In 1928 the Polish anthropologist Jan Czekanowski published in the ethnographical quarterly Lud1 a study of the Indo-European languages in which he employed the method of differential diagnosis by quantitative correlation determinations which he had long been using with success in physical anthropology and ethnography. This method, whatever its field, rests upon the recognition of isolable and definable features or traits, which we shall hereafter refer to as elements, whose presence or absence can be determined for a number of populational groups or territorial entities, such as races, tribes, cultures, castes, or, in the present study, languages. The distribution of these is tabulated in terms of plus for presence in a particular group, minus for absence, and the question mark for unknown. Then each group is compared with each of the other groups in terms of the four-cell segregation familiar to statisticians. That is to say, four values are determined: a represents the number of elements common to both groups, b the number present in the first but absent in the second, c the number absent in the first but present in the second, and d the number absent in both. In other words, a and d are agreements, positive and negative respectively; b and c are disagreements. These four values are then substituted in a suitable formula, and a coefficient of similarity between the two groups results. When the coefficients for each pair of the groups being considered are assembled, we get a classification of the relative degrees of similarity between the populational groups or territorial entities which, being objective, has genetic and historical significance. For example, to make this clear in linguistic terms, suppose for the four languages Baltic, Slavic, Indic, and Iranian, we get a high coefficient between Baltic and Slavic, and again a high coefficient between Indic and Iranian, but low coefficients for the four other possible pairs, Baltic-Indic, Baltic-Iranian, Slavic-Indic, and Slavic-Iranian. The coefficients thus make the four languages fall into two classes, Baltic-Slavic, and Indic-Iranian, and it is evident that each class has had a certain history common to its members but not shared by members of the other class.

Is phonemic procedure arbitrary? It can be, but I submit that it need not be. If one defines a phoneme as ‘one of an exhaustive list of classes of sounds in a language’1, one admits an endless variety of treatments and the choice of one or another treatment is arbitrary. The ideal of exhaustiveness is not in itself sufficient to define a unique scientific procedure: But if we also take the ideals of simplicity and self-consistency, we have the basis for a non-arbitrary method, particularly in view of the essentially systematic nature of phonetic totalities. The aim of a phonemic analysis may then be stated as: the smallest number of elements having the largest number of permutations and the most systematic relationships. To attain this ideal, it is necessary to consider always the totality2 of phenomena in the given language. A formulation as to any detail must be made on the basis of all pertinent and comparable data anywhere in the totality of the given phonetic system. In the present paper I propose to demonstrate this method in application to a selected type of phenomenon, long consonants, in different phonetic totalities.3

Professor Collinson's study of Indication is the third installment of a series of language studies entitled Foundations of Language, logical and psychological, an approach to the international language problem. That series is, in its turn, part of a more inclusive program of linguistic investigations, a further part of which comprises a series of Comparative Studies of certain ethnic and constructed languages.

The Achewa are found principally in the Kasungu, Dowa, and Lilongwe Districts of Nyasaland Protectorate, British Central Africa. Their territory extends eastward as far as Lake Nyasa and westward into North-Eastern Rhodesia. Ethnically, they are a division of the Nyanja group, which has its center in the Shire Valley and Highlands of the Southern Province and in the Central Province southwest of Lake Nyasa, extending as far north as the middle of the lake.

The Nuzi dialect of Akkadian is known from several thousand cuneiform tablets discovered in the ruins of ancient Nuzi, about 10 miles southwest of modern Kirkūk, Iraq. These documents, dating from the middle of the second millennium B. C., were unearthed in the course of several campaigns sponsored by the American School of Oriental Research in Baghdad, the Semitic Museum of Harvard University, and the Iraq Museum in Baghdad. Following the publication of a small number of these records which had found their way to the British Museum and the Louvre, Dr. Edward Chiera began systematic excavations at Nuzi in the spring of 1925; other campaigns were conducted from 1927 to 1931, and were directed successively by Chiera, Pfeiffer, and Starr.

This dissertation on the Additional Short Syllables in Ovid was suggested by a statement made by Dr. John F. Gummere in his dissertation, that a very convenient way to get a short syllable in place of a long one was to substitute the plural for the singular in the nominative and accusative cases of neuter nouns. He refers to the figures of R. B. Steele, who has shown that fifteen Latin poets have an average of 45 per cent of dactyls in the first four feet of their hexameter, that Ovid has 54.6 per cent of dactyls in his Metamorphoses, and that Vergil has 44.1 per cent of dactyls in his Aeneid. It will be seen that Ovid gained over Vergil 84 short syllables per 100 lines. In Chapter IV it will be seen that in five books of the Metamorphoses and three books of the Aeneid Ovid gained over Vergil 84.6 short syllables per 100 lines. This figure will therefore be used throughout this dissertation. Vergil's percentage has been chosen for comparison with Ovid's, first, because it is so near the average, and second, because Dr. Gummere's figures for the neuter plural in Vergil are available.

Sturtevant in HG 144, speaking of relatively transparent nouncompounds in Hittite, goes on to say ‘One may be almost equally sure that tayukas (ta-a-i-ú-ga-aš, da-a-i-ú-ga-aš, ta-a-ú-ga-aš) ‘two years old’ beside yukas (i-ú-ga-aš) ‘one year old’ is a compound whose prior member is akin to tan (ta-a-an, da-a-an) ‘iterum, second’; but our ignorance of the etymology of the second member makes its classification impossible'. This article is an attempt to explain the word yukas ‘yearling’.

Despite all that has been written on the subject, it is still an open question whether the prevailing types of strong preterites of class VII in NGmc. and WGmc., as Ole. hét, hlióp, helt, fekk, lét, blét, etc., have a common origin with Gothic reduplicated preterites of the types haíháit, áiáuk, haíhald, faífāh, lailōt, haíhōp, etc., or descend from IE unreduplicated formations, in particular lengthened-grade aorists. In recent decades this problem has come to be involved with another one, namely the origin of the pt. pl. stem in classes I-V of the strong verb, above all in classes IV and V for Germanic as a whole, and probably most scholars who accept unreduplicated formations as the source of the forms in the category first mentioned do so for the forms in the latter categories also. It is not proposed in this paper to restate the arguments that have been put forward in support of this view,1 or in refutation thereof,2 but rather, tentatively assuming its general correctness, to reconstruct in chronological order the various developments which the theory seems to require if it is to be brought into detailed correspondence with the known NGmc. and WGmc. forms which it purports generally to explain.

§1. General Remarks. In all its aspects except the behaviour of the pitch-accent, the phonetic system of the Chewa language is a fairly simple one. There are only five vowels, and the consonantal system is highly symmetrical.

In this paper∗ the writers will attempt to depict in their mutual relations the several types of verb-forms which evolve into the so-called medio-passive, deponent, etc. categories in the historical IE languages as these types seem to have existed in late IE times; and to indicate subsequent developments (e.g., growth of some types, decay of others) that appear to have occurred in the prehistory of the continuant languages. It is our purpose to confine ourselves more or less to descriptive morphology, avoiding, on the one hand, inquiry into the ultimate origin (non-finite verbal, substantival, etc.) of the forms here discussed; and, on the other, as far as possible, entanglement with semantic and syntactic dogma of every sort.

§6. Introductory. In morphology, the Chewa language conforms quite rigidly to the general Bantu pattern. It illustrates the typically synthetic language of agglutinative technique; i. e., the word is fairly elaborate in structure, with affixed elements that are loosely united to its nucleus. With the exception of the two vocalic suffixes of the verb form,1 there is hardly a single element that may be regarded as fused. The system of noun classes and concordances serving as relational elements may be said to stamp the Bantu family as a ‘simple mixed relational’ type; i. e., the relational elements may be classified as 'concrete relational· elements.2 We shall see that certain of the noun classes are more formal than semantic in application—at least in Chichewa.3 Another feature (of Chichewa at least) is the identification, very much as in English, of the relation of actor and action with that of subject and predicate.4