It is well known that the earliest and probably best descriptive grammar in existence is the Sanskrit grammar of Pāṇini (?350 b.c.). The subject of the following essay is Pāṇini's phonological analysis of the Sanskrit consonants. This illustrates some of the more important techniques used by Pāṇini and his colleagues. It also enables us to reconstruct some of the methods by means of which Pāṇini constructed his grammar.

In the summer of 1938, at the Linguistic Institute in Ann Arbor, Sturtevant gave a public lecture on his Indo-Hittite hypothesis, which was then very much in the foreground of discussion. The next year, C. C. Fries, as Director of the Institute, persuaded Sturtevant to allow the lecture to be printed in the announcement and course program for the Institute of 1939, as the first (and, as it turned out, the last) of what Fries hoped might be a regular feature of such programs.

The lecture, in its published form, had a limited circulation. At the time and for a little while thereafter, Fries received a number of requests for copies. But the publication remained largely unknown, and in particular seems never to have reached European scholars. For this reason, Fries has now suggested that the lecture be reprinted in Language, so that it may at last enjoy the wider circulation which it deserves as the most concise defense of his hypothesis that Sturtevant ever put together.

The article is reprinted here verbatim from the course program of the 1939 Linguistic Institute, except that the bibliographical references in the footnotes have been changed to the form customary in Language.—The Editor.

The term semantic structure is one that is often encountered these days. Some very rewarding effort has gone into the study of phenomena embraced by this term, but there is still no clear picture of the relationship between semantic structure, whatever it is, and language structure as we know it. One view, and probably the most explicit, has been that ‘semology’ constitutes a third subdivision of language, on a par with phonology and grammar (or ‘morphology‘). I am going to present here a somewhat different view, based on a proposed parallelism between phonology and grammar and their respective ties to the nonlinguistic universe. If this view is found to have some validity, it may serve to help us in our understanding of the interrelationships of all three of the phenomena listed in the title above.

This paper begins with a question. It gives empirical data in order to show that the question appears answerable in principle. It discusses some problems that require an elaboration of the preliminary solution. Finally, it summarizes a few of the implications of the study.

The approach to language study represented by postulates and corollaries was once the basis of Bloomfieldian linguistics, yet it is now somewhat out of fashion. Sets of postulates have a tendency to grow in number and unwieldiness, and transformational investigators favor a more operational analysis and statement. Yet transformational study itself rests upon a postulate which is at present implicit. Stated without setting it into the Euclidian language and style favored by Bloomfield, it is that the relations between utterances are capable of being stated in the form of manipulations. Such an approach to utterances forces us to assume that all utterances contain items in combination, and that all items can appear in other utterances, which are then new combinations.

In the Festschrift für Paul Kretschmer 267–74 (1926), in an article entitled Gedanken über den lateinischen ā-Konjunktiv, N. S. Trubetzkoy proposed very convincingly the theory that the ā-subjunctive of Italic and Celtic was, in reality, originally an optative not a subjunctive. His thesis was that beside the athematic optative in jē/ī (Lat. siem, siēs, siet / sīmus, etc., Gk. eíēn/eîmen, Skt. syām, etc.), there were two distinct optatives for thematic indicatives, one in oi (Gk. phéroi, Skt. bharet, Goth, bairai, etc.), which is the one generally recognized, and another in ā found in Italic and Celtic to the exclusion of the oi optative.

It is a well known fact that the nominal system of Proto-Indo-European (as well as the pronominal and verbal systems) has to be reconstructed with three numeri—singular, plural, and dual. In many of the Indo-European languages the dual has disappeared as an independent formal category, so that the old threefold contrast is found to be replaced by a binary one.

Glottochronology makes certain general statements about historical processes in linguistics and sets up mathematical models which are declared to correspond to the processes. These models are presented in the form of algebraic equations. Numerical data are then collected from particular historical situations and put into the equations, whence numerical results are derived which are then converted into specific historical statements. It is the thesis of the present paper that the mathematical models which these equations represent do not correspond sufficiently to the general statements, that they are therefore inadequate models, and that the results obtained by using them are unreliable in proportion as the models are defective.

It is the purpose of this paper to present some of the results of a statistical comparison of the basic vocabularies of a collection of Malayopolynesian languages. By reason of the very number of its members, the Malayopolynesian family has great importance among the language families of the world. A fair estimate would place the number of Malayopolynesian languages at about 500. If the number of the languages in the world is set at about 4000, then about ⅛ of all the world's languages are Malayopolynesian.

It is now nearly twenty years since my paper ‘The Indo-European semivowels’ was published in Lg. 19.83–124 (1943). Some recent publications, to be cited in §§15 ff. below, have suggested to me that a brief restatement of what I was trying to do there is needed.

This is a continuation of two articles published several years ago in Language (30.197–201 [1954]; 31.4–8 [1955]). In these articles I was concerned above all with a reinterpretation of the interrelationship of the various Armenian reflexes of PIE p t k kw. My conclusions were that it was best to assume that the first result of the Proto-Armenian sound shift of Proto-Indo-European voiceless stops was a set of voiceless spirants, and that some allophones of these spirants later merged with other phonemes, while the principal allophones were affected by a further shift, as a result of which they became voiceless aspirated stops. I interpreted Greek evidence as indicating a rather late date for this second shift (about the beginning of the seventh century a.d.). This interpretation has been challenged by Hans Jensen in an article in Aus der byzantinistischen Arbeit der Deutschen Demokratischen Republik I (1957), and I now deem it possible that the date of the second shift has to be set considerably earlier than I had assumed. On the other hand, I do not find it necessary to revise my basic theory concerning the results of the first or Proto-Armenian sound shift ; only the postulation of Proto-Armenian spirants provides us with a consistent explanation of the seemingly erratic distribution of reflexes of Proto-Indo-European voiceless stops.

Among the types of weak verbs occurring in Germanic, none has provoked more difference of opinion than that of Class III, and no other has so consistently eluded a believable analysis. As might be expected, most discussions of the subject lean heavily on the evidence provided by Gothic. In Gothic weak verbs of Class III, the most noticeable feature is the occurrence of a in some inflectional forms beside ai in others. Thus the present indicative active of the verb ‘have’ has sg. 1 hab-a but 2 hab-ais and 3 hab-aiþ, pl. 1 hab-am and 3 hab-and but 2 hab-aiþ. Similarly, the imperative has sg. 2 hab-ai and pl. 2 -aiþ but pl. 3 -andau. On the other hand, a always appears in the infinitive (haban), in the present participle (habands), and throughout the passive indicative (e.g. pl. 1–3 habanda). Conversely, ai occurs in all forms of the preterit (e.g. ind. sg. 1 habaida, p.p. habaiþs) and in nearly all optative forms, the sole exception being pres. act. 1 habau.

The Arabic verbal stem is a complex of two interdigitated discontinuous morphemes, a consonantal root with lexical content and a vocalic pattern with the content of voice (passive vs. nonpassive), which may be augmented by one or two morphemes having roughly aspectual content, or which may be unaugmented. The verbal stem plus a verbal paradigm expressing the subject constitute the verb. If the paradigm is a set of gender suffixes alone, the form is an imperative; the imperative plus a set of personal prefixes constitutes the jussive, a command or conditional form; the jussive plus a modal morpheme forms the future, either an independent ‘indicative’ or a dependent ‘subjunctive’, depending on the modal morpheme; a variant of the jussive, the ‘energicus’, is formed by adding another modal suffix. Finally, the verbal stem with another set of endings, a fusion of personal and gender morphemes, constitutes the past tense. Another set of suffixes may be added to any of these finite forms to express the pronominal direct object.

Until recently I had taken symmetry in phonemic analysis as something axiomatic and self-evident, needing no discussion. But in 1960 two well-known structuralists found fault with my patterning of the phonemes of Icelandic: on the one hand, the Russian phonologist M. I. Steblin-Kamenskij accused me of doing violence to phonetic fact in order to achieve symmetry; on the other, the American phonemicist A. A. Hill accused me of needlessly taking phonetic facts in a way destructive of symmetrical patterning. According to my Russian critic I am set on symmetry at all costs; according to my American critic, ‘it is clear that for Malone a phonetic “fact” must clearly outweigh any possible symmetry of arrangement’ (249). Confronted with these opposing criticisms of my procedure, I can no longer let symmetry take care of itself; I am forced to weigh its worth in the patterning of phonemes. Hence this paper.

Maranao is a member of the Tagalog family of Philippine languages; these, together with the Visayan languages, are members in turn of the Malayopolynesian family. These languages are characterized by verbal inflection to indicate not only time, kind of action, and psychological point of view of the speaker, but also certain grammatical relations between the verb and the topic of the sentence. When a substantive phrase is not the topic, its grammatical relation to the verb is indicated either by particles or by certain pronoun forms. In some of these languages, position in the sentence signals such relations, but in Maranao position carries little functional load.

Zellig S. Harris distinguishes three situations in the ordering of morphemes relative to each other: ‘restricted’, ‘contrasting’, and ‘descriptively equivalent order’. Candidates for all three types are found in Modern German. It is the purpose of this paper to examine some problems of order in German within the framework of a transformational grammar.