For the sake of argument, the extreme view is taken that the identification of phonological entities on the basis of articulatory or acoustic features is not in keeping with the strict structural linguistic tenet that only information from language itself be employed in determining the units used in linguistic description. The attempt is then made to show how the morphophonemes of Finnish are distinct from each other on the basis of their distribution within the word. The notions employed in the analysis are taken only from within language; i.e. no reference is made to nature or to the human body. The weaknesses of this approach are also pointed out.

Some of Pāṇini's rules contain negative compounds. In such cases there is ambiguity; the negative can be construed with the nominal following it in the compound, or it can be construed with a verb. According to the Mahābhāṣya, the first interpretation yields a positive rule providing an operation in a domain specified by the negative compound: non-x. The second interpretation yields a negative rule providing for the cancellation of an operation already provided for. Again, the first interpretation yields a one-step operation, while the second interpretation requires that two sentences be understood, providing two steps: tentative application of an operation and its subsequent cancellation. Both interpretations involve negation (pratiṣedha); the first type is called paryudāsa (pratiṣedha) 'limitation(al negation)', the second type prasajyapratiṣedha 'negation (subsequent to tentatively) applying'. The conclusions derived from the Mahābhāṣya are used to judge some formulations in the Kāśikā. In addition, it is shown that, while Patañjali usually demonstrates that both interpretations of an ambiguous negative compound will yield desired results, later commentators decide in favor of one or the other interpretation on the basis of economy. Finally, it is shown that the rules usually considered by Western scholars do not lend themselves to a clear understanding of the essential difference between paryudāsa and prasajyapratiṣedha.

Language is viewed as a system in which semological structures, manifested in experience, are symbolized by phonological structures, manifested in sound. Other animal communication systems exhibit a one-to-one symbolization of semological units by phonological units. The evolutionary expansion of human experience was accompanied by developments through which language came to diverge radically from the one-to-one model. Shorter term (non-evolutionary) change also has led language away from one-to-one symbolization, as exemplified here by the effects of phonological change and (on the semological side) of idiom formation. In consequence of both its evolutionary and non-evolutionary history, language exhibits an initial organization of experience within deep semology, a set of mutations which lead to surface semology, an initial phonological symbolization, and another set of mutations which lead to a final phonological organization manifested in sound. The place of separate semantic and syntactic components within such a system is questioned.

The Wu dialects of China are a group of dialects spoken in the southeastern part of Kiangsu province and in most of the province of Chekiang. The most characteristic feature of these dialects is the tripartite division of initial stops into voiceless unaspirated, voiceless aspirated, and voiced aspirated, a feature which constitutes a defining differentia of Wu. Within Wu, the voiced initials of Chekiang are more fully voiced than in Kiangsu, and there is less double reading because of borrowing from Mandarin. Another intra-Wu contrast discussed is the fronting of diphthongs cognate with Mandarin ou along the Grand Canal.

As the study of phonology has progressed, within the framework of generative grammar, certain principles have emerged governing the form and organization of phonological rules, and the manner in which these rules are to be interpreted. Several such principles are presented, with illustrative examples, from English, indicating the kind of evidence that supports them. Some possible alternatives are also discussed. Certain of these principles are quite abstract, and impose fairly heavy restrictions on the form and interpretation of grammars. If correct, they restrict the class of ‘possible human languages’ in non-trivial ways. The broader context in which these principles might have more general relevance is briefly touched on.

PIE intervocalic *w is preserved in the Primitive Irish of the earliest Ogam inscriptions, but is lost by the time of the Old Irish sources. The details of the loss are described here, with special attention to cases of interaction with the processes of vowel loss. It is shown that a number of problems can be resolved by the assumption that *w before *e and *i was palatalized to [w'], with subsequent unrounding to [i].

A system of day-names, indicating the sex and the day of the week on which a child was born, was carried from Africa to Jamaica in the seventeenth and eighteenth centuries. As personal names, they are now obsolete and almost extinct. All fourteen, however, survive as pejorative common nouns, retaining certain syntactic features (e.g. gender), but losing the purely semantic week-day feature. The series of historical changes, conditioned both by social influences and by the syntactic and semantic structure of the system, gives support to the subordination of semantic features to syntactic ones in modern linguistic theory.

Preliminary estimates of language divergence and word retention have been obtained by means of a log log transform from data consisting of the determinations of cognations in the lexicostatistical lists of Austronesian languages. Corrections have been obtained to the preliminary estimates by a maximum likelihood estimation method which assigns appropriate scores to represent the weight of evidence contributed by each determination of cognation.

The theoretical significance of a proposed computer program for the automation of discovery procedure in linguistics is first discussed. An outline is then presented of a type of heuristic computer program which would be suitable for the automation of the procedures employed in the initial phases of field work, particularly during the elicitation of paradigms. The basic characteristics of the trial and evaluation routines to be used in such a program are set forth. Finally, the basic assumptions underlying the proposed computer simulation of informant work are discussed.

The Proto-Algonquian word for ‘grandchild’ has harbored subtle problems of reconstructive detail for Algonquianists. The Ojibwa and Menomini forms disagree, but do not warrant setting up a fresh phonetic correspondence. On the basis of new Ojibwa data, a fresh solution is offered which actually inheres in the data already known. Given the protoform *n(e)-Hōšihsema, an explanation follows naturally from the respective formations of diminutives in Ojibwa and Menomini. The problem is submitted as illustrating a point of non-superficial reconstruction, having interest for general theory.

A new analysis of the Norwegian (and Swedish) word tones is presented, according to which Accent I is regarded as the unmarked member and is identified with primary stress, while Accent II is regarded as the only distinctive tone. The problem of functional burden is explored, and it is suggested that the relatively low importance of tone in communication is due to its structural dependence on the placement of stress. Rules are proposed for the distribution of tone, bringing out the invalidity of most of the minimal pairs presented in previous studies. It is shown that tone is a feature of the post-tonic syllable, and that its presence in the native and much of the borrowed vocabulary is determined by the morphology; in other vocabulary items, the occurrence of tone is highly arbitrary, being lexically and dialectically determined, with virtually no minimal pairs possible.

The continuing relevance of Bernard Bloch's ‘Postulates for phonemic analysis’ lies in the facts that they point to the basis of the minimal-pair identification test, the most powerful tool in phonological analysis, and that they further point to the basic necessity of sound-discrimination in the development of human language.

This paper contains no new information; it is merely an attempt to present, economically and clearly, one small portion of the design of the Yawelmani dialect of Yokuts. The implication is that the rest of the language could also be discussed in the relatively simple way demonstrated here. Newman's description (1944, 1946), from which I draw the facts, is notable for its sensitivity and realism, but is exasperatingly difficult to follow. However, if a more straightforward presentation is now possible, this is perhaps because we have developed some techniques of analysis and of statement not available to Newman three decades ago.

I am designedly using the term ‘genetic’ rather than ‘historical’ in this context because, after having been for thirty years a devotee of the discipline conventionally labeled ‘historical linguistics’, I have at present grave misgivings about the unqualified suitability of the label and, far more important, I have become alert to the implications of its possible inappropriateness.

To begin with, the label ‘history’, like many terms of ancient scholarship, is fraught with imprecision. It refers to an analytical discipline concerned with the study of the past, but also with a segment of that past regardless of any analysis, as long as there exist any written records. Only thus can we explain the fact that prehistory designates, at least in normal usage, another, earlier segment along the time axis, whereas such terms as prelinguistics (which mayor may not have been coined) and prephilology (which has actually been toyed with on one occasion) could meaningfully refer only to a prescientific stage of a discipline; that is to say, to cognition itself, not to an object of cognition.

Previous expositions of Japanese accent have assumed that unaccented forms of inflected words are to be equated in basic accent type with atonic nouns; they also assume that verb and adjective bases display two accent types, to be equated respectively with the atonic class of nouns and with the several tonic classes. It is possible to simplify the description by assuming that the phonemically unaccented forms have a morphophonemic final accent, automatically lost before juncture (which, in turn, may disappear). This leads us to reconsider the adjective bases, for which neither of the traditional patterns turns out to be atonic; moreover, an examination of new data reveals four different patterns, instead of the traditional two, with some bases exhibiting ambivalent behavior.

Samuel E. Martin has recently, for the first time, unequivocally demonstrated the Korean-Japanese relationship, but his sound correspondences are in the main based on comparisons between Middle Korean and Modern Japanese cognates. The Japanese forms which enter into his reconstructions are phonologically ‘antiqued’ on the basis of the traditional (and largely historical) orthography, but they do not take full cognizance of our present knowledge of the phonology of Old Japanese. The problem is particularly critical because of the nature of the Old Japanese vocalism, which, in contrast to the modern five-vowel system of the standard language of Tokyo, can be shown to have been an eight-vowel system. The present paper collects the available Old Japanese evidence for those Japanese forms which enter into Martin's comparisons, with the particular aim of seeing what light Old Japanese phonology may cast upon his reconstructions for the proto-Korean-Japanese unity.

Some postulates concerning the length of sentences, as advanced by Noam Chomsky, are considered in the light not only of grammatical possibilities, but also of behavioral probabilities; and certain implications for linguistic theory are suggested.

The phonemic inventory (overall pattern) of a language must account for every contrast occurring in any idiolect. Utterances occur in the phonemes of an idiolect, and phonemes are therefore the ultimate units of lexical items. But a three-stratum, twenty-seven level model for linguistic analysis indicates structural units—the morphophones—between phoneme and morpheme. Morphophones are seen as the basic units of the morphemes, and are themselves composed of dialectally different phonemic variants, which are non-contrasting in the same lexical items. Further, there are systematic alternations between morphophones themselves—substitutions, alterations, and replacements.

A system of linguistic description is proposed which distinguishes semological and grammatical levels. The first consists of units organized in multi-dimensional structures; those units can be combined into rudimentary ‘sentences’, each of which is related to numerous surface structures which are all paraphrases of each other. Implications are drawn for descriptive theory, lexicography, cross-cultural studies, and historical semantics.