Smith (1973) presents a detailed analysis of his son's phonological development between the ages of two and four.1 The book is impressive, not only for the care with which the analysis was done but also, and more importantly, for the clarity with which central acquisition issues were stated. The analysis of the child's productions was done in two ways: first as a mapping from the adult system and second as a self-contained system. In his introduction, Smith raises seven issues that any theory of language acquisition must address; one of these concerns the nature of phonological change. Smith states that when changes occur in the child's output, they do so in an ‘across-the-board’ fashion. On the basis of this (and other) evidence, Smith concludes that the child must have the adult surface form as his underlying lexical representation. The implication is clear: the child must thus perceive in an adult-like fashion and the deviance of his/her output is due to the articulatory difficulty of certain sounds and sound sequences (and in some cases to certain formal properties of his mapping system).

Over the last several years there has been increased interest in and reliance on the role of phonetics in explaining various aspects of phonology. Such notions as ‘natural rule’ and ‘phonetically explainable’ are commonly equated and are incorporated into arguments over the appropriateness of some given rule formulation or over the range of analyses permitted under alternative theoretical approaches. Those who have made the strongest, most explicit appeal to phonetics in this regard include Stampe (1969, 1973), Ohala (1971, 1972, 1974a, 1974b, 1975, 1978), Schane (1972), Harms (1973), and Hooper (1976). The common thread in these various appeals is the claim that some or all defensible phonological rules are phonetically explainable.

The main goal of linguistic research is to develop a theory of grammar, i.e. a set of universal principles to characterize human language. Since languages vary superficially, this goal is achieved only when it is shown that superficial differences among languages can be accounted for by the theory of grammar. Clearly, then, the more that superficial differences among languages can be accounted for by universal principles, the more the field will advance. This conclusion is best stated by Perlmutter (1971): … progress in linguistics depends on extracting as much as possible from grammars of particular languages and formulating general principles from which the facts of particular languages will follow as automatic consequences….

This paper argues that the theory of Universal Grammar must include both implicational universals and universals of language distribution in the description and explanation of word order.

Implicational universals are of the form: if a language has some property (or properties) P, then it will also have some property (or properties) Q. These ‘if P then Q’ statements require that a language must not have property P without property Q (*P –O).

According to Hooper (1976: 5), the theory of natural generative phonology places the strongest possible constraints on abstractness in phonological descriptions. She claims that ‘the major advantage of natural generative theory over previous theories is that it gives a realistic representation of linguistic competence by constraining the theory to allow only a small subset of the grammars allowed by the unconstrained theory’ (xi). She claims further that the subset of grammars which are consistent with the constraints of natural generative phonology (NGP) accurately represent a speaker's knowledge of his language. This paper considers how vowel harmony rules in languages such as Turkish and Hungarian would be formulated in NGP. It is argued that the analyses required by NGP for these and similar vowel harmony languages do not represent the internalized knowledge of speakers of these languages and, therefore, that such vowel harmony systems provide strong evidence against NGP.

Modern Greek has a Relative Clause Formation process by which the target of Relativization is deleted under identity with the head of the Relative Clause – these Relative Clauses are introduced by the invariant complementizer particle pu, which also introduces factive complements. (Greek also has a movement strategy for Relative Clauses, with an inflected Relative pronoun, but the details of this process are irrelevant here.) Examples of the deletion strategy are given below in (i):

In the last twenty years, Cummings's line ‘he danced his did’; has been at the centre of almost every major discussion of the deviant language of literature. In the last two decades, hundreds of pages and scores of articles have been written which have attempted to assess and motivate its linguistic status and aesthetic success. Without question, ‘he danced his did’ has been more closely scrutinized and extensively analysed than any other single phrase in English literature.

Two American linguists, Ferguson (1959) and Householder (1962) who have worked on the subject of Greek diglossia from a technical and theoretical point of view have predicted that in the conflict between Katharevousa (K), an artificially created archaic form of Greek and Dimotiki (D), the language based on the vernacular, the latter will finally win.

The debate between the supporters of these two language forms has recently entered a new phase due to the Karamanlis' government decision to recognize the D as the official language and as the medium of all levels of education. The linguistic activity which followed this event falls into three main types.

In this paper I set out to show that, contrary to expectation, Subject-to-Object Raising does not form part of the derivation of Accusative and Infinitive (A&I) complements in Latin: and that this, coupled with the apparent NP-status of these complements, causes some difficulties for two of the current ‘schools’ of transformational-generative grammar, namely Generative Semantics/Relational Grammar, and Chomsk's Extended Standard Theory.

After a brief note on the use of intuition in analysing a dead language, the basic positions of the above two schools are given, together with certain distinctions and terminology relevant to the Latin data under consideration.

Harris (1978) discusses the ‘descriptivist’ and ‘non-descriptivist’ interpretations of performative utterances; on the former account, a performative utterance is an assertion, and may thus be true or false, depending on whether the speech act is felicitous or not, and on the latter account, the performative utterance itself constitutes the act in question, and is not an assertion, and therefore has no truth value.1 He presents the following argument against the descriptivist position. A speech act (I apologize, say) may be reported by using the same verb non-performatively: He apologized. This report, however, can be true even when the reported speech act was infelicitous in some way, e.g., He apologized, but to the wrong person. He claims that utterances of this type create a dilemma for the descriptivist position. If the assertion reported is considered to be true in this case, then its truth value does not depend on the success or failure of the performance.

Theorists of performatives are very much at odds with one another when it comes to the crucial question which has divided them into two camps; that is, are performative utterances true or false? One side, including Austin (1962), Black (1963), and Harris (1978), denies that the question is relevant, while the other side, including Hedenius (1963), Lemmon (1963), Wiggens (1971), and Åqvist (1972), insists that the question is of primary importance for understanding the use of performatives. These two camps have been termed, respectively, the ‘non-descriptive’ and the ‘descriptive’ interpretations. In a recent issue of this journal, Harris (1978) suggested that the descriptive interpretation of performatives is self-contradictory.

In an important theoretical discussion, Sampson (1975) argues against what he terms the ‘single-mother condition’. This is the condition that no node may have more than one mother. It is central to the definition of a tree, as the term is normally understood in linguistic theory. Sampson argues that the condition should be abandoned, and that what he terms ‘semitrees’, graphs in which nodes have more than one mother, should be permitted as representations of linguistic structure.2 He argues for this innovation on the basis of certain anaphoric phenomena. In this note, I want to take a critical look at his argument. I will suggest that the semitree approach to anaphoric phenomena that he sketches is untenable, and hence that these phenomena provide no motivation for an abandonment of the single-mother condition.