In this chapter I would like to turn to the diachronic consequences of the theory outlined in the previous chapters. I intend to show that the approach to linking based on restrictiveness provides an elegant way to state generalizations about the relative likelihood of changes. The Optimality Hypothesis introduced in the last section of the last chapter defines what a local simplification would be: the lowering of the restrictiveness of linkers applying to a given argument. If change in linkers is local simplification, then we can, by way of the Optimality Hypothesis, identify decreases in the restrictiveness of linkers applying to arguments as the likely changes.

The organization of this chapter is as follows. First, I will consider some of the issues raised in some previous work on change in linking. Then, I will return to the topic with which I began Chapter 2, case alternations. Like much, if not all variation, the case alternations are changes in progress. In §6.3 I will lay out the predictions the restrictiveness-based approach to linking makes for linking change. Then I will show that the diachronic side of the case alternations is exactly what is expected on this view. I will then turn to a more general consideration of the consequences of the theory and provide an account of linking changes in Scandinavian in particular. Finally, I will consider the general character of changes in linking.

Previous work on linking change

Previous work in case change has raised three major issues that I would like to address in this chapter in the context of the restrictiveness-based theory of linking.

In this chapter I will provide an analysis of some central problems in Icelandic syntax within the theory developed so far. I would like to follow out one of the consequences of the theory in the last chapter, that of the possibility of double linking. If, as argued there, obliqueness and non-obliqueness (represented by [±obl]) is introduced to NPs through the linkers that link them, then the possibility of lexical case and word order doubly linking NPs is predicted. Just this double linking, I will argue, is the best way to explain the well-known combination of rich case and relatively strict word order in Modern Icelandic. In the next chapter, I will provide a diachronic explanation of the rise of Modern Icelandic double linking.

The essence of the analysis I propose is that lexical case in Icelandic does not carry the [+ obl] feature specification and that Icelandic is type 1. If the lexical cases are not oblique, they will be non-oblique by default, and so it follows that the lexically cased NPs will behave as terms. This will be true whether or not there is a word-order position to link them as well. I will argue that preverbal or pre-VP position is the only positional linker in Icelandic. Subjects, then, will typically be doubly linked, but objects will not. Finally, the approach advocated here will provide a predictive account of the interaction of case, passivization, and coordination.

Case, agreement, and word order

Icelandic was the first language considered in Chapter 2. There I provided an analysis of case in Modern Icelandic, and illustrated the basic workings of the theory using Icelandic data.

Preliminaries

The study of case has many facets, and the word “case” itself has come to mean many things. This study aims at a better understanding of the way case functions in syntax by providing a new theory of the syntactic functioning of case and other morphosyntactic devices.

Because the focus is on case as it functions in syntax rather than case as a morphological category, I will make certain assumptions which will figure prominently below. First, I will assume that morphological case and syntactic case do not always coincide. Morphological case will simply be the paradigms of affixes which may carry other information as well, e.g. gender and number. For example, the “dative” case as a morphological category is a set of endings. My primary concern is with syntactic case, and here the distinctions are not quite so visible. A syntactic case will be defined by its basic distribution and its interaction with other cases. For instance, if we observe that dative marks the goals of various verbs – what might be termed “indirect objects” – we have a situation very common in the world's languages: I(nom) gave you(dat) the book (acc). But quite often we then notice that the dative marks an NP that is an experiencer in a sentence based on a two-place predicate: I(dat) like the book (nom). At this point we have a choice. We can assert that the “ I ” NP in such a sentence is the “same” in some sense as the “you (dat)” in the sentence with the three-place predicate. The “sameness” might be identity of “grammatical relation,” for example indirect object (or 3 in Relational Grammar).

This chapter will explore some consequences of the theory introduced in the previous chapter. There each case was associated with certain syntactic features, implemented there as the value of the feature link, which determine that case's syntactic behavior. In this chapter, I will extend this treatment of case to cover languages of different types and, in the next, to languages that make extensive use of linkers other than morphological cases.

By “languages of different types” I mean languages whose cases show superficial differences from the system introduced in the last chapter, which had a nominative for arguments and an accusative as default. Some languages have a default nominative rather than the default accusative. And in some languages, the case of the object of a transitive verb is the same as the case of the subject of an intransitive verb, and a different case marks the subject of a transitive verb. The former case would be called “absolutive,” the latter “ergative.” This chapter extends the particular system of Chapter 2 to cover all these types. The main result of this chapter is that a generalization of the theory of the last chapter involving just two parameters correctly predicts four main types of two-case system. This four-way typology will lead to a range of predictions about the behavior of case in individual languages.

In the next two sections, I will generalize the theory of linking and case frames already formulated and spell out the predictions it makes. Then I will present examples of each of the major predicted case systems and how the predictions fare in each of the languages considered.

Introduction

The theory of grammar predicts that a class of ergative adjectives should exist alongside the established classes of ergative verbs (Perlmutter 1978a, b; Burzio 1981, 1986), and ergative nouns. Their existence is a consequence of the X-bar Theory of phrase structure within the more general Lexicalist Hypothesis of Chomsky (1970). X-bar Theory imposes the condition that all phrasal categories (VP, NP, AP, PP, QP, AdvP, IP, CP) have the same internal structure. And the Lexicalist Hypothesis demands that morphologically related verbs, nouns, and adjectives be represented in the lexicon as single, categorially unspecified, entries endowed with certain unique theta-marking and selectional properties.

One would thus expect that an adjective morphologically related to an ergative verb (e.g. morto “dead” related to morire “die”) would also be ergative, that is, it should have its subject generated in object position, under A′, just as the subject of the corresponding ergative verb is generated in the structural object position, under V′.

On the basis of various phenomena in Italian, I will argue that the ergative/unergative distinction does indeed extend to (superficially intransitive) adjectives, in spite of the fact that adjectives which are morphologically related to ergative (and passive) verbs are not themselves ergative. Indeed, one of the reasons why the existence of a class of ergative adjectives was not immediately recognized, and is occasionally even explicitly denied (Burzio 1986; Levin and Rappaport 1986; Stowell 1987) is the absence of this “predicted” class of ergative adjectives.

There are some peculiar restrictions on the occurrence of impersonal si in untensed and tensed clauses with specific time reference in Italian that have gone largely unnoticed in the literature and are not expected under the standard analysis of the construction. I will show that they are in fact simple consequences of very general principles, Theta Theory and Case Theory, under a finer analysis of impersonal si: one that provides, among other things, for two distinct uses of the morpheme, an argumental and a non-argumental one.

Although it introduces a further distinction in the class of si constructions, this analysis is in principle compatible with the program of unifying all uses of si (impersonal, middle, and reflexive). The unification approach and the one pursued here simply set themselves at different levels of abstraction. A question to which we return below.

After reviewing the fundamental restrictions on impersonal si in untensed clauses (section 1), we will suggest an analysis which accounts for them via principles that are already independently justified (section 2). In section 3, the differences in the interpretation of impersonal si in finite clauses with specific and generic time reference will be discussed and related to the same analysis in interaction with a more general theory of “arbitrary” (arb) interpretation. Section 4 contains a reexamination of so-called passive si, reinterpreted in part in the light of the preceding results on impersonal si. In section 5, some (provisional) comparative remarks are made on impersonal si constructions in other Romance languages.

The articles collected here, some of which have not been published before, deal with different aspects of the syntax of Italian, including properties of the structure of nominal phrases, impersonal si constructions, and the particular type of finite complements of perception verbs sometimes called “pseudo-relatives.”

In some of the articles the analysis of the Italian data may be seen to bear on more general, comparative and theoretical, questions. Often, independently established principles of the theory are applied to the intricacies of the data and argued to reveal unforeseen patterns, or solutions for certain long-standing anomalies (as in the study of relative clauses in chapter 2, or in the study of si constructions in chapter 4). In other cases, the facts of Italian are shown to offer a privileged access to certain theoretical conclusions, not easily reachable from the vantage point of other (well-studied) languages (as is the case with the notion of syntactic operator in chapter 3, and the expected existence of a class of ergative, or unaccusative, adjectives, in chapter 7).

Although written over a period of more than ten years, these essays are all cast in one or another version of the so-called Principles and Parameters theory, essentially stemming from the notion of parametrized principle in Chomsky's 1979 Pisa lectures (Chomsky 1981).

The revolutionary and heuristic values of this development in the generative program initiated by Chomsky over forty years ago have probably been greater than those of any other, previous or later, development.

A guiding assumption of much current work in the theory of grammar is that (most) apparent differences among languages are merely the superficial consequences of relatively few different choices open to languages at an abstract level, at the margins of a unitary core, invariant across languages. In comparing, in their essentials, the relativization systems of Italian, French, and English, we will aim at a fundamentally unified treatment of the three systems, in spite of the many overt differences existing among them. In particular, we will try to show that it is possible to assign to the three languages one and the same theoretical apparatus for relative clause constructions and to localize a number of significant differences among them in terms of the slightly different way the three languages utilize this apparatus – perhaps one or two parametric choices. The analysis is conducted more broadly within the framework of the Extended Standard Theory, specifically, within the version of it presented in Chomsky (1979a, 1981a) and recent related work.

We will begin by giving (section 1) a somewhat more detailed analysis of Italian relative constructions and will sketch, in so doing, what we take to be the “core” grammar of relative clause structures for this type of languages. The French and English systems will then be discussed (section 2) against the background of the general conclusions arrived at on the basis of Italian.