SOAS Research Online

Abstract

There are two thrusts to this dissertation, one backward-looking, one forward. The first is that HPSG is a better minimalist theory than is attainable for the Principles and Parameters framework, given its programmatic goals and its commitment to move alpha as the central explanatory mechanism. The second is that there should be no artificial dichotomy between the inherited features which participate in unbounded dependencies. In the introductory chapter, I point out certain fundamental problems relating to the assumption that grammaticality is determined via conditions on movement operations, and provide a sketch of the importance of structure-sharing in HPSG, a theory which makes no appeal to movement. In chapter one, I present the evidence for a second wh-question feature, and extend Johnson and Lappin's treatment to account for the important "Subjacency in Japanese" data. In chapter two, I indicate certain flaws in Johnson and Lappin's approach, related to the fact that their account is not "head-driven" in having inherited features amalgamated through selecting heads. I suggest that this may be rectified while preserving a unified account of inheritance if LOCAL is retired, allowing full synsem structure-sharing between fillers and gaps. In chapter three, I present a feature amalgamation principle, and a revised NONLOCAL feature principle in order to determine conditions on inheritance. I point out severe difficulties associated with the decision to abandon a unified treatment of NONLOCAL features. In chapter, four, I present a cross-linguistic treatment of data relating to wh-question sentences in which I demonstrate the advantages of employing alternative repositories for the amalgamation of wh-question features -- rather than a two-level approach in which a wh-feature is reentrant as a syntactic trigger. This offers a natural extension of the J&L account, including a unified explanation of pied-piping and the restricted wh-in situ option in English, and Japanese Subjacency. In Chapter Five, I review certain difficulties with the account, and suggest that it is necessary to have more generalized lexical binding of SLASH in order to successfully dispense with complementizers. I show that the main claims, that NONLOCAL features are subject to amalgamation by selecting heads, with inheritance via structure-sharing, are sound and provide the basis for further research.