Linguistics 581

      Syntactic Structure

      What do we mean by Syntactic Structure?

        First Intuition: We can build up sentences in regular ways:

      A finite number of patterns licensing an infinite number of sentences.

        Canonical word order.

          English:
            John read the book.
            Subject Verb Object (SVO)
          Japanese: SOV
          • John-ga hon-o kai-masi-ta
          • John book bought.
          Irish: VSO

      These are the most common.

      Others. Free word order too.

      But there are no languages in which the canonical word order is SVO for some verbs, SOV for others.

      When you pick a pattern there are certain ways in which it has to generalize. Just makes sense? It would be confusing to have different verbs have different word order? But there's already so much to learn when you learn a verb: conjugation, conjugation group, what the verb means, what prepositions go with it, if any. Why not word order too?

      Maybe this constraint reveals something about the mind, or maybe something about how interpretation works. We leave that open.

      Structure = patterns [Rules]

      Language has different kinds of patterns [Rules]

      Syntactic structure = how the words of the language are put together into larger units (order of words, and organization of words into units (phrases))

      Syntax

      Ambiguous and non-ambiguous sentences:

      A Structural account:

      Why have structure intervene? Why not just go directly to meaning?

      Correlations: Readings correlate with certain syntactic patterns.

      Related sentences that have the same or similar readings plausibly have many of the same phrases.

      Why should this be? Phrases themselves have meanings. Syntax tells us how to put meanings together. Therefore different phrasal configurations can give us different meanings.

      1. The man walked his happy dog.
      2. The dog walked his happy man.
      3. I know a man who speaks French.
      4. A man I know speaks French.

      Principle tasks:

      Tests for constituency

      We call a phrase a constituent of the sentence it occurs in. This emphasizes that it's a relation concept. What may be a phrase in one context may not be in another

      More correlations. What can be replaced by a pronoun (it, that, he, him, she, her, they, them) is a phrase.

      Example One:

      1. He explained all the terrible problems he had encountered to her.
      2. He wrote down all the terrible problems he had encountered. Then he explained them to her.

      The second sentence in (2) (Then he explained them to her) is the result of substituting the pronoun them for the words all the terrible problems he had encountered in (1). Moreover, them refers to the same thing as all the terrible problems he had encountered, so that the second sentence in (2) means approximately the same thing as (1), demonstrating that, in the right context, the substitution can be made without changing the meaning. We conclude all the terrible problems he had encountered is a constituent of (1).

      Example Two:

      1. The agency sees widespread use of the codes as a solution.
      2. Widespread use of the codes needed to be promoted. The agency sees that as a solution.
      3. * Widespread use needed to be promoted. The agency sees that of the codes as a solution.

      Sentence (2) provides evidence that Widespread use of the codes is a constituent in (1), because it successfuly substitutes that for widespread use of the codes without changing the meaning of (1). By doing that, it also provides an argument that the PP as a solution is modifying the verb, not the noun codes or the noun use. The third sentence shows that widespread use is not a constituent of (1), and by doing that it shows that of the codes is NOT modifying the verb.

      Example Three:

      1. He explained all the terrible problems he had encountered to her.
      2. He explained to her all the terrible problems he had encountered.
      3. * He explained all the to her terrible problems he had encountered.

      The italicized words in (2) form a constituent because they can be moved from the position they occupied in (1) without changing the meaning, while the italicized words in (3) are not a constituent, because movement fails for them.

      More correlations. What can be MOVED is a phrase.

      Example Four

      Sometimes what can be omitted without changing the meaning is a phrase:

      1. John won't put the vodka in the drink, but Bill will put the vodka in the drink.
      2. John won't put the vodka in the drink, but Bill will. [omitted= put the vodka in the drink.] of or as an immediate part of a verb phrase: completely
      3. * John won't put the vodka in the drink, but Bill will put. [omitted= the vodka in the drink.]

      Context free grammars

      We'd like to have a precise statement of what kind of constituents are possible in the language, and what kinds can occur inside other kinds.

      Context free grammars are a good vehicle for capturing constituency.

      • s → np (aux) vp vp
      • s → vp
      • s → aux np vp
      • vp → v (np) pp*
      • np → (det) adj* n pp*
      • np → propername
      • np → pronoun
      • vp → is/are adj
      • pp → p np

      Good things:

      1. One set of rules for NP, whether in subject position or not.
      2. Lots of sentences we haven't seen before now "generated"

      But CFGs create as many problems as they solve. Consider the sentence

        The agency sees widespread use of the codes as a solution to the problem.

      Look here to see the trees the grammar above assigns to this sentence. There is only one right tree. Which is it?