Linguistics 522

Background Lecture

Sample Trees from last homework assignment

    Chapter 6 Tree

    Chapter 7 Tree

XBar Theory: Complements and adjuncts

We assume the following PS-rules (phrase-structure rules):

  1. NP => (D) N'
  2. N' => N' PP (Adjunct rule)
  3. N' => N PP (Complement rule)
Read N' as N-bar.

The first thing these rules do is claim that there is a constituent intermediate between an NP and lexical N. This will be a constituent containing the head noun and its modifiers, the italicized sequences in the following NPs.

  1. the king of France
  2. the yellow bird
  3. the letter carrier
  4. the unattractive hippotamus in the living room
So far we only have rules for PP modifiers. Rules for other types of modifiers will be given below.

The other thing the 3 PS-rules do is distinguish between three kinds of PP modifiers, Specifier, complements and adjuncts. Here are some examples to motivate the distinction:

  1. specifier: the student
  2. complement: the student of physics

      Figure 1
  3. adjunct: the student with red hair

      Figure 2
There is a semantic intuition here which is not always clear but can roughly described as follows. The adjunct and the noun express two distinct properties of the individual described. It can be paraphrased in two separate clauses: The complement and the noun function together to express one property, which can be expressed in one clause, and not in two: This semantic intuition is not very solid and the more you think about it, the slipperier it gets. We are going to rely on some syntactic tests to back the intuition up.

In determing whether a modifier is a complement or adjunct remember the following:

  1. When asked to produce an argument that something is a complement or an adjunct, you are being asked to apply some of the tests below, not to report your intuition about whether one property or two is being attributed.
  2. It is also bad form, when asked to produce an argument that something is a complement or an adjunct, to draw a tree like the tree in Figure 1 and say "of physics is a complement because it is sister to the lexical head N". The tree in Figure 1 is a technical way to notate the claim that something is a complement. It is not an argument. You could also draw the tree

      Figure 3
    And this would make the competing claim that of physics is an adjunct. When you are asked to produce arguments that something is a complement or an adjunct, you are being asked to decide WHICH tree is the right tree, and then produce evidence that it is the right tree. Drawing the tree is not one of the arguments; it's what you're arguing for.
  3. Note: the notions complement and adjunct are both relational. A modifier is a complement of some head. The first decision to make in arguing that something is a complement is to decide what head it is the complement of. Some of the tests below only work when the head is of a particular category. For example, one- replacement is only relevant when the head is a noun. do so replacement only works when the head is a verb. Preposing only works when the head is a verb.
  4. The tests you use and the examples you produce must be discriminating. They should produce one result when applied to a complement and a distinct result when applied to an adjunct. Make sure you know which result correlates with which outcome and say that in your explanation of the test result. You can be compact, present 3 example sentences, and say "All these results are evidence that X is a complement," but if the results are split, you MUST say which result points to complementhood, which to aduncthood.
  5. Use minimal pairs whenever possible. For example, when applying the reordering test below, show both orders, thus demonstrating that any ungrammatical sentence you produce has no problems independent of the reordering. When using the obligatoriness test, always give one example with the modifier claimed to be obligatory, and one without. And so on.
  6. The property you are determining in performing adjunct/complement tests is a lexical property of the head word. This will become more clear in Chapter 8. That means it is a property of the head word when it has a particular meaning. In constructing examples you must never change the head word and you must never change the head word's meaning. And the syntactic category, and semantic relation of the modifier to the head word should remain fixed. As a consequence, if the modifier is a prepositional phrase, the preposition should not be changed, because that almost always affects the semantic relation. The purpose of the initial example, then, is to fix the head word, the meaning, the semantic relation, and the category of the complement. Almost everything else can be changed (and should be, in the service of producing a good example).
    1. Initial example: I had an argument with a colleague.
    2. Head word: argument Modifier: with a colleague
    3. Relevant example: I had forgotten about our argument with the IRS. (same head, semantic relation)
    4. Irrelevant example: I had an argument about a colleague. (different semantic relation)
    5. Irrelevant example: I gave her an argument with an agenda. (different semantic relation. The agenda is something the argument possesses, the colleague is your opponent in the argument. The agenda is not the opponent in the argument and the argument does not possess a colleague.)

We introduce a set of tests for the complement/adjunct distinction. As a set of properties that all converge on making the same distinction they constitute an argument for it. Some of them are also simple predictions that follow directly from making this structural distinction.
Order of
and adjuncts

If you look closely at the complement and adjunct rules you will see that in any tree containing both a complement and an adjunct, the complement must always occur closer to the noun. This predicts the following facts:

  • a student of physics with red hair
  • * student with red hair of physics
Consider another example.
  • John's desire for justice
Let's test whether in Switzerland is a complement or an adjunct, given the little we know so far.
  • John's desire for justice in Switzerland
  • ? John's desire in Switzerland for justice
The second NP is certainly more marked, so we have some evidence for the claim that for justice is a complement.
Iterability of

The adjunct rule has the interesting property that it can be applied any number of times. It has an N' as both the mother and the daughter.

  • the boy with red hair
  • the boy with red hair with long arms
  • the boy with long arms with red hair
  • the boy with red hair with long arms in the den
In contrast, the Complement rule can only be applied once:
  • * the student of physics of astronomy

Note: Do not confuse this with the claim that there can only be one complement. This is wrong:

  1. John handed Mary the ball.
  2. * John handed Mary.
  3. * John handed the ball.
There are clearly verbs that take two obligatory complements, which must both be sisters to the lexical head:
This is not two applications of the complement rule. It is one application of a single rule that allows two complements:
    V' -> V NP NP
But only some heads, like hand, are eligible for this rule. The noun student is clearly not one of them.

Note: The following test involves replacing an Nbar with one. It is only relevant when the head is a noun.

One is an anaphoric element that seems to be able to take an entire N-bar as its antecedent.

  • Bill has a yellow car with green racing stripes. Sue wants a blue one. (one = car with green racing stripes)

We see the following contrast

  • The student with red hair was smarter than the one with blue hair.
  • * The student of astronomy was smarter than the one of biology.
Do-so replacement

This works just like one- replacement only it is for use when the head is a V instead of an N.

do so replace V's, just as one replaces N's,

  1. * John relied on Mary, and Fred did so on Sue.
  2. John danced on the stairs and Sue did so on the stage,
The italicized words are what do so replaces in each sentence, and the fact that it replaces dance successfully is evidence that dance is a V', and therefore that on the stairs is an adjunct. Meanwhile the fact that do so cannot replace relied successfully is evidence that relied is not a V', and therefore that on Mary is an adjunct.

We can coordinate adjuncts with adjuncts and complements with complements:

  • The student with red hair and with green eyes was quite striking.
  • A student of astronomy and of physics must be smart.
But we cannot mix.
  • * The student with red hair and of physics
Something bogus about this test? Yes, many things can go wrong. But you are allowed to use it. Notice, coordination of likes with likes is grammatical for both complements and adjuncts. So the test only tells us something when we get an ungrammatical result. And what that ungrammaticality tells us is that the two modifiers are unlike. We then need the further assumption that one of them is an adjunct to conclude the other is a complement.

So far we've talked about predictions made by our structural assumptions. Now we just point to a syntactic fact that seems to be sensistive to our complement/adjunct distinction, without being able to say exactly why it's sensitive.

  • John decided on the boat.
  • On the boat, John decided.
Notice the first example is ambiguous. It either means (a) that John's decision took place on the boat, or (b) that of two alternatives John was considering (say, a boat or a coat hanger), John chose the boat-alternative. The second example has only the (a) reading. We usually explain this by saying that on reading (b) the PP on the boat is a complement, and that verbs dont like having their complement PPs preposed. This is a well-accepted tets, but it only works reliably for verb heads and PP modifiers. Don't use it in other circumstances.

In the theory developed in Chapter 8 it will become clear that ONLY complements can be obligatory. There will simply be no way to say that an adjunct is obligatory.

Therefore it can be argued that all of the following modifiers are complements:

  1. John gazed at/upon Mary.
  2. * John gazed.
  3. The book belonged to Mary.
  4. * The book belonged.
  5. John often resorted to flattery.
  6. * John often resorted.
  7. The purser reported to the first mate. (means: has as immediate supervisor)
  8. The purser reported. (meaning change)
  9. John doted on Mary.
  10. * John doted.
  11. John relied on Mary.
  12. * John relied.
  13. John was familiar with plate tectonics.
  14. John was familiar. (meaning change)
  15. John worded the letter with great care.
  16. * John worded the letter.
This argument may be used for heads of any category (notice the adjectival head familiar), but obligatoriness most often arises with verbal heads.

Notice the implication is valid in only ONE direction. If something is optional:

  1. John ate beans.
  2. John ate.
We cannot say whether it is a complement or an adjunct.

Our textbook says that "all" PP complements are marked with of.

As the list of examples under Obligatoriness above shows, this is wrong.

Ignore this statement and never use it in an argument. On pain of excommunication.

I'm not sure what he was thinking. Perhaps he meant: Almost all PPs marked with of are complements. This is a lot closer to true. But notice the almost,

The ruling here: Dont use the identity of the preposition, of or otherwise, as an argument. It is an unreliable indicator at best.

A Complicated tree from the homework

7 k from 194 (Ch. 6)

    Tree 1

The same tree in the theory of Ch. 7

    Tree 1

Complements and adjuncts prenominally

prenominal np

We have the following order facts for some prenominal NP modifers.

  • the Cambridge physics student
  • * the physics Cambridge student
Again we see a distinction in order, and we see the same semantic relation that we saw filled by an of-phrase filled by a prenominal NP. That is, the following are paraphrases:
  • the physics student
  • the student of physics
A movement account is possibly right!

Other complement/adjunct tests agree, all pointing toward complementhood of Physics in physics student and the adjuncthood of Oxford in Oxford student

  • Iteration
    1. A Cambridge high quality middle class student
    2. * A physics astronomy student (cannot mean a physics and astronomy student)
  • Free order with other prenominal modifiers
    1. A high quality Cambridge middle class student
    2. * A high quality physics middle class student
  • Coordination
    • Several Cambridge and Oxford students
    • Several physics and astronomy students
    • * Several Cambridge and astronomy students
  • One-replacement
    • * The physics student was way smarter than the astronomy one.
    • The Cambridge student was way smarter than the Oxford one.