La valeur emphatique de l'indo-européen *kwid est attestée par cid dans le Rgveda, par čit̠ de l'Avesta, par čiy du vieux perse, par oὔκι d'Homère et d'Hérodote, par cid en vieil irlandais. La nuance indéfinie de ce petit mot à l'origine, restée visible en présence d'un interrogatif ou dans le cas du redoublement, s'est ailleurs modifiée, passant de ‘en quelque chose’ à ‘un peu’, d'où par litote ‘plutôt, bien sûr, même, en particulier’ (Bartholomae, Ai. Wb. 593 'und zwar'). Dans RV IV.3.4 Macdonell lui attribue un sens atténué, Geldner au contraire met le mot en valeur: tváṃ cin naḥ śámyā (pada śámyai) bodhi... svādhḥ 'merke du wenigstens auf diesen Opferdienst von mir'. Et pour I.10.9 ν̄ cid dadhiṣva me gíraḥ, Macdonell traduit: ‘Even now take to thyself my songs’, là où rien ne légitime le sens de ‘maintenant encore’; mais Geldner écrit: ‘Nimm doch ja meine Lobreden an’.

Par rapport à l'usage du sanskrit classique (très variable d'ailleurs suivant les genres et les époques), le Rgveda (RV) ne présente qu'un nombre relativement limité de composés. Laissant à part les formations nombreuses, mais peu caractéristiques, où le membre antérieur est un invariant—préverbe, préposition, préfixe, adverbe ou nom adverbialisé—et qui n'entrent pas dans la définition stricte de ‘composé nominal’, on ne trouve en moyenne que cinq composés pour 150 mots (ou: pour 80 mots pleins), à peine davantage dans les portions récentes de la Samhitā. La proportion varie suivant le style: les passages descriptifs (ainsi 2.21) ou emphatiques en présentent davantage, les hymnes simples (comme l'hymne à Pūsan 6.54 ou le dialogue 'affectif Yama-Yamī 10.10) en ont fort peu.

The problem. Sanskrit noun stems ending in u, when subjected to secondary derivation, typically replace u by av, whereas, typically, stems ending in i simply drop the i before a suffix beginning with a vowel or with y (rather than replacing the i by ay or e respectively). I have not found any attempt to explain this disparate treatment, which, to judge from the implication or the express statement of our standard reference works, prevails equally in the Rig-Veda and in all later stages of Sanskrit. To point out, as the grammars do, that in this regard i-stems behave like stems in a, ā, and ī is not to explain the disparity but to heighten it.

There is a group of asseverative particles, usually postpositive, serving normally to stress the preceding word, which seem to be allied: Hittite -ma, Sanskrit sma smā, and Greek μά.

La comparaison de la flexion pronominale hittite avec celle des langues apparentées fait ressortir immédiatement deux traits distinctifs: d'une part le hittite se conforme au modèle indo-européen en attribuant des flexions différentes aux noms et aux pronoms; de l'autre, la flexion pronominale n'a en hittite à peu près rien de commun avec le système admis comme indo-européen. Cette situation mérite examen, et l'objet de ces pages est même de montrer que certaines correspondances tenues pour certaines sont fallacieuses, et que c'est en soulignant ces différences qu'on prépare une interprétation meilleure des faits hittites aussi bien que de la préhistoire des pronoms en indo-européen.

Among the Anatolian proper names transmitted to us in the Old Assyrian tablets from Cappadocia, those in -uman and -l(i)ka, in and -aššu, and finally those in -iyat/iet are particularly characteristic. Their analysis, undertaken elsewhere, has shown that they contain the following identifiable theophorous elements: Aššiyat, Ḫalki, Ilaliya, Inar(a), Išput, Perwa, Tarawa, and possibly Šiwat. The question arises whether these gods can be ascribed to a definite ethnic and linguistic layer of early Anatolia, such as is known to us from the Boğazköy texts.

[The Proto-Tocharian imperfect, based on IE optative formations, is preserved in Kuchean (Toch. B) but lost in Turfanian (Toch. A) except for two relics, the imperfects of the verbs ‘be’ and ‘go’. The remaining Turfanian imperfects are in origin identical with preterit formations found in Turfanian or in Kuchean or in both, which are all derived from IE perfects and aorists.]

Given the following groups of corresponding phonemes or phoneme sequences in Sanskrit, Greek, and Germanic: (1) v/ϝ/w, r/ρ/r, n/v/n, (2) u/v/u, r/ρa~aρ/ur, a/α/un, and (3) uv/v(ϝ)/(uw), ir(ur)/αρ/ur, an/av/un: the segments of sound lying behind group 1 were once in complementary distribution both with those behind group 2 and with those behind group 3. For the older stages of Indo-European, therefore, one needs to reconstruct only three entities: w, r, n (and likewise y, l, m for other sets not listed here), each with three positional variants of varying syllabicity. The argument has been most fully presented by Franklin Edgerton. In connection with that argument there has been some discussion whether the syllabic element in group 3 above should or should not be written ъ. Assuming for the moment that ъ has to be reconstructed in the neighborhood of stops to account for quattuor, πíσυρ∈s, and the like, the dilemma resembles that posed by the [k] of length in the description of some types of English : if we are impressed by its similarity to other [k]-like segments in environments where it is not automatically present, we say that there are some environments in which [k] is automatic but is phonemically /k/ nevertheless; but if we are more impressed by the automatic nature of its presence in length, we seize upon any phonetic detail that may distinguish this [k] from other [k]-like sounds elsewhere and see in the sequence [] merely the allophone of // before /θ/.2a If ъ is not recognized in quattuor, the problem hardly arises in the case of the nasals and liquids. For (i)y and (u)w it still exists, in a somewhat different form; but this question is here left out of account.

It is an honor to be invited to contribute to a number of Language dedicated to Franklin Edgerton, and it is a pleasure for me to accept the invitation. For that purpose I have selected three notes dealing with the language of the Iliad, which are held together by the fact that all three end with problems of the sort that Edgerton has dealt with so successfully in his reconstruction of the Panchatantra and in his edition of the Sabhāparvan of the Mahābhārata—problems of textual criticism.

An index may or may not be a trustworthy shift for finding something a reader remembers having read without recalling just where he read it. No index is a substitute for reading. Accordingly the epigraphic items discussed in this paper were found by reading the entire contents of CIL 13 (in seven parts and two supplements, 1899-1943). The texts are selected for their linguistic interest.

Except for the o-stem class, the formation of the IE genitive singular shows a clear and regular development from PIE -os/-es for all noun classes, in Germanic as well as in the other IE languages: Goth. sunaus, gumins, fadrs, nahts; ON sonar, hana, (nom. nǣtt) ; OHG fridō, hanen or hanin, fater, naht; OS suno, gumen; OE suna, guman, fæder, gēs (nom. gōs); Lat. hominis, patris, noctis; Gk. πατρός, κuvóς, etc. The gen. sing. of IE o-stems, however, is attested in forms whose explanation is problematical. This case is built in a number of markedly different ways in the IE languages: for Greek and Indic the indicated ending is the archaic -sjo (e.g. Horn. Gk. ΐπποιo, Skt. asvasya from ); for Illyrian Messapic it is -ihi (e.g. blatθihi ‘Blossii’; cf. the gen. sing, of consonant stems in -as, e.g. platoras); for Latin and Keltic it is -ī (e.g. Lat. servī, Ir. maicc = older maqi from nom. macc 'son'); while in Lithuanian, Lettish, and Slavic, the ablative singular has replaced the genitive (e.g. Lith. diẽvo, Lett. dìeva, Slav, boga 'of God'). In Germanic there is a form that has not yet been satisfactorily accounted for, namely Goth. dagis, ON dags, OE dœges, OS dages, OHG tages, etc. Because the IE languages differ so widely in the form of the gen. sing, of o-stems, most scholars tend to believe that no common proto-form for all the IE languages can be reconstructed.

Although modern scholars are becoming gradually less reluctant to regard the inscription on helmet B of the Negau trove as a solved riddle, I would propose a more cautious attitude. The only result to be expected from a new discussion of this enigmatic object is clarification, with a warning not to make use of its text for any far-reaching theories in historical, runological, mythological, or linguistic research.

It has been customary in Romance philology to study and analyze texts in terms of their orthographical tradition, even where—as in the case of the famous Oaths of Strassburg of 842—their phonology has been the subject of intensive discussion and interpretation. At the present stage of linguistics, however, this is not sufficient; since every written document represents an act of speech, it is our job to determine as closely as possible exactly what we assume that act of speech to have been, and to analyze its systematic aspects just as we would those of living informants, as far as our evidence permits, making use of historical and comparative linguistic data for synchronic interpretation. This analysis we should then present in consistent structural terms, to make it most usable in further historical and comparative work. It is our intention to do so here for the Oaths of Strassburg, presenting (1) each of the Oaths in its traditional orthography, with a translation; (2) a discussion of the phonemic system that we assume for the Oaths on the basis of descriptive and historical criteria; (3) a tentative phonemic transcription of the Oaths; and (4) a discussion of the implications of this re-interpretation for our classification of the language of the Oaths.

It is a matter of frequent occurrence that words etymologically identical appear in varying forms in a given language. The pairs person: parson, royal : regal, shirt: skirt are familiar examples in English. In the first and third there is a clear difference of meaning, while in the second the distinction is rather one of style. The source of these doublets is different in each case. In the first a sound-change which has resulted in the modern standard English pronunciation of clerk, Berkeley, Hertford affected likewise person. But the change was not completed. The changed form, now written parson, has been retained only for the incumbent of a parish, while the common form person serves the other meanings. In the second we have an instance of learned borrowing, while the third presents Scandinavian and English forms side by side, with a notable differentiation of meaning. The first of these three types is of peculiar interest, for the language, in such cases, has added to its vocabulary from within its own resources, not by means of suffix or composition, but out of the working of phonology. It is as though the speakers saw an opportunity, and availed themselves of it.

The purpose of this article is twofold: to examine the various patterns of declension found in Russian and establish the morphemes in them; and, by this presentation, to exemplify a possible method of morphemic analysis.

[Proto-Dravidian had a class of kinship nouns which occurred only in the possessed construction (inalienably possessed) ; this probably was a syntactic rather than a morphological construction. The personal and reflexive pronouns which occurred as attributes in this construction were only the plural ones; distinction of number in the possessor was not indicated, and could be gathered only from the context. The evidence is drawn from Old Tamil, Kota, Gondi, Kolami, Kuwi, and Kurukh.]

Among the Burmese versions of the Pāli Jātaka there is one, the nga:ya. nga:ze ‘the five hundred and fifty’, which is a faithful rendition of the original and which retains the gathas in a Burmanized Pāli. These gathas are accompanied by Burmese commentaries, which present points of linguistic interest both in the Pāli and in the Burmese.

1. Proto-Malayo-Polynesian *R was reconstructed by Dempwolff to account for the correspondence of Tagalog g with Toba-Batak r and the lack of a corresponding phoneme in Javanese (i.e. loss of a consonant and contraction of the abutting vowels); thus, *beRas, Tg. bigás, Jv. wòs, TBt. boras ‘polished rice’. Certain contradictions to Dempwolff's formula appeared not only in the Javanese cognates, but in those of Ngaju-Dayak and Merina, a dialect of Malagasy. The present article will examine these contradictions, to see whether they suggest hitherto unrecognized PMP distinctions.

Possibly the most widely disseminated fact about the Chinese language is that each syllable has a definite musical accent or tone forming an inalienable part. This fact emerges from even the most casual inspection of a Chinese dictionary, in which the syllabic symbols are each labelled with one or more tones, and where it is obvious that the intelligibility of a particular syllable depends on its having the correct tone. Persons who try to talk standard Chinese have to be told that a large number of these tones disappear in connected speech, which only goes to show the universal sloppiness of people. A pedantic speaker, who is at the same time literate, can restore the tones by visualizing the graphs with which the speech would be written down. But it must be concluded that for illiterates the intelligibility of their communications cannot depend on a theoretical knowledge of the tones for syllables that have no tones. An examination of Tangsic brings to light in that Wu dialect a situation rather different from what is ordinarily imagined. (For a definition and general description of the Tangsic dialect see Lg. 28.457-8.)

The Miao-Yao languages are spoken by tribes scattered among the mountainous regions of southwestern China (in the provinces of Hunan, Kueichow, Kuangtung, Kuanghsi, Yünna, Szuchuan), in Viet Nam (in the area of Tonkin), and in Thailand. These tribes bear various Chinese names, with either Miao or Yao as their last element. On the basis of the information that I now have, I have tentatively classified this group of dialects into two main divisions: the first preserves the syllable-finals -p, -t, (k), -m, the second has lost them. The dialects of the second division in turn form two subdivisions: one displays a highly complicated system of initials, the other a comparatively simple one. In this paper ten dialects are discussed: MKC (Hungchan Miao in Kaop'o, Kueichu, Kueichow), MKS (Yi Miao in Kechengchai, Kuangshun, Kueichow), YLP (Hei Yao in Yaolu, Lipo, Kueichow), YYT (Tahua Yao in Hsishankai, Yungts'ung, Kueichow), MTK (Hei Miao in Shihtungk'ou, Taikung, Keuichow), MCF (Hei Miao in Chengfeng, Keuichow), MJC (Hei Miao in Kaot'ungchai, Jungchiang, Kueichow), YYL (Yao in Yaoling, in the districts of Lienchow, Lienshan, and Yangshan in Kuangtung), YHN (Ch'ing-yi Yao in Haining, Tonkin, Viet Nam), YTP (Tapan Yao in Tonkin, Viet Nam). Most of these dialects belong to the second main division; MKC, MKS, and YLP differ from MTK, MCF, MJC, and YYT by having nasal-prefixed initials (stops and affricates) and initial consonant clusters with -l- or -r- as the second element. The dialects of the first main division are YYL, YHN, and YTP; they are cited only to show the distinctive finals -p, -t, (-k), -m. The many seemingly irreconcilable irregularities in the phonology of the three Yao dialects YYL, YHN, and YTP may be due to the inaccuracy of the records made by Wong and Savina.