This paper is concerned with strategies for relative clause formation, both in the universal and language-specific sense. It builds on the pioneering work of Keenan & Comrie (1977) and proposes some extensions to this work. The basic issues addressed here concern what exactly ‘case-coding’ strategies in relative clause formation are, why languages should use such strategies, and what form the strategies can take. Language-specific illustration is given from Modern Welsh, a language which exhibits some relativization strategies not discussed by Keenan and Comrie, but which nonetheless broadly supports their proposals.

Notions of ‘head’ and ‘modifier’ have a long and respectable history in syntax, although they have changed considerably in recent years. But the application of similar notions to morphology is relatively new. The extension of these notions to morphology has been prompted by two developments:

(i) Intimations from dependency grammarians that their approach to language ought to be generalizable to morphology (Andersen, 1980); and, far more importantly

(ii) The development of lexicalist theories of word-formation using an X-bar notation.

One of the major innovations within post-SPE generative phonology has been the development of frameworks where phonological units are organized in a non-linear fashion. Taking autosegmental phonology (Goldsmith, 1976) as our main exemplar of such frameworks, we wish to address the following question: What is the appropriate interpretation of autosegmental representations? There is, of course, a further question about what we mean by INTERPRETATION: formal, phonetic or computational interpretation? Although we will concentrate on the first of these, we believe that all three aspects should be regarded as closely inter-connected and mutually constraining. The question of interpreting autosegmental representation has in fact been recently posed by Sagey (1988), and we shall take her proposal as our starting point. While it is uncontroversial to suppose that the relationship between units on a given autosegmental tier is one of temporal precedence, Sagey claims that it is more problematic to pin down what is meant by association between tiers. She argues, cogently we believe, that if association is taken to be a relationship of simultaneity between durationless units, then standard analyses of complex segments and gemination lead to logical inconsistency. Instead, association should be taken as temporal OVERLAP between units with duration.

The phenomenon of stress in Old English (OE) has been the subject of thorough and extensive study for well over a century. Indeed the foundation for any modern study remains the work of Eduard Sievers (1885, 1893a, b), well summarized in Campbell (1959). The present paper is not concerned with a revision of the ‘facts’ of Sievers' account, although we shall note below instances where we disagree with those facts, but rather with a linguistic explanation of those facts. Sievers' account is essentially a statistical presentation of varying stress patterns, and he made little attempt to provide explanations of either frequent patterns, or non-existent ones. The framework in which we shall operate is that of lexicalist metrical phonology. Within that framework we shall attempt to demonstrate that Old English stress was organized in a way very different from that in Modern English. Most particularly we shall suggest that there is a central rule (the Old English Stress Rule = OESR) which, in contrast to the central rule for present-day English (PDE), operated from left-to-right. This, we shall suggest, has direct implications for the operation of other features of stress derivations, such as Destressing. Further, we shall argue that it is probable that level-ordering has no role to play in the stress phonology of Old English.

The major assumption of notional grammar is that the syntactic classes established for each language on the basis of their distributional properties are labelled on notional grounds: the denotata of the prototypical members of classes meet universal ontological definitions. For example, one might suggest that: ‘As far as nouns are concerned, the prototypical denotata are persons, animals, and other discrete physical entities…’ (Lyons, 1989: 161). Despite the work of Lyons (notably also 1966, 1977) and others, involving a tradition of some antiquity, adequate explicit definitions for the range of classes are lacking; much remains intuitive. This should not prevent us, however, from examining the syntactic consequences of the notionalist assumption. Nor, of course, should the existence of non-prototypical class members (such as nouns which denote non-physical entities or whose denotata are of limited temporal extension - see further below). What follows explores the consequences for a specific area of classification of one articulation of a notional theory.

A nonrestrictive relative clause (henceforth NRR) is shown in (I) and a restrictive relative clause (henceforth RR) in (2).

(1) The swans, which are white, are in that part of the lake.

(2) The swans which are white are in that part of the lake.

Example (1) implies that all the swans under discussion are white. Example (2) implies that the white swans are being distinguished from some other not white swans which are also under discussion. There are many superficial differences between restrictive and nonrestrictive relative clauses; in this paper I show that there is no need for construction-specific stipulations which distinguish between them. The differences arise from the fact that the RR is a modifier, while the NRR is not, and in fact has no syntactic relation to its host/antecedent. Co-indexing (involving a referential index) between the relative clause and its antecedent is central to this account. I examine the requirement that a relative pronoun must have an antecedent, which in the case of a NRR is the sole manifestation of the relationship between the relative clause and its host), and suggest that this holds at a level of discourse structure.

One of the less rewarding of our common interdisciplinary pursuits is lifting theoretical concepts from subjects not our own, and using them in contexts very distant from those they were intended for. Such borrowings often turn from theoretical claims into sloppy metaphors, leading to varieties of ‘vulgar X-ism’, the resuit of overenthusiastic appropriation with insufficient sense of the subtlety or precise applicability of the originals. Spencer's ‘Social Darwinism’, vulgar-Freudian or vulgar-Marxist literary analysis and sociology are nice examples. Linguistics, being less unique than linguists often think, is no exception: Praguian and neo-Praguian functionalism may be a kind of vulgar Darwinism, extending notions of ‘adaptation’ or ‘selective pressure’ to the inappropriate domain of language Systems (see Lass, 1980a). But every once in a while such transfers seem to work, like Darwin's borrowings from late eighteenth-century Scottish economie theory; if not always through direct applicability, then by focusing on new ways of interpreting old data, or providing a basis for linking disparate phenomena as instances of a new (putative) natural kind.

In this article data are presented from the Amele language and a number of other different Papuan languages to show that in these languages the medial verb form is marked for a binary distinction of realis versus irrealis modality. Furthermore it is demonstrated that in these languages the realis-irrealis distinction interacts with categories of tense and mood marked on the fina! verb in such a way as to divide them into the two basic domains of realis and irrealis modal status. It is also demonstrated for two of the languages (Amele and Nobonob) which have sentence particles with an illocutionary force function that these particles do not interact with the realis-irrealis modal system marked on the medial verb. In the final discussion and conclusion a schema is suggested to account for the facts of future tense, imperative mood and counterfactual mood being categorized as irrealis in these languages. The schema suggests that events can be referred to in the unreal world as either events that are potentially true/false in the real world, or events that used to be true/false in the real world, or events that are not true/false in the real world.

This paper explores the fit between syntax and semantics in one small corner of English. It addresses two related questions: To what extent is the distribution of bare infinitive (BI) complements, as in (1) and (2) below, semantically motivated?

(1) I saw/heard Mary slam the door.

(2) I made/let John cross the street.

Chomsky's (1981) i-within-i condition (also written as i/i) blocks co-indexation of a phrase with one of its proper subconstituents:

(I) *[…a1…]i

We argue here that the i-within-i condition as stated in (I) and used in current work is both empirically inadequate and theoretically incoherent.2 Many of the data that the i-within-i condition has been taken to account for should, instead, be accounted for by a range of unrelated constraints. However, a condition on the interpretation of the reference of free pronominals and anaphors - which we state as a definition of circular chains - is, in fact, motivated and will adequately account for the data involving referential circularity that have previously been accounted for with the i-within-i condition.

In Roca (1988) I take to task Harris (1983) and Den Os & Kager (1986) and propose a novel approach to Spanish (nonverbal) word stress according to which (i) the algorithm is not quantity sensitive and thus differs from that of Latin, the parent language, (ii) the system does not obey the Uniformity of Algorithm Principle, (iii) the algorithm operates on the morphological stem, rather than on the whole word. As regards (i), I made the suggestion (418, ibid.) that the preferred paroxytone stress in native forms with penultimate closed syllables must be formalized as a lexical prohibition on extra-metricality markings in the given environments. Clearly, the formalization of this proposal and the exploration of its consequences are central to the analysis. Moreover, while the arguments for a relaxation of the Uniformity of Algorithm Principle still stand, departures from the principle are doubtless costly and therefore best avoided. Finally, in Roca (1988) there is no discussion of the consequences of (iii) above for the cyclic data presented in Harris (1983), the compatibility of which with (iii) is consequently unclear. All these facts warrant a re-examination of the issue, which I propose to carry out in the present paper.

In his paper ‘On Horn's dilemma: presupposition and negation’ Burton-Roberts (1989a) presents an ambitious programme, formulated right at the outset. He seeks to establish three points:

(i) Under the ‘standard logical definition of presupposition’ a pre-suppositional semantics is INCOMPATIBLE with a SEMANTICALLY AMBIGUOUS NEGATION operator (SAN), on pain of the theory being rendered ‘empirically empty and theoretically trivial’,.

(ii) From this it follows that the one unambiguous negation is presupposition preserving. Cases that have been identified as presupposition-cancelling negation should be re-analysed as ‘instances of a pragmatic phenomenon’, not unlike what has been proposed in Horn (1985), that is as METALINGUISTIC NEGATION (MN).

(iii) This pragmatic analysis of MN ‘itself implies a presuppositional semantics’, that is to say ‘a presuppositional theory of truth-value gaps’.

In his response to ‘On Horn's Dilemma’ (Burton-Roberts, 1989a, henceforth OHD), Seuren makes it clear that he believes that the argument of OHD is vitiated by my failure to pay attention to his own logical system for presupposition. That system embodies (a) a pre-theoretical conception of P-failure (as in The king of France is wise) as being tied to a THIRD TRUTH-VALUE (‘Radical Falsity’) and (b) his view that a semantic ambiguity of negation (SAN) exists and should be welcomed. OHD, by contrast, was concerned with the connections between (a) the Strawson-Frege definition of P, (b) a pre-theoretical view of P in which P-failure is tied, not to a third truth-value, but to LACK of truth-value, and (c) Horn's view of negation as being semantically UNambiguous in the object language. So I disregarded Seuren's theory as being irrelevant to the argument of OHD. I shall make up for it here, though, showing that Seuren's system, and his own claims for it, provide ample corroboration of those arguments and, indeed, of the arguments of Burton-Roberts (1989b) against any formally trivalent approach to the logic of presupposition. I set the scene first with a brief account of the recent history of the distinction between trivalence and gapped bivalence (Section 2). Then I consider in Section 3 the argument of OHD and Seuren's response, and go on to examine Seuren's system (Section 4). Finally, and very summarily, I compare the pragmatic account of special negation outlined in OHD with the semantic account favoured by Seuren (Section 5).