Linguistics 522

Lecture 5

X-Bar Syntax

Phrase-structure rules: Complements, Specifiers, and adjuncts

Much of our energy last time expended on motivating a constituent intermediate between N and N'' in NPs. That was N'.

We argued that there were two sorts of Noun modifiers that could attach to a nominal constituent to make up an N', complements and adjuncts.

We argued that that when you attached a D (determiner) to an N', that made it an N''.

Now we argue analagously for AP, VP, and PP.

  1. We argue that there are constituents intermediate between the heads and full phrases in AP, VP, and PP. These are A', V', and P', respectively.
  2. We argue for the complement/adjunct distinction in V'. What about the others?
  3. We argue for something analagous to D in N'', something that added to A', V', P' makes a A'', V'', P''.

X-Bar Syntax

Stated in terms the phrase-structure rules we have been using till now, we have sets of PS rules, all with analogous structures:

    1. A'' => Spec A'
    2. A' => A' Adjunct
    3. A' => A Complement
    1. V'' => Spec V'
    2. V' => V' Adjunct
    3. V' => V Complement
    1. P'' => Spec P'
    2. P' => P' Adjunct
    3. P' => P Complement
We use Spec (for Specifier) for the constituent added to the single bar level category to make t he double-bar level.

Using X to stand for ANY of the major categories, the pattern is this:

  • [X'' Spec [X' [X' X Complement] Adjunct]]
Voila. X-Bar syntax.
and adjuncts

  1. Specifiers italicized:
    1. AP: how proud of her son was she?
    2. PP: right under the bridge is where you'll find it.
    3. VP: She must be thinking of you.
  2. Complements italicized:
    1. AP: How proud of her son was she?
    2. PP: Right under the bridge is where you'll find it.
    3. VP: She must be thinking of you.
  3. The only obligatory element of a phrase (generally) is the lexical head. Phrases withOUT complements or specifiers;
    1. NP: Students rose in protest,
    2. AP: She was proud.
    3. PP: She went in,
    4. VP: She must think.

We use the term subcategorization to denote the kind of complement a category takes as complement. We have seen that verbs can subcategorize NP (transitive verbs), PP, AP, and S.

  1. John devoured the trout.
  2. John despaired of an answer.
  3. Mary seemed kind.
  4. Mary proved that she was inoocent.

We note in passing that some verbs subcategorize V' (not V'' or V):

  1. I saw John run down the road.
  2. *I saw John be running down the road.
  3. *I saw John have finished his work.

Arguments for V Complement versus adjunct distinction

Preposing Preposing test. Adjuncts prepose easily.
  1. He decided on the boat. (ambiguous)
  2. On the boat, he decided. (one reading)
  3. He explained last night.(ambiguous)
  4. Last night he explained. (one reading)
  5. He laughed at the clown. (complement)
  6. He laughed at the office. (adjunct)
  7. *At the clown he laughed. (complement)
  8. At the office he laughed. (adjunct)
  9. He worked at the office. (adjunct)
  10. He worked at physics. (complement)
  11. At the office he worked. (adjunct)
  12. *At physics he worked. (different reading)

Only complements may pseudo-passivize:

  1. Physics needs to be worked at.
  2. *The office was worked at. (If office is the place where work happened)
  3. The clown was laughed at.
  4. *The office was laughed at. (If office is the place where laughter happened).
  5. The boat was decided on after lengthy deliberation. (no reading on which the boat is the place where trhe deciding happened)
  6. Last night couldn't be explained by anyone. (No reading on which last night is the time at which something (else) was explained.
Do-so anaphora

It appears that do so anaphora is to wo V-bars what one-anaphora is to N-bars. That is, do so takes V' antecedents:

  1. John will buy the book on Tuesday and Ringo will do so on Wednesday.
  2. * John will put the book on the table and Ringo will do so on the chair.
Order facts

As we expect from our X-bar rules, complements must precede adjuncts

  1. He worked physics at the office.
  2. * He worked at the office at physics.
  3. He laughed at the clown at ten o clock.
  4. * He laughed at ten o clock at the clown.
(not semantic)

Semantics may decide what adjuncts go with what heads. But it isn't enough to decide what complements go with what heads. We need to allow for arbitrary syntactic selection.

  1. He asked the man next door what time it was.
  2. * He asked of the man next door what time it was.
  3. * He inquired the man next door whether he was leaving.
  4. He inquired of the man next door whether he was leaving.
  5. He desired justice.
  6. * He desired for justice.
  7. * He longed justice.
  8. He longed for justice.
  9. Mary's desire for justice was well-known.
Note that the verb desire requires an NP complement. But this doesn't seem to be a semantic fact because the semantically closely related verb long and noun desire both require PP complements with for. So if you have a modifier that is acceptable with one head but not another semantically close head, that is evidence that the modifier is a complement.

Adjuncts are never obligatory.

  1. John treated Mary badly.
  2. John treated Mary. (different meaning)
To keep the interpretretaion of treat meaning behaved towards we need to include an adverb like well, or badly. This is evidence that badly is a complement of treat.
Ellipsis Verb phrase Ellipsis ellides V-bars or V-double bars.
  1. Who might be going to the cinema when?
  2. John might be on Tuesday.
  3. Who will put the book where?
  4. * John will on the table.
  5. Who might have gone to the cinema when?
  6. John might have last night.
(4) is an attempt to ellide a non-constituent put the book, which is only a part of a V', and is therefore ungrammatical.

Let's assume emphatic reflexives like this:

  • John baked the cake himself.
are adjuncts

Does this predict the following facts?

  1. John will bake the cake himself for the party.
  2. John will bake the cake for the party himself.
  3. John will put the candles on the cake himself.
  4. *John will put the candles himself on the cake.
Answer: Yes. Why?

Structure of AP

[AP> (D) A (Complement) Adjunct*]
  1. It's hard to believe he's that angry.
  2. It's hard to believe he's so angry.
  3. She was very angry.
  4. She was much kinder to John.
  5. He was too tall.
  6. How tall was she?
  7. She was as tall as John.
  8. She was a foot taller than John.
  9. She was a foot taller than John.
  10. He was more impressive close up.
  11. She acted quite independently.
Recall that the category A includes both adverbs and adjectives. There are specifiers of both
Complements We assume complements are introduced by the following PS rules:
    A' => A (PP)
    A' => A (S)
    A' => A (V')
Crucially, complements are sisters of A, not A'. Here are some examples:
  1. He is fond of Mary.
  2. He is proud of Mary.
  3. She is angry at John.
  4. She was kind to John.
  5. She is happy that he would show up.
  6. She is eager to run marathons.
  7. She made up her mind independently of me.
Note that adverbs, too, may have complements.
The A'' constituent

The rules above embody the assumption that there is a constituent A'' that includes a head adjective and ALL its modifiers, complements, heads, and adjuncts alike. A key argument for this constituent:

  • Preposing
      Fond of Mary though he is in some ways
The A' constituent

The rules above embody the assumption that there is a constituent of A'' which excludes the determiner and includes both the the head and its complement and adjunct modifiers. We call this constituent A'.

Some arguments motivating this constituent:

  1. Anaphora
    1. John is very fond of Mary in some ways, but he is less so in other ways.D[so = [A' very fond of Mary])
    2. John used to be very fond of Mary in some ways but he is rapidly becoming less so. (so = [A' very fond of Mary in some ways])
  2. Coordination
    • John is very rich and thin.
    • The pole is very long and thin.
    • John is very proud of Mary and beholden to her.

There are two types of adjuncts

  1. Attribute adjuncts
      She utterly foolish.

      She severely critical of the president

  2. Postadjectival adjuncts
      He is fond of Mary in some ways

We assume these are introduced by the following rules:

  1. A' => ADVP A' [attributive adjunct]
  2. A' => A' ADVP
  3. A' => A' PP
Crucially all adjuncts are sisters of A' and are introduced by recursive rules.

Arguments distinguishing A Complements from A adjuncts

Obligatoriness Obligatory modifers, we assume, are complements. Why this is natural will be clearer later when we flesh out our theory of the lexicon.

Complements conjoin with complements, adjuncts with adjuncts, and the twain shall never meet. This follows if we assume a conjunction rule something like:

    PP => PP and PP

The data:

  1. John is fond of Mary and of Sue.
  2. John was utterly and completely foolish.
  3. *John is fond of Mary and in some ways.
< Note that this test has to be taken with a grain of salt, beacuse there are strong constraints even on the coordination of adjuncts:
    *She was fond of Mary in some ways

The rules introducing complements and adjunct predict complements do not iterate and adjuncts do: K

    She was severely directly critical of the president
    *She was proud of Mary of John

The rules introducing complements and adjuncts predict that complements must come closer to the head adjective:

    She is fond of Mary in some ways
    ? She is fond in some ways of Mary
Proform We assume so can take only A's (not As) as its antecednt.
  1. John is very fond of Mary in some ways, but he is less so in other ways.D[so = [A' very fond of Mary])
  2. John used to be very fond of Mary in some ways but he is rapidly becoming less so. (so = [A' very fond of Mary in some ways])
  3. * Jean is very fond of Mary but she is less so of John. (so = [A fond], which contradicts our assumption that the antecedent of so is always an A')
If your grammaticality judgements do not in fact make the last example ungrammatical, then so anaphora does not provide a testr distinguishing A complements from A adjuncts (although it still gives us evidence motivating the A' constituent). m We must also revise our description of so as a Pro-A.
    so. takes either A' or As as its antecedents.

Arguments for specifier vs. adjunct analysis

Note that a lot of things we've called A' specifiers are sometimes called adverbs:

But so are many of the things we've called attributive adjuncts. How do we tell them apart?

In general specifiers according to our rules can't iterate:

In some cases they may appear to:

But in this cases, we assume the first very is a specifier of the second very. The second very, then, is a specified specifier.

But this makes iterability hard to use as negative test. If something iterates to the left of a head, we may always call it a specified specifier. Still, if something fails to iterate, we've got good evidence it should be a specifier:

A weird fact about many A' specifiers is that many of them cause a funny kind of preposing of AP out of NPs:

  1. They chose so large a milkshake that they couldn't finish it
  2. I couldn't believe they chose that large a milkshake.
  3. How large a milkshake do you want?
Instead of occuring after the determiner as attributive APs usually do these, these APs occur before it. Not all do: But it's a good positive indicator. Preposing out of NP indicates specifier-hood.

P'' and its specifiers, complements and adjuncts

Our PP modifers will look a lot like our modifiers for other categories.

We assume phrase structure rules analogous to those above:
  1. P'' => (D) P'
  2. P' => ADVP P' [attributive adjunct]
  3. P' => P' ADVP
  4. P' => P' PP
  5. P' => P S
  6. P' => P NP
  7. P' => P PP
  8. P' => P

Examples that Radford discusses (with some fleshing out):

  1. Complements
    1. NP prepositional objects
      1. during the war
      2. before the war
    2. Sentential complements (subordinating conjunctions as prepositions)
      1. before the war ended
      2. when Sam arrived
    3. prepositional phrase complements
      • out from under the table
      • out of toilet paper
      • away from Boston
  2. Specifiers
    1. right under the textbook
    2. less at odds with his friends
    3. so completely in the wrong [these example discussed further below]
    4. two years before the war
    5. The bodyguards stood quite close behind him.
    6. The rabbit burrowed quite deep under the surface
  3. Adjuncts
    1. prepositional adjuncts
      1. at odds with his friends
      2. out of touch in some ways
      3. the man in the park on Tuesday
    2. attribute adjuncts
      1. completely at odds with his friends
      2. completely in the wrong
      3. partly out of the drawer

There is another analysis available for

So could be a specifier of completely rather than of the preposition. Our rules allow both and both seem semantically possible.[Draw the alternative tree, Radford's is on p. 77] Radford notices this systematic spoecifier ambiguity in earlier chapters, but doesn't fully address the problem.

Notice we need so as a specifier of completely independently:

Notice that both analyses account for the word order facts Radford cites.

Arguments that something is a P modifier

PP modifiers often occur in positions where they could be interpeted as modifying something else, for example whatever the PP is modifying. In such cases, we need ways to argue that a modifier really is a modifier for the PP, and not something else. For example, consider in some ways in

How do we know that in some ways isn't modifying the copula is instead of the preposition? Or in How do we know on Tuesday isn't modifying the noun man? Here's an argument.

Call this a semantic dependency argument: The modifier's presence depends on the preposition's presence:

  1. He's out of touch in some ways
  2. ? He is in some ways.
  3. The man in the park on Tuesday was a policeman.
  4. ? The man on Tuesday was a policeman.

Arguments for complement vs. adjunct analysis for Prepositions

Obligatoriness Obligatory modifers are complements.
  1. Lay the strip along the side of the drawer.
  2. * Lay the strip along.
  3. They were always at odds with their friends.
  4. They were always at odds

The rules introducing complements and adjunct predict complements do not iterate and adjuncts do:

    He was at odds with his friends in some ways
    * He was at odds at loggerheads.

The rules introducing complements and adjuncts predict that complements must come closer to the head adjective:

    He was at odds with his friends in some ways.
    He was at odds in some ways with his friends. He was so out of touch in some ways. * He was so out in some ways of touch.
Proform We assume so can take only P's (not Ps) as its antecedent.
  1. I know that he's at odds with his colleagues, but he's less so with his friends.(so = [P'at odds])
  2. * I know that he's at odds with John, but he's less so odds with Mary.(so = [Pat])
This time the judgements on the so test seem much more robust. Interesting.

Arguments for specifier versus head

  1. The dispute dates from before the war.[HEAD]
  2. I've put your book over in the corner.[SPECIFIER]
  • * The dispute dates before the war.[from omitted]
  • Put your books in the corner.[over omitted]
We've seen that specifiers can be optional. But heads determine the distributional properties of their phrases. They are not optional.

Note that using obligatoriness to argue that something is not a specifier has to be done with caution. We argued in class that the italicized constituent was a specifier:

  • He put the book six inches from the end of the table
  • He put the book from the end of the table.
Note that in this context the specifier is obligatory. Sometiems specifiers are obligatory. We saw this in the case of NP specifiers:
  • The man was awake.
  • * Man was awake.

We see that the same verb date syntactically selects a PP headed by from.

    The dispute dates from the time of the war.
This motivates calling the head of from before the war from as well in the previous example.

Over takes complements of its own, Radford argues, as in:

  • I've put your books over there in the corner.
If there is a complement of over, then in the corner can't also be.

This argument has several flaws. First, we know heads can take two complements,as give does. Second, the so-called head in the previous example is Optional:

  • I've put your books there in the corner.

What this suggests is that over in over there is a specifier of there, which is why it is optional.

General Issues

So far we allow recursive modifiers to attach only to X' constituents.

Radford discusses the possibility that there are recursive modifiers of X and of X''. He suggests words like even as an N'' modifier and enough as an A modifier and our famous verb particles as V modifiers:

A disappointment for this analysis is that none of these is recursive:

Rule Constraints

We continue to write PS-rules as we have, reemembering that our category names are really abbreviations for sets of features.

We formalize X-bar theory as a set of constraints on Rules. We call the intervening bar-levels between lexical X and the highest level X projections of X. We call the highest-level projection the maximal projection of X.

  1. Endocentricity constraint(EC): The output of a rule must contain a category of the same category and either of the same bar level or lower.
  2. Modifier maximality constraint(MMC): All the nonheads in a rule must be maximal projexctions (X-double bar categories:Stowell).

Radford observes that the MMC requires us to revise:

to be

Which seems right:

Eliminating categorial rules

The gist of this is that we move to very general rules like this complement rule

We now allow any kind of complement and any number of them with any category head, overgenearting wildly.

We make the same move for all the rules, leaving ourselves with the problem of explaining why the rules overgenerate, for example, accounting for why adjectives, adverbs, and nouns never take NP complements.