I. I. In a paper dealing with the analogical elimination of allomorphy and the explanatory (in)adequacy of generative phonology in accounting for it, Hogg hypothesized that ‘a necessary condition for analogical change is that the alternation which is subject to the analogy must not be synchronically rule-governed’ (1979: 58). This observation was elevated to the status of a principle which was formulated as follows:

Analogical extension or levelling of some phonological rule can take place only if that rule is no longer present in the synchronic grammar of the language (Hogg, 1979: 8).

Starting from the well-known difference between I ‵thought he was married (which he is) – where the complement is given a factual interpretation – and I thought he was ‵married (but he isn’t) – which is counterfactual – this article examines the syntactic, semantic, contextual and prosodic conditions under which factual and counterfactual interpretations can emerge. Factuality is distinguished from factivity, and concerns a person’s (usually the speaker’s) subjective assessment of whether the propositional content of a sentence or sentence fragment conforms or conflicts with his perception of reality. The modal nature of factuality is stressed, as is the contribution of tense and intonation to the emergence of factuality interpretations.

It has been argued in a number of studies that the facts of loan phonology in Japanese and in Miami Cuban Spanish provide strong support for David Stampe's theory (see especially Stampe, 1973, Donegan & Stampe, 1979) of ‘natural phonology’ (cf. Ohso, 1971 Lovins, 1973, 1974; Bjarkman, 1976). To the evidence adduced in these studies, I would like to add some evidence from English. The English evidence is especially compelling, since it involves not only actual nativizations, but (intuitions about) IMPOSSIBLE nativizations.

The feature [lateral] has a very limited morphophonemic or classificatory role which makes its presence in the universal phonetic feature inventory questionable. A mode of elimination is suggested which involves (i) treating all laterals as [+ distributed]; (ii) limiting [distributed] as a distinctive feature to continuants; (iii) redefining [anterior] in terms of [labial], [apical] and [dental]; (iv) analysing approximant rhotics as glides (Kahn, 1976); (v) adopting Halle and Stevens' re-analysis of [coronal]. This also provides a natural characterization of retroflex sounds and of sibilants. A conceptual problem in child phonology is discussed in the light of the above proposals.

This paper shows how the occurrence of parasitic gaps in preposed adverbial clauses and licensed by a subject trace can be used in determining the type of adjunction of such adverbial clauses. The sometimes very subtle grammaticality judgments are supported by some evidence from traditional grammars.

Defining the constraints on the use of the modals will and would in futurate if-clauses in English poses problems to both descriptive and theoretical linguists. One ‘rule’ often referred to is that while future will and counter- factual would do not normally occur in conditional sub-clauses with if, volitional and dispositional will and would can be used in that context.

This paper attempts to account for this general rule and for the so-called exceptions to it in terms of two different syntactic positions of if-clauses in the sentence. It is suggested that the contrast can be formalized within current versions of -theory.

One of the most salient differences between Irish English and Standard English (in the sense of Received Pronunciation, Gimson, 1980:89 ff.) lies in the realization of coronal segments. I use this term in the standard sense of Chomsky and Halle (1968:304) and intend it to be understood as a convenient means of referring to several groups of sounds which happen to have in common that they all involve the raising of the point or blade of the tongue from a putative neutral position. In using this term I do not necessarily pledge my support to its effectiveness in phonological description (see 3.1 below for a discussion of distinctive features with reference to Irish English). Precisely what segments are involved here will be clear from the remarks below. Before starting however, a word on the term Irish English is called for: by it I mean the variety of English spoken in the Republic of Ireland (on this see Barry, 1982:101 ff.) as the phenomena which I will be discussing are either not at all or only partially found in the variety of English spoken in Northern Ireland (consider the realizations of / t /and / d / discussed below which are unique to the Republic). Furthermore I have allowed myself the generalization Irish English although there is considerable variety in the English spoken in the Republic of Ireland. But the term Irish English is used in a fairly restricted sense here: it refers to urban middle-class speech. This excludes contact Irish English (in the areas officially designated as a ‘Gaeltacht’ (Irish-speaking area) in which both Irish and English are spoken), rural Irish English and lower class urban English. References to, and comments on these latter varieties are labelled specifically as such.

The history of generative accounts of vowel harmony in Classical Manchu is discouraging. The early treatments (Vago, 1973; Odden, 1978; Finer, 1979) assumed the wrong type of vowel harmony (front vs. back rather than relative height harmony). Hayata (1980) proposed the correct type, but was unable to justify it adequately within the generative framework. These shortcomings are due to deficiencies in the practice of generative phonologists. To demonstrate the correct type of vowel harmony in Classical Manchu requires considerations of (1) the actual surface representation and/or (2) facts in related dialects and languages. Neither type of evidence is frequently used in the practice of generative phonologists, although the former is almost always valued in theory. An implication is that attempts to improve phonology must weigh practice as well as theory.

The claim made by Harris (1978) that the descriptivist position in respect of performatives is self-defeating has engendered some spirited controversy through the years since then (see Edmondson 1979, 1983, Wachtel, 1980; Spielmann, 1980; Taylor & Wolf, 1981). Let me declare right at the outset that I am with the non-descriptivists. Nevertheless, in what follows, I wish to show (a) that no brusque argument of the sort originally advanced by Harris or subsequently marshalled by the others can help settle the dispute either way, and (b) that the controversy so far has been riddled with a good deal of futile cross-talk which has blinded the participants in the intellectual free-for-all to some of the real issues at hand.

Diachronic change emerges from the construction of grammars by successive generations of learners by exposure to primary data. Learners receive little information concerning what is ungrammatical in their language. To propose learnable systems, linguists should use positive data and not depend on negative information (the learnability criterion). Learnability is relevant to historical linguistics under the view that links diachronic change to grammar construction by the learner. This perspective is used to establish the system of two types of free relatives in thirteenth-century Spanish (quanto que vs. quanto-constructions), and the changes that led to the disappearance of the first type in later centuries.

In some languages, vowels occurring in some specified domain, which is usually the word, must share some phonetic property or properties. Such languages are said to have vowel harmony.

Luganda, a Bantu language of Uganda, is unusual in having a vowel harmony system which is entirely morphologically conditioned. There is one subsystem which operates in the noun phrase and another irregular subsystem which operates in the verb phrase. The aim of this paper is to show how the non-linear theory of phonology developed by writers like Clements (1980, 1982) can throw some light on this apparently irregularity-ridden system.

It is well known that a cleft whose focal item functions as predicate nominal in the WH/that-clause is often unacceptable, or at least questionable:

(I) (a) *It's a genius that he is. (Leech & Svartvik, 1975: 181)

(b) *It's a conductor that John is. (Stockwell et al. 1977:107)

(c) *It's the football coach that John is. (Emonds, 1976:140) In this article I will investigate different types of clefts highlighting predicate nominals. I will point out five different restrictions that may lead to partial or complete unacceptability and develop a theory that can account for them.

A major controversy in syntactic theory concerns the nature of control verbs, verbs like try, which govern equi-NP-deletion in classical transformational grammar. For recent versions of the extended standard theory and, in particular, the government-binding theory, such verbs take a sentential complement with a PRO subject. (Cf. Chomsky 1980, 1981, 1982 and Koster & May 1982.) On an alternative analysis, originating in Bresnan (1971) and Brame (1975, 1976), and developed by Bresnan (1978, 1982), Gazdar (1982) and others, they take VP complements. A similar dispute arises over raising verbs, verbs like seem, which govern subject raising in classical transformational grammar.

The problem of how to model syntactic variation has received much attention over the last few years. Some of the well-known difficulties stem from attempts to extend to the syntactic level quantitative techniques which were originally developed to handle phonological variation. In particular, the application of the sociolinguistic variable model to the syntactic domain has come in for a good deal of criticism (e.g. Lavandera, 1978; Romaine, 1981). According to the classical Labovian formulation, all variants of a given variable must share ‘sameness of (cognitive) meaning’ (1972a: 271). Considerable doubts have been voiced about whether this requirement can consistently be met at the syntactic level. The problem is thrown into even greater relief when it comes to studying the sort of syntactic variation that occurs in situations of interface between standard and vernacular varieties. It has not generally been noted that assuming direct semantic equivalence between standard and nonstandard syntactic variants presupposes that they are embedded in structurally identical grammars. That is, apparently alternating standard and vernacular forms are simply treated as distinct surface realizations of the same underlying structure. The justification for this approach rests squarely on the belief that it is possible to offer a structural definition of the intuitively available concept of ‘dialects of the same language’. According to what we might call the PANLECTAL IDENTITY HYPOTHESIS, dialects of a single language are considered to share a common grammatical‘core’ and differ only in matters of low-level realization (e.g. Agard, 1971). Variation arising out of interface between dialects that are allegedly related in this way is thus describable in terms of variable rules which operate at low levels of realization.

1. On the (no doubt reasonable) assumption that cross-linguistic variation is subject to limitations due to the existence of universal constraints on linguistic rules and regularities, it is obviously desirable as a matter of descriptive principle to draw the limits within which individual languages may vary from one another as narrowly as the available cross-linguistic empirical evidence (plausibly interpreted) allows. Focusing on two interestingly interrelated aspects of linguistic regularities, viz. on possible conditions on agreement and on possible interactions of phonology and syntax, this paper once more tells a familiar tale: drawing the limits as narrowly as seems feasible vis-à-vis a relatively broad range of data may unfortunately turn out to be unjustified in the light of further, previously unavailable or – worse – ignored evidence. One always hopes, though, that having to redraw the limits is not tantamount to admitting that languages after all perhaps do vary without assignable limit.

This is a book written by a philosopher for philosophers; it needs a philosopher to review it properly. A review by a non-philosopher in the pages of a linguistics journal is justified, however, by the fact that the core of the book consists of a series of proposals concerning one of the fundamental questions of lexical semantics: what are we to make of the apparent variability in the meaning of a word from context to context? (This fact is not immediately inferrable from the title.) The present review focuses exclusively on this topic; it should be borne in mind, however, that the author's exposition of his views on word-meaning occurs as part of a more general discussion concerning the nature of analogy, and its role in literary, religious and legal discourse.