X-bar theory

For other "bar" connected with formal languages see Bar (computer science)

X-bar theory is a component of linguistic theories of Chomskian geaneaology which attempts to identify syntactic features presumably common to all those human languages that fit in a presupposed (1965) framework. It claims that among their phrasal categories, all those languages share certain structural similarities, including one known as the "X-bar", which does not appear in traditional phrase structure rules for English or other natural languages. X-bar theory was first proposed by Noam Chomsky (1970) and further developed by Ray Jackendoff (1977). An X-bar theoretic understanding of sentence structure is possible in a constituency-based grammar (= phrase structure grammar) only; it is not possible in a dependency-based grammar (= dependency grammar).

The letter X is used to signify an arbitrary lexical category (part of speech); when analyzing a specific utterance, specific categories are assigned. Thus, the X may become an N for noun, a V for verb, an A for adjective, or a P for preposition.

The term X-bar is derived from the notation representing this new structure. Certain structures are represented by X (an X with an overbar). Because this is difficult to typeset, this is often written as X′, using the prime symbol. In English, however, this is still read as "X bar". The notation XP stands for X Phrase, and is equivalent to X-bar-bar (X with a double overbar), written X″, usually read aloud as X double bar.

Core concepts

There are three "syntax assembly" rules which form the basis of X-bar theory. These rules can be expressed in English, as immediate dominance rules for natural language (useful for example for programmers in the field of NLP—natural language processing), or visually as parse trees. All three representations are presented below.

1. An X Phrase consists of an optional specifier and an X-bar, in any order:

XP → (specifier), X′

     XP                XP
    /  \      or      /  \
spec    X'           X̄  spec

2. One kind of X-bar consists of an X-bar and an adjunct, in either order:

(X′ → X′, adjunct)

Not all XPs contain X′s with adjuncts, so this rewrite rule is "optional".

   X'                   X'
  / \        or        / \ 
X'   adjunct    adjunct   X'

3. Another kind of X-bar consists of an X (the head of the phrase) and any number of complements (possibly zero), in any order:

X′ → X, (complement...)

  X'                                X'
 / \             or                / \  
X   complement           complement   X  

(a head-first and a head-final example showing one complement)

How the rules combine

The following diagram illustrates one way the rules might be combined to form a generic XP structure. Because the rules are recursive, there is an infinite number of possible structures that could be generated, including smaller trees that omit optional parts, structures with multiple complements, and additional layers of XPs and X′s of various types.

    XP
   / \
spec  X'
     / \
    X'  adjunct
   / \
  X   complement
  |
head

Because all of the rules allow combination in any order, the left-right position of the branches at any point may be reversed from what is shown in the example. However, in any given language, usually only one handedness for each rule is observed. The above example maps naturally onto the left-to-right phrase order used in English.

Note that a complement-containing X' may be distinguished from an adjunct-containing X' by the fact that the complement has an X (head) as a sibling, whereas an adjunct has X-bar as a sibling.

A simple noun phrase

The noun phrase "the cat" might be rendered like this:

     NP
    /  \
  Det  N'
   |   |
  the  N
       |
      cat
 

The word the is a determiner (specifically an article), which at first was believed to be a type of specifier for nouns. The head is the determiner (D) which projects into a determiner phrase (DP or DetP). The word cat is the noun phrase (NP) which acts as the complement of the determiner phrase. More recently, it has been suggested that D is the head of the noun phrase.

Note that branches with empty specifiers, adjuncts, complements, and heads are often omitted, to reduce visual clutter. The DetP and NP above have no adjuncts or complements, so they end up being very linear.

In English, specifiers precede the X-bar that contains the head. Thus, determiners always precede their nouns if they are in the same noun phrase. Other languages use different orders. See word order.

A full sentence

For more complex utterances, different theories of grammar assign X-bar theory elements to phrase types in different ways. Consider the sentence He studies linguistics at the university. A transformational grammar theory might parse this sentence as the following diagram shows:

The "IP" is an inflectional phrase. Its specifier is the noun phrase (NP) which acts as the subject of the sentence. The complement of the IP is the predicate of the sentence, a verb phrase (VP). There is no word in the sentence which explicitly acts as the head of the inflectional phrase, but this slot is usually considered to contain the unspoken "present tense" implied by the tense marker on the verb "studies".

A head-driven phrase structure grammar might parse this sentence differently. In this theory, the sentence is modeled as a verb phrase (VP). The noun phrase (NP) that is the subject of the sentence is located in the specifier of the verb phrase. The predicate parses the same way in both theories.

Substitution test

Though X-bar clauses may seem arbitrary and unneeded, their existence can be confirmed by substitution. To the above sentence, "He studies linguistics at the university," someone could reply, "Oh, she does, too." The word "does," here, stands for the entire V-bar phrase, "studies linguistics at the university", thus implying the existence of this phrase as a complete unit of the whole sentence. In other words: if the V-bar phrase above were not defined as such, the sentence would have three separate phrases directly underneath S: the Noun phrase, the Verb phrase, and the Prepositional phrase. To substitute for two of them, together, as shown, implies that these two, together, make up one phrase within the sentence.

Reduction

In 1981, Tim Stowell tried to derive X-bar theory from more general principles in his MIT thesis Origins of phrase structure, a pathbreaking but ultimately unsuccessful effort according to Andras Kornai and Geoffrey Pullum. In the mid 1990s, there were two major attempts to deduce versions of X-bar theory from independent principles. Richard Kayne's theory of Antisymmetry derived X-bar theory from the assumption that there was a tight relation between structure and linear order. Noam Chomsky's paper Bare Phrase Structure attempted to eliminate labelling (i.e. bar-levels) from syntax and deduce their effects from other principles of the grammar.

Quantity of sentence structure

Theories of syntax that build on the X-bar schema tend to posit a large amount of sentence structure. The constituency-based, binary branching structures of the X-bar schema increase the number of nodes in the parse tree to the upper limits of what is possible. The result is highly layered trees (= "tall" trees) that acknowledge as many syntactic constituents as possible. The number of potential discontinuities increases, which increases the role of movement up the tree (in a derivational theory, e.g. Government and Binding Theory) or feature passing up and down the tree (in a representational theory, e.g. Lexical Functional Grammar). The analysis of phenomena such as inversion and shifting becomes more complex because these phenomena will necessarily involve discontinuities and thus necessitate movement or feature passing. Whether the large amount of sentence structure associated with X-bar schemata is necessary or beneficial is a matter of debate.

Endocentric structures only

When the X-bar schema was introduced and generally adopted into generative grammar in the 1970s, it was replacing a view of syntax that allowed for exocentric structures with one that views all sentence structure as endocentric. In other words, all phrasal units necessarily have a head in the X-bar schema, unlike the traditional binary division of the sentence (S) into a subject noun phrase (NP) and a predicate verb phrase (VP) (S → NP + VP), which was an exocentric division. In this regard, the X-bar schema was taking generative grammar one step toward a dependency-based theory of syntax, since dependency-based structures are incapable of acknowledging exocentric divisions. At the same time, the X-bar schema was taking generative grammar two steps away from a dependency-based understanding of syntactic structure insofar as it was allowing for an explosion in the amount of syntactic structure that the theory can posit. Dependency-based structures, in contrast, necessarily restrict the amount of sentence structure to an absolute minimum.

See also

• Bare Phrase Structure
• Linguistics
• Minimalist syntax
• Phrase structure rules
• Syntax
• Sentence diagram

External links

• http://www.ucalgary.ca/~mcginnis/301/f04/Xbar.pdf
• http://www.ling.upenn.edu/~beatrice/syntax-textbook/index.html
• http://kornai.com/Papers/xbarver.pdf