Recently, linguists have discovered (or, more accurately, rediscovered) the role that historical linguistics can legitimately play in providing explanations for the facts of synchronic language types. Greater awareness of synchronic variability and its source in historical changes in progress has focused attention on the concept of a language as a system not of static structures, but of interacting processes which the individual speaker becomes involved in as he/she acquires the language and enters the language community (Weinreich, Labov & Herzog, 1968). These processes are hypothesized to be universal, in that they may occur in any language family at any time period (though the initial actuation of a change is still a riddle, ibid. 102).

Anaphora clearly involves syntactic, semantic and pragmatic factors. Although it is generally acknowledged that pragmatic factors are predominant in discourse anaphora, it is equally widely held (especially among Government-Binding (GB) theorists) that only syntactic and semantic factors are crucial to intrasentential anaphora. In this article, I shall argue, in the spirit of an ongoing debate about the ‘division of labour’ between grammar and pragmatics regarding anaphora (Reinhart, 1983a, b, 1986; Kempson, 1984a, b, 1988a, b; Levinson, 1987a, b, 1991; Yan Huang, 1987a, 1989), that contrary to this popular but erroneous view, the contribution of pragmatics to anaphora is much more fundamental than has been commonly believed, even at the heart of intrasentential anaphora, at least with respect to languages like Chinese. Such a position, if established, would seem to decrease the plausibility of Chomsky's (1981, 1982, 1986) claim that anaphora, and zero anaphora in particular, have a privileged access to the alleged underlying principles of the innate Universal Grammar (UG), the biologically determined endowment of the human mind.

There has been much discussion, mainly by Montague grammarians, of the fact that ‘transportation’ of adverbials affects use or semantic interpretation of perfect sentences. Consider, for example, the following perfect sentences: (1) with a durational PP in sentence-final position; (2) with a durational PP in sentence-initial position; and (3) with a durational wh-phrase in sentence-initial position:

(1) John has been in Tucson for two years.

(2) For two years, John has been in Tucson.

(3) How long has John been in Tucson?

(4) (a) John has been in Tucson for two years up to now.

(b) John was in Tucson for two years in the past.

The key assumption in the standard generative approach to historical linguistics (King, 1969) is that each sound change is incorporated directly into the native speaker's grammar as the final phonological rule, moving up gradually into the grammar as further changes are implemented. Restructuring of underlying representations by later generations during acquisition is theoretically permitted, but infrequently invoked, with the result that the historical phonology of a language will be almost directly mirrored in the order of its phonological rules. The only extractable generalizations are then that the ‘highest’ rules will correspond to the oldest changes, and that a sound change and the rule into which it is converted will tend to be identical or at least show a high degree of similarity in formulation. This approach casts no light at all on the problem of the implementation of sound change.

This article argues in favour of a syntactic analysis of secondary predication as first proposed in Williams (1980) and against the semantic approach to predication defended most recently in Napoli (1989). Most of the evidence to be adduced in the proposed analysis comes from Spanish. We posit that secondary predicates are base-generated inside VP in Spanish, confirming Culicover & Wilkins's (1984) and Roberts's (1988) analysis for English. Under this hypothesis, object-oriented (O-) predicates occur inside the c-domain of V as sisters to the verbal complement, whereas subject-oriented (S-) secondary predicates attach to VP as sisters to the subject. Based on this asymmetry and adopting Jaeggli's (1981) assumption that initial wh-traces must be lexically identified, we explain the fact that only O-predicates can undergo wh-movement in Spanish. This restriction on wh-movement follows from the syntactic configuration assigned to secondary predicates in which both predicates are governed by V, but only O-predicates are c-commanded by V.

In the search for linguistic universals which typifies the generative approach to grammar, nothing is as revealing as phenomena which manifest themselves in a wide variety of languages. By exploring the similarities and differences that we find in a single phenomenon cross-linguistically, we can gain insight into the nature of the linguistic universals that are responsible for the phenomenon in question. This study is an investigation into one such phenomenon, morphological causativization (henceforth causativization), which may be illustrated by the following examples from a variety of languages.

Chomsky (1981: 188, 220; 1986a: 166) formulates the Binding Theory essentially as in (1).

(1) (A) An anaphor must be locally bound

(B) A pronoun must not be locally bound

(C) An R-expression must not be bound

The notion ‘bound’ is defined as ‘c-commanded by a co-referential element’. As for the notion ‘locally’, that of Chomsky (1986a) differs somewhat from that of Chomsky (1981), and much recent literature addresses the issue, especially in connection with the phenomenon of ‘Long Distance Anaphora’. (For relevant discussion see Burzio (1989c and references therein) and also Levinson, this volume.) For most of our purposes, it will be sufficient to assume Chomsky's (1981: 188) ‘within its [i.e. the anaphor's/pronoun‘s] governing category’, or even the formally simpler (though empirically less adequate) ‘within the same minimal clause’. The empirical effects of the Binding Theory in (1) can then be illustrated as in (2a, b, c) – instances of local binding, non-local binding, and no binding respectively. In each case the connecting line expresses intended co-reference, much as co-indexation in later examples. Each ungrammatical case is accounted for by the principle indicated in parentheses.

Since Paul Grice published ‘Logic and conversation’ in 1975, there have been a number of attempts to develop his programmatic remarks on conversational and conventional implicatures further (see Gazdar, 1979; Atlas & Levinson, 1981; Horn, 1985; Sperber & Wilson, 1986; and especially Levinson, 1983, and the references cited therein). The result has been a growing understanding of the relationship between semantics and pragmatics, and more generally of human reasoning in everyday language use. Many aspects of natural language understanding that were previously thought to be part of the conventional meaning of a given expression can now be shown to be the result of conversational inference. And with cancellability as the diagnostic test, a number of traditional problems in the study of meaning are yielding to more satisfactory analyses. Even more ambitiously, implicatures are penetrating into core areas of the syntax, as pragmatic theories of increasing subtlety are proposed for ‘grammatical’ phenomena such as Chomsky's (1981, 1982) binding principles (see Reinhart, 1983, and Levinson, 1987a, b, 1991).

In an earlier article (Levinson, 1987b), I raised the possibility that a Gricean theory of implicature might provide a systematic partial reduction of the Binding Conditions; the briefest of outlines is given in Section 2.1 below but the argumentation will be found in the earlier article. In this article I want, first, to show how that account might be further justified and extended, but then to introduce a radical alternative. This alternative uses the same pragmatic framework, but gives an account better adjusted to some languages. Finally, I shall attempt to show that both accounts can be combined by taking a diachronic perspective. The attraction of the combined account is that, suddenly, many facts about long-range reflexives and their associated logophoricity fall into place.

Greenberg (1963) formulated a number of implicational universals that refer to the order of various syntactic elements. He classified languages on the basis of their order of subject, object and verb into three types, which he labelled I, II and III, and which correspond to what are commonly known as VSO, SVO and SOV languages, respectively. Since that time, evidence for the existence of the three other logically possible orders, VOS, OVS and OSV, has been presented (see Keenan, 1978; Derbyshire & Pullum, 1981, 1986), thus leaving us with a typology of six types. Lehmann (1973, 1978) and Vennemann (1974, 1976) collapsed these six types into two types OV and VO. Implicit in this move is the idea that the fundamental parameter is the order of verb and object and that the position of the subject is less important. In treating VSO, VOS and SVO as subtypes of the general type VO, the claim is that these three types are similar to each other in their other word order characteristics and different from OV languages. More recently, a number of linguists, including Comrie (1981: 90, 94–95; 1989: 96, 100–101), Mallinson & Blake (1981: 379), Siewierska (1988: 18–19) and Payne (1990: 19), but most particularly Hawkins (1980: 199; 1983: 30), have criticized Lehmann and Vennemann for collapsing VSO, VOS and SVO languages into a single category VO. They all argue that the available evidence does not support the claim that SVO languages pattern like VSO and VOS languages. The purpose of this paper is to argue that, although some of these criticisms are not without merit, Lehmann and Vennemann were largely right: with certain well-defined exceptions, the word order properties of SVO languages differ little from those of VSO and VOS languages. In short, it will be shown that with respect to a large number of word order characteristics, we do find a basic split between VO and OV languages.