Introduction

The issues discussed in this chapter lie at the intersection of two lines of inquiry. The first line of inquiry can be characterized as a search for theories of locality under which all or some of the locality conditions can be given a unified formulation. One such theory, presented in Manzini (1988, 1989), is partially summed up in section 2. A crucial feature of the theory is that sentential subjects are generated in a non-VP-external position. This allows for the empty category principle (ECP) and binding theory to be unified under the same definition of locality, one that does not mention the notion of subject.

The second strand of the discussion is related to recent theories of parametrization in binding theory. Section 2 briefly sums up the theory presented in Manzini & Wexler (1987) and in Wexler and Manzini (1987), according to which parametrization effects are due to the presence of a parameter, or multiple disjunction, in the definition of locality.

The central proposal in this chapter is that the subject-based definition of locality in Chomsky (1981) and the non-subject-based definition under which ECP and binding theory are unified are not mutually exclusive; rather, the two definitions represent coexisting values of the locality parameter. Anaphors such as English himself, but also Italian se stesso, etc., are associated with the subject-based value.

Natural language and universal grammar is the first of two volumes with the common subtitle Essays in linguistic theory (henceforth Essays). Volume II, entitled Semantics, subjectivity and localism, will be published shortly.

The title that I have given to Essays I is not as innocent as it might appear to be at first sight; and there are those of my colleagues who might see it as being almost wilfully provocative. One of my general aims, in bringing the various chapters of Essays together, is to demonstrate that the expressions ‘natural language’ and ‘universal grammar’ are often employed nowadays loosely and uncritically (if not equivocally) in the case of the former and tendentiously in the case of the latter. What is (or, in my view, ought to be) meant by linguists when they use the highly ambiguous phrase ‘natural language’ (either generically or non-generically) is a question that is directly addressed in Chapter 4 of Essays I. All the chapters in both volumes are concerned with the structure of what are normally, but imprecisely, referred to as natural languages. Almost all of them, also, are concerned with (so-called) natural languages within the framework of what would be referred to traditionally as universal grammar. But, as I explain in Chapter 7 of Essays I (and, in greater detail, in one or two of the chapters in Essays II), my view of universal grammar differs from the generativist or Chomskyan view, which currently holds sway in linguistic theory and is propagated (all too often uncritically) in many textbooks.

Introduction

The aim of this chapter is twofold: on one hand I propose an account for the distribution of third person pronouns, clause-bound and long-distance anaphors within PPs; on the other, to explain such distribution, I individuate thematic properties peculiar to prepositions, which differentiate them from other lexical heads such as adjectives and nouns. In this way, some phenomena concerning the behaviour of PPs in small clause constructions, and in predicative contexts in general, will be captured.

In section 2, I analyse the distribution of anaphors and pronouns. In section 3, I present the central hypothesis of this work, i.e. that the behaviour of such elements is related to the thematic properties of prepositions. In sections 4 and 5, I consider the distributions of PPs in small clauses and in predicative constructions. In section 6, I discuss some differences between locative and non-locative prepositions. Finally, I try to handle some potential counterexamples, suggesting that independent principles can account for certain apparent anomalies. This chapter is a portion of a larger work in which I will also consider cross-linguistic evidence (English, Dutch and French), which not only supports my view of the phenomena analysed here, but can also shed light on the peculiarities of the languages considered. In this chapter, for space reasons, I only examine data from Italian, briefly mentioning in a note the most salient aspects in which the other languages diverge.

Introduction

Part of the importance of linguistic theories, as well as other scientific theories, is that they tell you which facts are interesting, why they are interesting, and where to look for other interesting facts. The standard binding theory (BT) of Chomsky (1981) has been extremely important in this respect since it has sparked a great deal of cross-linguistic research into the binding properties of pronouns, reflexives, and other NPs, overt and non-overt. As a result, we have learned a lot about the nature of NPs in various languages, what their similarities are and how they differ, both within a given language and cross-linguistically.

It is well known that one does not have to look very hard or very far to find, say, reflexives that do not obey the same restrictions as reflexives in English. The socalled long-distance reflexives found in various languages are a case in point (see, e.g., Thráinsson (1976a, c), Napoli (1979), Yang (1983), Hellberg (1984), von Bremen (1984), Giorgi (1984), Maling (1984), Anderson (1986), Rögnvaldsson (1986), Everaert (1986a), Barnes (1984, 1986), Sigurðsson (1986a), Sportiche (1986), Sells (1987), Manzini & Wexler (1987), Wexler & Manzini (1987), Pica (1987), Koster (1987), Kuno (1987), and references cited there). The question is, however, what the existence of such reflexives implies for BT. There are various possibilities, and different suggestions can be found in the literature cited above.

This volume is a collection of original articles on the nature of the anaphoric systems in a variety of languages and from a number of different perspectives. The aim of the editors is to provide a new impetus to the study of long-distance anaphors, a phenomenon with ramifications that are rather puzzling from the perspective of the binding theory in its canonical form (as in Chomsky (1981)). All of the contributions are concerned with extending that theory in a manner that is as restrictive as possible. Some of the solutions solidly remain within the domain of the structural binding theory. In other cases, developing a restrictive theory required recognizing different components in binding, and assigning a specific role to pragmatic factors. Many of the articles converge, however, in important respects, leading to considerable simplification of the overall picture. The patterns found in the languages covered are so consistent that they cannot be accidental (although one would certainly wish to investigate a larger number of languages).

Most of the articles originated as contributions to a workshop on long-distance anaphora organized by the Department of Linguistics of Groningen University, 18–20 June 1987. The editors added a first chapter in which the main results of the volume are put together. The original plans were for this chapter to be written by the editors together with Tanya Reinhart, who participated in the workshop as a discussant.

American linguistics has proudly and more or less consciously adopted the pragmatic position; the philosophy of justification by results, of first getting things done and only then, if at all, asking what in fact has been done. In the preface to his collection of articles by American linguists, Martin Joos brings out this point well. He goes on to remark: ‘Altogether there is ample reason why both Americans and (for example) Europeans are likely on each side to consider the other side both irresponsible and arrogant. We may request the Europeans to try to regard the American style as a tradition comme une autre; but the Americans can't be expected to reciprocate: they are having too much fun to be bothered, and few of them are aware that either side has a tradition’ (Joos, 1957: vii). As a representative of one European tradition in the enviable position of having secured a captive American audience for an hour or so, I propose to put before you views that absorption in the fun might otherwise prevent you from considering. To those of you who, having heard these views, might feel inclined to say that they are ‘of only theoretical interest’ and that the linguist's job is to describe what actually occurs in particular languages without troubling himself about what might occur (for I have heard this said), I would suggest that the history of science is full of examples to support the opinion that the actual cannot be properly described, perhaps not even recognized, except in the framework of what has previously been envisaged as possible.

I will begin, if I may, with a quotation from one of my own works:

The above quotation comes from the editorial Introduction which I wrote for New Horizons in Linguistics 1 some twenty years ago (Lyons, 1970a: 8). When I came to rewrite this Introduction, more recently, for New Horizons in Linguistics 2, I noted that, even when the earlier volume was being written, “there were many linguists who felt that the principle of ‘autonomy’ had outlived its usefulness”; and that “this feeling is probably more widespread today than it was then” (Lyons, 1987b: 4–5).

What I am concerned with in this chapter is not language in the most general sense of the term ‘language’ but with what can be described more fully as natural human language. Arguably, this fuller description is redundant in respect of either or both of the two adjectives, ‘natural’ and ‘human’. Indeed, this is the view that most linguists and many philosophers of language would take. But it is worth making the point explicit and concentrating for a moment upon the implications of both of the qualifying adjectives, without prejudice to the question of whether there is any language, properly so called, that is non-natural or non-human.

Without dwelling upon the details let us say that a natural language is one that has not been specially constructed, whether for general or specific purposes, and is acquired by its users without special instruction as a normal part of the process of maturation and socialization. In terms of this rough-and-ready operational definition, there are some thousands of distinct natural human languages used in the world today, including English, Quechua, Dyirbal, Yoruba and Malayalam – to list just a few, each of which is representative, in various ways, of hundreds or thousands of others. But Esperanto, on the one hand, and first-order predicate calculus or computer languages such as algol, fortran and basic, on the other, are non-natural. Many non-natural languages are parasitic, to a greater or less extent, upon pre-existing natural languages.

Introduction

This chapter articulates the following hypothesis: the conditioning factor for whether an anaphor is ‘long-distance’ or ‘locally’ bound resides in a distinction between what will be called containment conditions and connectedness conditions on anaphors. The hypothesis is based mainly on properties of the anaphoric systems of Norwegian and Icelandic (sections 2.1–2.2), and the importance of the distinction is demonstrated also by interpretive facts concerning Norwegian constructions containing anaphors (section 3). Based on preliminary consideration of long-distance anaphors in certain other languages (section 2.3), our conjecture is that the hypothesis holds universally.

The distinction between containment and connectedness conditions can be seen in the perspective of what relation modules exist in grammar: connectedness conditions deal with what we may call the argument module, i.e. relations defined in terms of the relation ‘argument of’, a module extensively explored in recent work (e.g. Chomsky (1986a) and related research). Containment conditions, in contrast, involve what we will call the command module, a family of relations tied together by a general notion of ‘command’, to be elucidated as we proceed. The significance of containment conditions for anaphors brings to light the status of the command module as an independent and potentially significant module in the grammar.

The topic that I have chosen to talk about – the relationship between deixis and anaphora – turns out to be even more appropriate than I had hoped. Dr Bullowa, Dr Widdowson and Professor Halliday have all touched upon either deixis or anaphora, or both, in their contributions to the Symposium; and they have all made points to which I can refer in the development of the thesis that I am presenting.

I will argue that deixis is both ontogenetically and logically prior to anaphora. By this I mean that the deictic use of pronouns and other such expressions precedes their anaphoric use in the earliest stages of language acquisition and, furthermore, that anaphora, as a grammatical and semantic process, is inexplicable except in terms of its having originated in deixis. That deixis precedes, and is in some sense more basic than anaphora, is something that the previous speakers would probably concede immediately. But it is not at all uncommon for linguists to describe the meaning of pronouns, as far as possible, in terms of anaphora and to treat that part of the use of pronouns which is irreducibly deictic as a theoretical embarrassment that is best forgotten. It is certainly the case that generative grammarians have been inclined, until recently at least, to underestimate the role played by deixis in the interpretation of utterances; and the very term ‘pronominalization’, which figures so prominently in works on generative grammar, is loaded in favour of the view that pronouns are, first and foremost, substitutes for nouns (or nominals).

One of my aims in this chapter, which complements the preceding one, is to motivate a distinction between two terms that are currently employed by most linguists as synonyms and to use this terminological distinction as a peg upon which to hang some comments about the present state of linguistics. The terms in question are ‘linguistic theory’ and ‘theoretical linguistics’. Another aim is to comment further upon the theoretical term ‘language-system’ in relation to Saussure's terms ‘langue’ and ‘langage’.

The distinction between ‘linguistic theory’ and ‘theoretical linguistics’ is by no means the only terminological distinction that I shall be drawing, here and in other chapters of this book. I do not wish to give the impression, however, that my sole (or primary) concern is at any point purely terminological. I am much more interested in the metatheoretical or methodological issues that the use of one term rather than another, or of one term in addition to another, helps us to identify. As far as the terms ‘theoretical linguistics’ and ‘linguistic theory’ are concerned, I wish to suggest that, if they are kept distinct, each of them can be usefully employed to refer to what have now emerged, or are in process of emerging, as two rather different, but equally important, sub-branches of linguistics.

When my Introduction to Theoretical Linguistics (1968a) was published, more than twenty years ago, it was hailed by Bar-Hillel as “the first [book of its kind] … to carry the long overdue adjective ‘theoretical’ in its title” (1969: 449).