This paper examines the applicability to the history of linguistics of Thomas Kuhn's conception of the history of science. It concludes that his notion of REVOLUTION, borrowed from the history of the non-sciences, can be applied to the history of linguistics; but the same is not true of his other key notion, the PARADIGM. The possession of paradigms, according to Kuhn, is what distinguishes the hard sciences from fields in the humanities and social sciences which have not achieved scientific maturity. Kuhn regards a paradigm as (1) resulting from an outstanding scientific achievement on the part of a single innovator, and (2) commanding uniform assent among all the members of the discipline. If these two requirements are to be everywhere met, the concept cannot be applied either to the history or the present state of linguistics. Serious objections can also be raised to other features of Kuhn's theory, such as the view that shifting allegiance from one paradigm to another is a largely irrational process. The paper recommends, then, that linguists abandon the theory.

Shibatani 1973 has suggested that surface phonetic constraints are a necessary part of a psychologically real grammar. I claim that, in addition, such constraints are sufficient—that they are the only true generalizations in a generative grammar, and that morpheme-structure conditions play no independent role (and thus have no place) in phonological theory. Both synchronic and diachronic arguments for this position are presented. An implication of this claim for phonological theory, aside from the obvious change in the form of a grammar, is that, without the possibility of referring to the simplicity of MSC's, analyses based on certain assumptions about the nature of underlying representations can no longer claim support from the theory.

In the markedness model of SPE (Chomsky & Halle 1968), redundancy rules lose their ability to characterize regularities and must convert ‘optimal’ phonemes to ‘real’ ones. Lexical redundancy and low-level segmental redundancy are confused in the SPE model, in which values are specified for all features in lexical matrices. In assigning values to features with no phonetic correlates in certain major classes, absolute marking conventions create unnatural classes and unpronounceable segments; their output should be expressed in u's, and interpretive marking conventions should assign values only to meaningful features.

Phonetic theory has generally been based on the assumption that speech can be divided into discrete segments, each phonetically homogeneous with respect to all linguistically significant parameters. Among the nasal consonants, however, are segments which must have internal sequential structure.

Primary nasal consonants are first discussed here, and it is concluded that these should be described as non-continuants. Various possible characterizations of complex nasal segments (pre- and post-nasalized stops) are then considered, and it is concluded that no feature homogeneous over the segment is adequate to describe them. The conclusion that these segments involve internal sequencing suggests that rules should exist which can produce and manipulate such intra-segmental structure. Applications of such rules are briefly suggested.

This paper offers a generative treatment of the phonological behavior of the prefix in- of Modern French. It is shown that the seemingly perplexing distribution of some of its allomorphs before stems with initial sonorant consonants is related to the predictability of the meaning of complex words from the meanings of their component parts, the semantic structure of the words (relative scope), and the productivity of word-formation rules. The formal solution presented comprises a series of morphophonological rules (nasalization, assimilation, degemination), and a proposal for the way in which in- is incorporated in the lexicon. The use of the Elsewhere Condition (Kiparsky 1973) is shown to provide a simplification of the rules, corresponding to a meaningful explanation of the relation between the rules of nasalization and assimilation.

In discussing rule opacity as a factor influencing phonological change, Kiparsky 1971 mentions three common fates of opaque rules within the phonological system: loss; re-ordering with respect to other rules to yield greater transparency; and morphologization. In this paper I discuss another sort of systematic reaction to the presence of an opaque rule, a reaction that has so far been overlooked in the discussion of rule opacity: the elimination of opacity-inducing environments in inflectional paradigms through analogic affix replacements. I argue that effects of this sort will continue to be overlooked until we recognize, and deal explicitly with, non-phonological aspects of inflectional change.

A number of syntactic constructions claimed by linguists to be restricted to main clauses are shown to occur, in fact, in a variety of subordinate clause types, but only under certain mysterious conditions—basically, when the speaker desires to be understood as committed to the truth of the subordinate clause. Formalization of this notion, however, remains elusive. The analysis of this phenomenon by Hooper & Thompson 1973, while attractive and initially explanatory, is shown to be, before the final analysis, incapable of accounting for the range of embedding environments allowed, which differs from one restricted construction to another. It is shown in the present paper that an adequate solution will involve a complex interaction of several factors—syntactic, semantic, and pragmatic—and the range of data to be accounted for is partially delineated.

While reduplication of definite direct objects is routine in the Balkan languages (e.g. Albanian E panë kështjellën? ‘Did they see the castle?’, literally, ‘it they-saw castle-the?’), reduplication of indefinite direct objects has been regarded as ungrammatical or at best exceptional (e.g. Albanian (*E) panë një kështjellë? ‘Did they see a castle?’, literally, ‘(*It) they-saw one castle?’) This paper attempts to show that the latter type of reduplication is not necessarily ungrammatical but that there are instead two kinds of indefinite direct objects, namely specified and non-specified ones. The former may undergo reduplication, the latter may not.

Considerable syntactic evidence from English has been presented in the literature for a rule of NEG-Raising. It is shown here that equivalent or analogous evidence exists in French (time-adverbials, tags, corrective responses, negative particles and polarity items, mood). NEG-Raising and its domain are then reconsidered from a functional perspective, whereby the transformation is shown to be a HEDGING device—operating only in sentences containing a type of pseudo-performative, which also has the function of hedging. Such an analysis not only is supported by the original data, but also automatically accounts for otherwise inexplicable and unrelated syntactic facts.

It is proposed that NP's determined by the generic indefinite article represent abstract concepts, and as such are not inherently different from indefinite NP's appearing in copulative predicates. A derivation for generic indefinite NP's that reflects this is formulated (and evidence is adduced to show that any cannot underlie generic a); this is shown to be explanatory as regards the phenomenon itself, its relationship to other kinds of generic NP's, and its relationship to other kinds of indefinite NP's. It also enables us to formalize the notion of ‘generic sentence’.

Indo-European verbs inflected in the middle voice have taken on passive meaning; explanations for this have been proposed, but each presents difficulties. Here an alternative paradigm of language change is set forth to explain the process. This paradigm is, among other things, dependent on the fact that more than one structural description may be assigned to a given sentence. If the hearer assigns a different structural description to a sentence than the speaker does, and if a rule based on this ‘deviant’ structural description spreads, then the language exhibits change. Under certain circumstances, then, agentless OV constructions may be mis-assigned the constituent structure of a passive SV construction—where the object of the active, transitive sentence is interpreted as the subject of its passive, intransitive counterpart.

The century-old failure of historical linguistics to discover regularities of semantic change comparable to those in phonological change, as described by Grassmann or Grimm, has forced us to entertain as ‘semantic laws’ proposals that express mere tendencies, or are so restricted to a particular time, language, or narrow inventory, that the ‘law’ is indistinguishable from a description of a discrete historical event. But in the lexical field of English adjectives referring to sensory experience, there has been a continuing semantic change so regular, so enduring, and so inclusive that its description may be the strongest generalization in diachronic semantics reported for English or any other language. On the basis of very similar evidence from Indo-European cognates and from Japanese, the possibility exists that the regularity described here might characterize more than just these languages. It qualifies as a testable hypothesis in regard to future semantic change in any language.