However, there has been work in syntactic theory in which the lexicon has played a more prominent role, such as Bresnan's Lexical Function Theory [15]. In addition, there is Levin's [73] work on how semantic classes of verbs determine their syntactic distribution.
Lexical Functional Grammar (LFG) developed in reaction to the basic principles of transformational grammar regarding grammatical structure, specifically the use of transformations to ensure that the same functional relations apply at ``deep structure'' for sentences that are paraphrases at surface structure (e.g., active vs. passive). The main features of LFG follow: there is no one-to-one correspondence between surface structure and predicate argument structure; functional primitives are incorporated rather than reducing grammatical function to phrase structure; and, only lexical rules can affect function argument correspondences, not syntactic rules. LFG has two levels of syntactic description: a constituent structure (or c-structure), which is the same as a conventional phrase structure tree, and a functional structure (or f-structure), in which grammatical functions are explicitly encoded. The latter is the input to the semantic component and consists of attribute value pairs, possibly nested.
An important aspect of LFG is that transformations are in effect compiled into the lexicon. It illustrates that grammatical coverage is not sacrificed during the shift in emphasis from syntactic transformations to lexical specifications of predicate argument structure. However, the theory is still heavily syntax-oriented: since the predicate argument structures are only at the level of subcategorizations. Therefore LFG does not directly concern issues regarding finer sense distinctions, outside of predicate subcategorizations. For instance, little has been said regarding adjuncts, which are (verbal) complements not strictly subcategorized. It seems that the distinctions of interest are more likely to be found in relation to adjuncts since there is relatively little variety in the subcategorizations involving arguments alone. In this respect, the work by Jackendoff [61] on representing various adjunct types provides better insight into the relation between syntax and semantics, as discussed below.