Limitations on Embedding in Coordinate Structures (original) (raw)

Chomsky (1959a) presented an algorithm for constructing a finite transducer that is strongly equivalent to a Chomsky-normal-form context-free grammar for all sentences generated by that grammar with up to any specified finite degree of center embedding. This article presents a new solution using a variety of COORDINATE GRAMMAR to assign nonembedding (paratactic) structures strongly equivalent to those assigned by an embedding grammar, which can in turn be directly computed by a finite transducer. It proposes that the bound on center embedding is really a consequence of a bound on alternation between right and left embedding, called here ZIGZAG EMBEDDING. Coordinate grammars can also be used to assign nonembedding structures equivalent to those with up to any specified finite degree of coordinate embedding (the occurrence of a coordinate structure as a member of a coordinate structure of the same type). It concludes that coordinate grammars or the finite transducers strongly equivalent to them are psychologically real, and that the existence of a finite bound on the degree of zigzag and coordinate embedding is a consequence of the increasing size and complexity of such grammars or transducers as the bound increases.

Minimalism in grammatical theorizing (Chomsky 1995) led to simpler linguistic devices and a better focalization of the core properties of the structure building engine: a lexicon and a free (recursive) phrase formation operation, dubbed Merge, are the basic components that serve in building syntactic structures. Here I suggest that by looking at the elementary restrictions that apply to Merge (i.e., selection and licensing of functional features), we could conclude that a re-orientation of the syntactic derivation (from bottom-up/right-left to top-down/left-right) is necessary to make the theory simpler, especially for long-distance (filler-gap) dependencies, and is also empirically more adequate. If the structure building operations would assemble lexical items in the order they are pronounced (Phillips 1996; Chesi 2004, 2012), on-line performance data could better fit the grammatical model, without resorting to external “performance factors.” The phase-based, top-down (and, as a consequence, left-right) Minimalist Grammar here discussed goes in this direction, ultimately showing how strong Islands (Huang 1982) and intervention effects (Gordon et al. 2001, Gordon et al. 2004) could be better explained in structural terms assuming this unconventional derivational direction.

Usage-based approaches to syntactic theory suggest that grammatical systems are best conceived of as abstractions from individual usage events, such that preferred structural choices in language performance are conventionalized into productive grammatical patterns, regularities and constraints. Consequently, morphosyntactic universals, as inferred from the comparison of synchronic grammatical descriptions, are predicted to be firmly rooted in language performance. The advent of rich and systematic corpus data for some languages opens up the possibility to substantiate this hypothesis on a sound empirical basis, and complex sentences provide a particularly wide range of constructions and phenomena amenable to such a research programme. The present paper surveys major previous contributions in this area and provides new data or perspectives on a variety of pertinent construction types, including relativization, purpose, avertive and complement clauses, as well as converbal constructions.

This paper addresses the issue of Split Intransitivity (si) and Unaccusative Mismatches (uMs), proposing a constraint-based approach to si and ums within a recent framework of Head-driven Phrase Structure Grammar. I argue against the widely accepted dichotomous distinction of intransitive verbs, which has been advanced by the Unaccusative Hypothesis [Perlmutter (1978)]. I then propose a quadripartitive distinction of intransitive verbs on the basis of the distribution of subject argument in the semantically motivated argument structure, and show that this quadripartitive distinction allows a better understanding of si and ums. The main idea of this proposal will be summarized as the Quadripartitive Split Intransitivity Hypothesis (Qsm).