with regards to main and subordinate clauses: It also shows up in the argument that German is SOV, which I also hate).

[Yng]

There's also fronting, which in some theories is explained by movement to the CP slot. Which is convenient for English but not necessarily for other languages (all too frequently a problem with syntax theories, of course).

[merijn]

At some point in the history of Mainstream Generative Grammar the functional categories dominating VP "exploded", from being just IP and CP to now some people believing there are more than 12 such categories. People are in disagreement to what extend these categories are universal, and to what extend these categories are always present, but most people agree that there are 3 domains; the functional categories immediately dominating the VP including the vP (that is little v) form the thematic domain, here is everything located that has to do with the thematic structure of the verb. The functional categories dominating the thematic domain form the inflectional domain; here tense, mood and aspect is dealt with (as well as subject agreement in most western languages). The highest domain is the complementizer domain, this is where everything is regulated that has do with the context of the clause, this means the complementizers for subclauses, topic and focus, and wh-fronting. Some people have proposed exceptions to this idea of three domains; there are people who argue that aspect is dealt with in the thematic domain and not in the inflectional domain, and some linguists have proposed a "low focus" position, a focus position that is IIRC in between the thematic domain and the inflectional domain.


++++++


wrt German (and Dutch) SVO versus SOV, if you believe SOV is the underlying order, you have two exceptions:
1. questions, for those you need to move the inflected verb to the front
2. main clauses, for those you need to move a topic to the front and the inflected verb to the second position

If you believe that the underlying word order is SVO you have the following exceptions
1. main clauses, for those you need to move the not-inflected verbs to the back
2. main clauses with a topic other than the subject, for those you need to move the topic to the front and the inflected verb to the second position, in addition to moving the non-finite verbs to the back
3. questions, for those you need to move the inflected verb to the front, as well as moving the non-finite verbs to the back
4. infinitival phrases (Es ist schwer, dich zu vergessen) move the infinitive to the back.

As you can see there are much more exceptions if you follow SVO as the underlying order, than if you follow SOV, and what's more exceptions 1 and 2 of the SOV order, that is all its exceptions, are the same as 3 and 2 of the SVO as underlying order. In other words, there is nothing to gain from analyzing German as underlying SVO

[linguofreak]

Except for the fact that *every* sentence has a main clause with an inflected verb, which, except in questions, always resides in the second position, whereas *not* every sentence has a non-inflected verb, or a topic other than the subject, or an infinitival phrase, or a subordinate clause (which you missed in your list of exceptions for SVO). Sure, with SVO you have more *situations* in which exceptions appear, but with SOV, the *single most common* situation, appearing in *every flippin' valid sentence* is an exception.

What you gain from analyzing German as underlying SVO is that *there exist sentences that do not contain exceptions*. An SOV analysis guarantees that every single sentence in the German language will have at least one exception.

[zompist]

Since reading Tomasello, I can't take the "this must be a transformation because the analysis is more elegant" argument as seriously. I suspect most transformations are historical artefacts, much like relics of sound change.

As Tomasello points out, for instance, English-speaking children tend to learn questions first, before statements, which makes it unlikely that they're deriving them via transformation.

[linguofreak]

Since reading Tomasello, I can't take the "this must be a transformation because the analysis is more elegant" argument as seriously. I suspect most transformations are historical artefacts, much like relics of sound change.

As Tomasello points out, for instance, English-speaking children tend to learn questions first, before statements, which makes it unlikely that they're deriving them via transformation.

Wait, are you replying to me or merijn?

If to me, I don't see it so much as "instances of SOV in German must be transformations from SVO because the analysis is more elegant", as "instances of SOV in German must be transformations from SVO because the opposite analysis leads to absolutely absurd conclusions." (Namely, that there is an exception to the "normal" word ordering of German in every valid sentence).

The "what do children learn first" argument also applies here: Sentences with a main clause and a single inflected verb are going to be learned before subordinate clauses and sentences with uninflected verbs.

I find it interesting (although somehow not *entirely* surprising) that children learn questions before statements. If that extends to German speaking children as well, I'd be more inclined to accept an argument that German is underlying VSO than underlying SOV (which isn't to say that such an argument wouldn't be absurd, just that an argument for underlying SOV is more absurd).

[zompist]

I was replying to merijn, but it applies to you too if you believe in transformations.

Tomasello says-- based on detailed study of child language acquisition-- that children learn language construction by construction. It's surprisingly late (school age) before they learn abstractions and generalize what they know about syntax.

So I'd actually suggest that German is neither SVO nor SOV. Children learn how questions work, how simple statement work, how clauses work, by direct imitation... there is no need to suppose they are applying transformations at all. There is never a point where they are forced to reorganize their knowledge, no point where they have to decide on a single underlying order.

(To be precise, by 'constructions' he means something even narrower than 'questions' or 'statements'. He suggests that verbs are learned one by one, for instance. The child doesn't start with concepts like subject/object or agent/patient; he starts with things like the verb 'give' and the roles it has-- giver, gift, recipient. It's not till school age that concepts like subject/object are really understood.)

That isn't to say that transformations are meaningless; they can be valid patterns in the language even if speakers do not actually use them. (One way this could happen is if they're historical relics, as I said, but there can be other explanations.)

[Radius Solis]

Sounds like an interesting book. And suspiciously close to what I've been thinking, too, ever since I started reading more about syntax - that the whole idea of transformations may instead be a giant misinterpretation of what's merely evidence left behind by historical changes, just as allophony is nothing more than self-perpetuating artifacts of old sound changes. And so I've been thinking for some time that maybe a better way to investigate syntax might lie in dropping the "synchronics only, please" ethic and examining grammar's diachronic evolution, modeling the situation as akin to that of sound change. Because that could potentially shed some light on it.

I also hold that if people are going to declare they have a theory of cognition, they should have gained it by studying cognition. So in that regard this Tomasello gets my thumbs-up even before I know who he is.