Parallel Worlds of Syntax: Lexical-Functional Grammar as an Alternative to Chomskyan Theories

Lexical-Functional Grammar (LFG) stands as a major theoretical alternative to Chomskyan transformational grammars. Developed by Joan Bresnan and Ronald Kaplan, LFG is a non-derivational, constraint-based framework that rejects the core Chomskyan concepts of Deep Structure, Surface Structure, and syntactic movement. Its foundational principle is the parallel analysis of language on distinct levels of representation, primarily the constituent structure (c-structure), which models the sentence's tree-like form, and the functional structure (f-structure), which represents abstract grammatical relations like subject and object in an attribute-value matrix. This parallel architecture allows LFG to offer elegant, movement-free accounts of linguistic phenomena. For example, the active-passive alternation is not explained by moving a noun phrase from object to subject position, but rather by a lexical rule that alters the mapping of a verb's arguments to grammatical functions. This places significant explanatory power in the lexicon, which contains rich entries that drive the syntactic analysis. By separating grammatical function from syntactic position, LFG is exceptionally well-suited to analyzing the world's diverse languages, including those with free word order, and is often considered a more psychologically plausible model of human language processing.

1. Introduction: The Quest for a Model of Grammar

For much of the 20th century, the field of theoretical syntax was dominated by the generative enterprise initiated by Noam Chomsky. Theories like Transformational Grammar (TG) and its successors (Principles and Parameters, the Minimalist Program) proposed that the syntax of human language is a computational system involving derivations. Sentences begin as an abstract 'Deep Structure' and are transformed through rules of movement into the 'Surface Structure' we actually hear and speak. While incredibly influential, this derivational approach faced challenges, particularly concerning its psychological plausibility and its ability to elegantly model the diversity of the world's languages.

In the late 1970s, Joan Bresnan and Ronald Kaplan proposed a radical alternative: Lexical-Functional Grammar (LFG). LFG is a non-derivational, constraint-based framework that rejects the idea of syntactic movement. Instead, it posits that language is organized through multiple, parallel levels of representation that are related by principles of correspondence. It shifts the explanatory burden from complex transformational rules to a rich, highly structured lexicon.

2. The Core of LFG: Parallel Structures

The central innovation of LFG is its model of parallel architecture. Every sentence is analyzed simultaneously on at least two distinct but interconnected levels: the Constituent Structure (c-structure) and the Functional Structure (f-structure).

A. Constituent Structure (c-structure): The Shape of the Sentence

The c-structure is the most familiar component to students of syntax. It is a phrase structure tree that represents the hierarchical organization of words and phrases in a sentence. It captures constituency (which words group together) and linear order. These trees are generated by context-free phrase structure rules, similar to those in X-bar theory.

For the sentence "The girl reads a book," the c-structure would look something like this:

      S
     / \
    NP   VP
   / \   /  \
 DET  N  V    NP
  |   |  |   /  \
 The girl reads DET  N
               |    |
               a   book

This structure shows that "The girl" is a Noun Phrase (NP) and "reads a book" is a Verb Phrase (VP), but it doesn't explicitly state that "The girl" is the subject or that "a book" is the object. That information lives in the f-structure.

B. Functional Structure (f-structure): The Grammatical Roles

The f-structure is the heart of LFG. It is an attribute-value matrix (AVM) that encodes the abstract grammatical functions and features of a sentence, independent of their position in the tree. Attributes are grammatical functions (like SUBJ, OBJ, TENSE) and features (like PRED, NUM, PERS). Values are either other f-structures or atomic symbols.

For "The girl reads a book," the f-structure would be:

[
  PRED    'read<SUBJ, OBJ>'
  TENSE   PRESENT
  SUBJ    [ PRED   'girl'
            NUM    SG
            DEF    + ]
  OBJ     [ PRED   'book'
            NUM    SG
            DEF    - ]
]

This f-structure tells us that the predicate (PRED) is 'read', which requires a subject and an object. The subject's predicate is 'girl', and the object's predicate is 'book'. This level of representation is purely functional, abstracting away from the word order of the c-structure. This makes LFG exceptionally powerful for analyzing languages with free word order (non-configurational languages), where the subject might not always appear before the verb.

3. The Lexicon: The Engine of the Grammar

In Chomskyan grammar, the lexicon is often seen as a simple list of words with minimal features. In LFG, the lexicon is the primary engine of the grammar. Each lexical entry is rich with information that contributes to building both the c-structure and the f-structure. This is the "Lexical" part of Lexical-Functional Grammar.

A simplified lexical entry for the verb "reads" would specify:

1. Its c-structure category: V
2. Its f-structure information, including its predicate form: (↑ PRED) = 'read<SUBJ, OBJ>'
3. Equations that map its arguments in the tree to their grammatical functions. For example, it specifies that the NP preceding it in the VP maps to the subject ((↑ SUBJ) = ↓) and the NP following it maps to the object ((↑ OBJ) = ↓). The arrows (↑ and ↓) are formal devices for referring to the f-structure of the mother node (↑) and the current node (↓).

4. LFG in Action: The Active-Passive Alternation without Movement

A key test for any syntactic theory is how it handles phenomena like the active-passive alternation.

• Active: The dog chased the cat.
• Passive: The cat was chased by the dog.

Chomskyan (Transformational) Account: In a simplified transformational account, the passive sentence starts with an underlying structure similar to the active one. The NP "the cat" moves from the object position to the empty subject position to satisfy syntactic requirements (e.g., Case theory). This movement leaves a 'trace' in its original position.

LFG Account: LFG provides a more direct, movement-free explanation. The difference between active and passive is not a syntactic transformation but a lexical rule. The grammar contains a rule that takes an active verb's lexical entry and produces a passive participle entry. This rule changes the mapping of arguments to functions.

• Lexical entry for chased (active): Maps its first argument (agent) to SUBJ and its second argument (patient) to OBJ.
• Lexical entry for chased (passive): The lexical rule modifies this. The first argument (agent) is now optional and expressed in a prepositional phrase (OBL_AG, oblique agent), while the second argument (patient) is mapped to SUBJ.

F-structure for Passive: "The cat was chased by the dog"

[
  PRED      'chase< (OBL_AG), SUBJ >'
  TENSE     PAST
  VOICE     PASSIVE
  SUBJ      [ PRED   'cat' ]
  OBL_AG    [ PRED   'dog' ]
]

Notice that 'cat' is the SUBJ in this f-structure, directly reflecting its grammatical role in the passive sentence. There is no trace of movement because no movement occurred. This approach is considered more psychologically plausible, as it doesn't require the processor to mentally reconstruct a deep structure and track movement chains.

5. Well-Formedness and Conclusion

LFG ensures grammatical sentences through a set of universal well-formedness conditions on f-structures:

• Completeness: An f-structure is complete if it contains all the grammatical functions that its predicate requires.
• Coherence: An f-structure is coherent if all the grammatical functions it contains are required by its predicate.
• Uniqueness: Each attribute in an f-structure may have only one value.

In conclusion, LFG presents a powerful, formally precise, and psychologically grounded alternative to derivational grammars. By separating the form of a sentence (c-structure) from its grammatical relations (f-structure) and placing explanatory power in the lexicon, it offers elegant solutions to syntactic puzzles and provides a robust framework for computational applications and the analysis of diverse language typologies.

Browse public study notes