Extending the Applicability of Lexical Mapping Theory

LFG grants syntactic functions a central role and has developed a theory of argument structure, Lexical Mapping Theory (LMT), which is independent of phrase-structure trees and thus able to account for morpholexical derivations. Yet some fundamental phenomena falling within the scope of morpholexical analysis – such as morphosemantic (meaning-altering) operations, phenomena referred to elsewhere as ‘demotions’, or subjectlessness – are currently denied satisfactory LMT accounts. This paper offers a way of extending LMT to phenomena which are awkward or impossible to handle with the current widely accepted versions of LMT. While retaining the main component of LMT – the feature decomposition of syntactic functions – I suggest the following set of revisions: (1) restoring the early LFG distinction between argument positions and semantic roles; (2) allowing the semantic participants to change order and re-associate with different argument positions for non-default (morphosemantically altered) mappings; (3) fixing the order of (syntactic) argument positions; (4) reformulating the principles of argument-to-function mapping to make fuller use of the markedness hierarchy of syntactic functions and render the Subject Condition redundant; and (5) using a mechanism of increasing markedness to account for morphosyntactic operations referred to as ‘demotions’ in RG. I demonstrate that these revisions make LMT a cleaner formalism which is immediately applicable to some important phenomena that have so far escaped (good) analyses. 1 Revision 1: Restoring the early LFG distinction between argument positions and semantic roles LFG’s argument structure is the locus of the mapping between semantic roles and grammatical functions. Because it maps from some kind of semantic or conceptual representation to a syntactic representation of grammatical functions, it is widely accepted that argument structure is a representation of the syntactic arguments of a predicate and that it contains some amount of semantic information, even though researchers still do not agree on how much. See Dalrymple (2001:197-200) for an overview of two major approaches to the content and representation of argument structure within LFG: Jackendovian and Dowtyian; and Butt (2006:Chapter 5) for a critical account of Jackendoff’s and Dowty’s linking theories and the way they have been combined with LMT. Although the discussion of the semantic component of LMT has concentrated on the source and classification of the semantic content ascribed to the arguments (drawing from the Conceptual Semantics framework of Jackendoff 1983, 1990; or 1 I gratefully acknowledge the current ESRC grant RES-051-27-0211. I also wish to thank Miriam Butt and Yehuda Falk for their very helpful comments and questions on this paper, some of which remain to be addressed in further work. the Proto-Role classification of Dowty 1991), another relevant issue concerns the degree of association (i.e. either fusion or separation) of the semantic information and the syntactic argument positions. It is the second issue which falls under the scope of the proposed first revision. Early LFG representations of argument structure implied a dissociation of argument positions and semantic roles, for example (Bresnan 1982:6): (1) (SUBJ) (OBJ) | | ← lexical assignment of grammatical functions ‘LOVE ( arg 1 , arg 2 )’ ← predicate argument structure