A Reader in Nineteenth Century Historical Indo-European Linguistics

CHAPTER TWELVE

HEINRICH HÜBSCHMANN

ON THE POSITION OF ARMENIAN IN THE SPHERE
OF THE INDO-EUROPEAN LANGUAGES

"Ueber die Stellung des Armenischen im Kreise der
indogermanischen Sprachen," Zeitschrift für vergleichende
Sprachforschung auf dem Gebiete der Indogermanischen
Sprachen, 23.1 (1875), 5-49

Editor's Introduction

Hübschmann's is another of the articles published in 1875 which indicate the maturing of linguistics. Making use of the increased control over the data in the Indo-European languages, and over linguistic methodology, Hübschmann by this article established Armenian as an independent branch of the Indo-European group.

Hübschmann's minute attention to data enabled him to sort out the evidence for distinguishing between native Armenian forms and those borrowed from Iranian. In this way his is one of the last important articles to deal with a problem which vexed early Indo-European linguistics: identification of the various strata in a language so that its original relationships could be determined. By sorting out the non-native forms, and establishing the phonological correspondence between Germanic and the other Indo-European languages, Grimm, Rask and the early historical linguists laid down the methodological principles for their field. Using these principles in a more difficult area, Hübchmann at once demonstrated their validity and gave a definitive solution to the problem he was investigating.

It is clear why Hübschmann's solution was so satisfactory: after dealing with morphological characteristics he concentrated on phonological correlations rather than on the vocabulary. His procedure might still be emulated by linguists seeking to establish genetic interrelationships in other language groups.

Even with his accomplishment, Hübschmann's understanding of Indo-European phonology was not completely accurate. Although he was aware of secondary palatalization in Indo-Iranian, he still assumed a single short vowel a for Proto-Indo-European, with a split into the vowels of the European languages. The correct view became generally apparent shortly after his article was published. Other misconceptions in his article are obvious to the reader. But since Hübschmann's aim was to determine the relation of Armenian to the other dialects, his conception Of Proto-Indo-European was not crucial in achieving this aim.

Nor were some of his methodological views, such as those on residues. A few of these he dismisses as chance phenomena. The sections of his long article with such comments are not reproduced here. But it is to Hübschmann's credit that he recognized before his article was printed that he had slighted some of the material in Armenian. Subsequently he rectified any omission by his comprehensive Armenische Etymologie (Leipzig, 1897), the first part of his planned grammar of Armenian. Moreover, residues are still being explained in Armenian today, thanks to his solid work on its phonology.

Hübschmann's concern with thorough descriptive analysis enabled him to clarify other interrelationships, such as that of Ossetic as well as that of Afghan (which in his first excursus, pp. 43-46 [not included here], he demonstrated to be an Iranian and not an Indic language). The pioneering work that was necessary in Iranian as well as Armenian may be recognized from the preliminary note to his article, in which he deals with the transcription for Armenian. Though he modified it, he was unfortunately prevented by the editors of the Zeitschrift from revising the transcription for Iranian, which in the form he used suggests pronunciation like that of Sanskrit.

Besides his insistence on careful descriptive techniques, Hübscmann's conception of the interrelationships between the languages in one family was admirable. The family tree model as proposed by earlier linguists seemed far too rigid. His identification of shared characteristics in Armenian and the European languages as well as in Armenian and Indo-Iranian gave excellent support to the wave theory which had been proposed three years earlier by Schmidt. The resulting conception of the position of the early Indo-European dialects prepared for the more realistic view of interrelationship between languages which followed further studies in dialect geography.

The interest in broadening the study of the Indo-European languages at this time is strikingly illustrated by the editorial comment at the beginning of the twenty-third volume of the Zeitschrift, on its increased scope beyond Germanic, Greek and Latin. Hübschmann's own boldness concerning the Indo-European family may be demonstrated by the concluding comment in the second excursus to his article, pp. 46-49 (not included here). After suggesting that Phrygian may have been closely related to Armenian, he ends with the statement: "Possibly these languages formed a separate branch with other languages of Asia Minor, which in accordance with our contributions on Armenian above, might be placed between Iranian and Balto-Slavic. " The separate branch has indeed been uncovered, but with a position in the Indo-European family somewhat different from that which Hübschmann had forecast. His concentration on interpreting the Iranian and Armenian data permitted little further speculation of this sort. But the interpretations he provided of these data remain permanent contributions on the position of these languages.

Johann Heinrich Hübschmann (1848-1908), after a post-doctorate period of four years at Leipzig, spent his entire career at Strasburg. Scornful of academic jockeying, he rejected offers to move closer to the contemporary centers of linguistic research. Though he dealt with many of the Indo-European languages, his concern with Armenian extended beyond linguistics to Oriental studies. His primary achievement was in the elucidation of Armenian. Though it has never been one of the languages of central interest to Indo-Europeanists, Hübschmann holds a position of great respect not only as founder of scientific Armenian studies but also for his capable application of linguistic method.

My attempt to assign to the Armenian language its position among its relatives is not the first. The Armenians themselves have proposed views about it which flatter their national vanity but lack every scientific foundation. And European scholars of previous centuries have made everything of this language since they could do nothing with it. But immediately after the establishment of linguistics by Bopp, Petermann in his Grammatica linguae Armeniacae (Berlin, 1837), on the basis of etymologies given at the beginning of it was able to furnish the proof that Armenian is an Indo-European language. Nine years later, in 1846, and independently of the work of Petermann, Windischmann published in the Abhandlungen of the Bavarian Academy (IV, 2) an excellent treatise about Armenian, in which he comes to the conclusion that Armenian goes back to an older dialect which must have had great similarity with Avestan and Old Persian but to which foreign elements had been added early. But while Pott doubted that Armenian is an Aryan language and only wanted to admit a strong influence of Aryan on Armenian, Diefenbach on the other hand observed that this assumption did not suffice to explain the close relationship of Armenian to Indic and Persian, a view which Gosche also adopted in his dissertation: De Ariana linguae gentisque Armeniacae indole (Berlin, 1847). Three years later in the Zeitschrift der Deutschen Morgenländischen Gesellschaft IV, p. 347 ff., under the title "Vergleichung der armenischen consonanten mit denen des Sanskrit" de Lagarde gave a list of 283 Armenian words with their etymologies (which he also had found independently of Windischmann), without however dealing in greater detail about the character of the language. In the preface to the second edition of his Comparative Grammar, 1857, Bopp designated Armenian as Iranian and attempted, though without success, to explain its inflectional elements. Fr. Müller, who since 1861 had busied himself successfully with the etymological and grammatical explanation of Armenian in a series of treatises (Sitzungsberichte der Wiener Akademie), penetrated much more deeply in the essence of this language, which he explained as certainly Iranian. In general Patkanoff follows him in his summarizing treatise "Über die bildung der armenischen sprache," which was translated from Russian into French, Journal asiatique, XVI, Série 6, 1870, p. 126 ff. Even though de Lagarde in his Gesammelten Abhandlungen (1866), p. 291, asserted that three components are to be distinguished in Armenian: the original basis; an Old Iranian alluvium resting on it; and a similar New Iranian, added after the founding of the Parthian kingdom, nonetheless he did not give the distinguishing characteristics of these three layers, and for this reason his opinion has not been taken into further consideration. In any case Müller's view, that Armenian is Iranian, has not been disproved, and must be designated as the best established and the prevailing one at present.

The aim of the following is to investigate whether it is tenable.

It is a primary defect of Müller's investigations that he has not undertaken to demonstrate that the Armenian words which correspond phonologically with the corresponding Persian are not borrowed from Persian. If however the oldest Armenian that we know contains loanwords from Aramaic and Greek, we may expect that since the Armenians lived for centuries under the influence of the mightier and more cultured Persians, they also would have taken from their language no small number of words.¹" If this is admitted, then one can also suspect a great number of words to be borrowed; and if one has given way to this suspicion, then faith in the Iranian character of the language also disappears rapidly. And this suspicion can be very easily supported. In numerous Armenian compounds, for example, we find the word dast 'hand', while the usual word for 'hand' is dzer'n; now dast corresponds to Persian dasta, which in contrast with Av. zasta, Skt hasta is demonstrated to be specifically Persian through the sound change from z to d, and accordingly must be a loanword in Armenian. Accordingly also dastak, dastakert are foreign words, OP *dastaka, *dastakarta. It is further clear that Armenian regularly prefixes a or e to words with originally initial r: for this reason r'azm 'battle array, battle', r'ah 'way', r'ocik 'content'², which we also find in Persian, are loanwords from Persian, just as all words beginning with r' in Armenian are foreign words, cf. r'abbi 'master'. Further, since final h in Persian corresponds to original ç or dental, final h in Armenian to original s or tr (and original ç corresponds to Armenian s, original dental between vowels to Armenian dental or y), final h in Persian is accordingly etymologically different from that in Armenian;³ accordingly Arm. akah 'well-informed' = NP âgâh < âkâça, gah 'throne, seat' = NP gâh, Av. gâtu, zrah 'armor' = Av. zrâdha, NP zirih are loanwords from Persian. Further, if below we find the sound law that Skt j = Av. z = Arm. ts and accordingly Skt jan 'be born' = Av. zan = Arm. tsn (< tsin, tsen), then azat 'free' = Pers. âzâd, Av. âzâta, from the root zan, must be regarded as a foreign word. Similarly if it is demonstrated below that Skt han = Av. jan in accordance with the sound laws would have to be represented in Armenian by gan and is so represented, then Arm. zen- 'slaughter, offer' is suspected of being borrowed because it corresponds to Av. jan, NP zan-. If in the same way Skt aj in accordance with the sound laws is Av. az, Arm. ats, then gavazan 'stick' = Av. gavâz cannot be an original word -- it would have to be kovatsan -- and also not xarazan 'whip', instead of which išatsan would be expected. Finally, Skt. yaj = Av. yaz 'worship' in accordance with the sound laws would have to be lats or dzats in Armenian (Skt. j = Av. z = Arm. ts, see below; originally initial y becomes Arm. l or dz, z; where y is initial in Armenian, it is a newly added prefix, as can be easily demonstrated); the form however is yaz and accordingly it is borrowed. The same is true of yašt 'offering' = Av. yêsh'ti. Also to be considered as loanwords: dev = Av. daêva, instead of which tiv would be expected in accordance with the laws of the sound shift which are to be set up below; likewise, I am convinced, bag- 'god' = Av. bagha, and den⁴ 'religion' = Av. daêna, words which came to Armenia with the Zoroastrian religion. Likewise, without being able to furnish proof, I would also like to look on words like thošak = Pers. tôshah 'viaticum', ambox = Pers. anboh 'quantity', zĕndan = Pers. zindân 'jail' as having come from Persia to Armenia; but of words like dipak 'brocade' = Pers. dîbâh, Arab. dîbâj, crag 'candle' = Pers. cirâgh, Arab. sirâj, thuthak 'parrot' = Pers. tûtak, tûtî, kerpas 'silk' = NPers. kirpâs, Arab. kirbâs, Skt karpâsa, Gk. kárpasos etc. there can be no doubt that they are foreign material. If it has been so easy for me to separate as loanwords no small number of the words⁵ treated by Fr. Müller, how greatly would this number be increased if an expert like de Lagarde would undertake to separate the foreign elements from the entire Armenian lexicon? Possibly also two groups of these could be distinguished, an older and a younger, and in this way the two layers would be found which according to de Lagarde were deposited on the Armenian basis.

If now we have become suspicious of the lexicon, we may turn with greater confidence to the grammar; for in all living languages this is surely the palladium that a foreign influence cannot touch. How wild is the lexicon of Afghan and New Persian, or English, and how clearly does the grammar teach us that in the former we have Iranian at hand, in the latter Germanic! And we may expect to find clarification from the grammar much more readily in Armenian, because it displays a relatively rich inflection. For Armenian still has four cases of nouns distinguished by endings and five of pronouns; and in verbs, without considering the infinitive and participles, it distinguishes by means of inflection active and passive, indicative, subjunctive and imperative, present, imperfect, simple and compound aorist, and corresponding to these double futures. Since I must treat of the grammar here briefly, it may be permitted to adduce a paradigm for the inflection of the noun and the verb:

a) Noun:
Stem: mardo "human" (Gk. broto-), anwan "name" (= anman).

b) Verb: ger-el 'take captive'

m, s, y, mkh, ykh, n serve as primary verbal endings, and i, ir, r, akh, ikh, in as secondary.

If however one views the total structure of Armenian, it gives the impression of a language which has undergone great changes⁷, having lost much of the old material of stem and word formational elements; but it replaced what was lost by new inflectional elements. In this way the subjunctive turns out to be a new formation from the present stem and the subjunctive of the substantive verb: em (pronounced yem) = Lat. sum, iʒem = Lat. sim- accordingly gerem -- geriʒem, aḷam -- aḷaiʒem; similarly, the future is formed from the aorist stem and the aorist subjunctive, with little change of the coalescing components: gereʒ + iʒem = gereʒiʒ instead of gereʒiʒem, 2. p. gereʒ + iʒes = geresʒes instead of gereʒiʒes; and the imperfect similarly might be a new formation from the present stem and the imperfect of the substantive verb: em 'sum', êi 'eram' - gerem - gerêi, but Fr. Müller claims to find a formation with the suffix ya in the imperfect: berêi = berey -i,⁸ with reference to the a-class, which forms ayi not êi: aḷam - aḷayi. Moreover, the main factor in new linguistic formations, analogy, has of course been powerfully effective, just as it also essentially brought about the remodeling of the Old Armenian inflection to the New Armenian. For example, the passive marker is i; if it is added to present stems in -u, wi results (zenu-l, pass. zenwi-l, l-nu-l, pass. lnwi-l); and this wi, which of course was originally only the present marker of the passive of a very limited number of verbs, has become the general passive marker in modern Armenian; cf. NArm. kordzwil 'be done' = OArm. gortsil, act. gortsel. In this situation it is readily understandable why the elements of the Armenian inflection are still so obscure to us. I do not know how one is to explain the ʒ⁹ which forms the compound aorist and the ʒ in iʒem, etc. If one identified with them the s of the Indo-European aorist and the sy of syijn, the opt. of as, then Armenian could not be Iranian, for in Iranian s would have to be represented by h and sy by hy. Equally obscure are the secondary verbal endings; on the other hand the primary are clearer, among them m = mi, n = nti, y = ti, mkh = masi; accordingly ykh (= tkh) could go back to tasi, the original Indo-European form assumed by Schleicher, in contrast with which Sanskrit and Avestan show tha. But ykh is probably an analogical formation to mkh of the first person, and kh is to be regarded as added on later, so that y likewise goes back to tha or a similar form. The suffix of the 2. p. sg. s refutes the Iranian character of Armenian, since Iranian shows h rather than s; yet also Ossetic, certainly an Iranian language, has s in the same form, for which explanations must still be provided.

Among the case forming suffixes of the plural, ʒ too is unclear, kh probably goes back to as (or in accordance with Fr. Müller to âsas, Iran. âhah) s to ans; in the instrumental we have the instrumental marker of the singular, to which the plural marker kh was added. Among the suffixes of the singular, m of the dative-locative goes back to the pronominal -hmâi, hmi; the ê of the ablative prepares difficulties. Fr. Müller would like to derive it from âdha, a shape of the ablative suffix found occasionally in the Avesta; I would rather think of the adverbial suffix tas = Av. tô, if ê can really not be = at. The instrumental suffix b remains to be considered. While this suffix was formerly identified with the one suffix of the Indo-European instrumental bhi, recently Fr. Müller and I have attempted to see in it a new formation, to be sure for no other reason than that this suffix contradicted the Iranian character of Armenian which had been asserted by us. For like Aryan in general, Iranian too does not know the instrumental suffix bhi. Our conclusion was accordingly: because Armenian is Iranian, it may not have the instrumental suffix bhi. But suppose one should rather conclude: because Armenian has this suffix, it is not Iranian. Now in accordance with Armenian sound laws, b surely points to bhi; and an original anmanbhi, martabhi had to become Armenian anmanb, martob, subsequently anwamb, mardov, as the instrumental of anun, mard actually is attested. And since in its function as well the case with b is a pure instrumental, there can be no objection to the equation: Arm. b = IE bhi. Some scholars have claimed to find this suffix bhi in Greek, Germanic and Balto-Slavic. But φι could also be a reflex of the other suffixes compounded with bhi (bhiam, bhiams, bhiâms, bhis). In Germanic the instrumental in mi = bhi is actually not found.¹⁰ Accordingly it remains only in Balto-Slavic, where bhi is found as OCS mĭ, Lith. mi. Accordingly bhi as instrumental suffix of the singular can be assigned with certainty only to Armenian and Balto-Slavic.

Result: In the inflection of Armenian we cannot demonstrate any specific Iranian characteristics; on the contrary it differs in an important point with Aryan and agrees with Balto-Slavic.

Since the inflection does not give us enough information about the character of Armenian, we will turn to the phonology.

Part I

In order to decide whether or not Armenian is Iranian with reference to its sounds, the question must first be answered: what are the characteristic features of the sound system of Iranian in contrast with the other Indo-European languages?

They are as follows:

A.1. The dental s, when not protected by a directly preceding following consonant, consistently becomes h, and

2. correspondingly sv becomes hv,

3. but when i, u or ai, au precedes, it becomes sh. In the latter point Iranian agrees with Sanskrit (except for final position where Sanskrit preserves the s); but Slavic between vowels always develops instead of that sh the fricative ch (sluchŭ = sraosha). In the change of s to h on the other hand, Greek agrees with Iranian; but unlike Iranian, in Greek this change is not carried through consistently. Further, also in Celtic sv becomes hv, chw; cf. Cornish huir, Breton choar 'sister' = NPers. khvâhar; Welsh chwech (= sves) 'six'.

4. Iranian shows a disinclination for aspirates but an inclination for the formation of spirants, of which it is particularly fond of kh, gh, f and w. Yet the oldest Iranian dialects, those of the Gathas and of the Old Persian cuneiform inscriptions, do not yet know the voiced spirants (gh, dh, w, which are present in the usual Avestan);¹¹ and Ossetic has shifted the voiceless stops (k and t) to aspirates. Baloci too knows aspirates (see at the end of this discussion), but they probably arose through the influence of Indic.

5. In consonant clusters spirants arise from stops through the influence of following t, sh, r, v; accordingly original kt, pt, tt become kht, ft, st; khsh, kra, pra, tra become khra, fra, thra.

6. Notable is the lack of l in Old Iranian,¹² the shift of çv to sp, and in contrast with Indic the form of the preposition pati (= Skt prati), of the adverb us, uz (= Skt ut, ud, but similarly in Old Persian), words like gaosha with the meaning 'ear', etc.

B. The aspirated voiced stops are lacking in Iranian; through loss of aspiration they fell together with the voiced stops and like them often became spirants subsequently. Balto-Slavic also merged voiced stops and voiced aspirated stops.

C. The change of the original palato-velars k, g, gh to the palatals c, j and s, z must be considered a primary characteristic of Iranian. But Sanskrit shares in the formation of palatals from palato-velars, and Balto-Slavic in the change of the palato-velars to slit fricatives.

Accordingly every single one of these characteristics is found in other Indo-European languages, and only the occurrence of all of them makes up the character of an Iranian language. We ask now whether all these characteristics occur in Armenian.

A.1. Original s generally appears in Armenian as h; cf. hin 'old' = Av. hana, Lat. senex; mahik, diminutive of mah 'moon', = Skt mâsa, Av. mâoṅha; this h is lost for example in the root arb 'drink' = original sarbh (Lat. sorbeo, Lith. srebiù), evthn 'seven', Ossetic awd = original saptan. In inflection this h shows up as kh (now pronounced as an aspirate), just as also in Persian h is closed to the spirant kh. s has been maintained as s in amis 'month', mis 'meat', us 'shoulder', in which the maintenance of s is explained by an originally preceding n: amis developed from mens, mis from memsa and us from amsa. The s in the accusative plural may probably be explained similarly: mards (now pronounced mártŭs) developed from mardins = martans. Accordingly the maintenance of s in these cases would not contradict the Iranian sound law set up above, even though Avestan would also change s to h after n; cf. maiih = mans, aorist stem of the root man, = man + s. In one case to be sure (before original v) s seems to be maintained even contrary to the sound law: skesur 'mother-in-law', cf. Avestan qapra 'father-in-law', NPers. khusur; here v may first have changed to g¹³ and this to k after the s on the pattern of skund 'puppy' = çvan-. Windischmann, Grundlage des Armenischen, p. 20, already wanted to regard the colloquial by-form kesur as the original form, and derive k from kh = sv; s would then have been added inorganically. But this explanation does not seem probable to me.

2. sv becomes kh or v in Armenian, both probably having arisen from hv: khoir, now pronounced khuir, NPers. khvâhar, pronounced khâhar, originally svasar, and veʒ = sechs = Gk sweks, Welsh chwech.

3. Aryan sh = s after i, u and their various grades is found in dž = Av. duzh from dush, cf. dž-goh 'discontented' and in zguiš 'careful' = *uzgaosha, actually 'with pricked up ears' -- two genuine Iranian formations, of which the latter itself would prove the Iranian character of Armenian. And zguiš is so well established in Armenian that one cannot readily assume it to be borrowed. This sh shows up also in uš 'memory, reason' = Av. ushi 'reason'. Elsewhere this š may indeed have developed further to s, e.g. ls-el 'hear' = Av. srush in sraosha (Lith. klausà OCS sluchŭ), nist 'sitting' = nsit = niseda = nishadah, as in Ossetic where in ghos 'ear', ars 'bear', aχsawa 'night', ast 'eight' s is found instead of sh. Accordingly de Lagarde is probably right in deriving gusan 'singer, musician' from the root Skt ghush 'make noise, resound'. Yet this material is not sufficient in order to discuss this point adequately.

4. Armenian is fond of the (voiceless) aspirates, of which it possesses a complete set: kh, th, ph; but of the voiceless spirants it knows only χ. Yet of the voiceless spirants Afghan possesses only χ (kh), but not f, which is frequent in Ossetic and Persian.

5. χt = original kt is found in uχt 'vow, treaty', Av. ukhti, aχt 'suffering, sickness' = Av. akhti; and dstr 'daughter' (beside duχt = Pers. dukht) may also be derived from duχtr, when we find bast beside bayt, drast beside draχt ('garden, paradise' Pers. dirakht 'tree'). ft for original pt cannot be found in Armenian, since f is missing; yet Avestan too still has pt instead of ft: gerepta, haptan, but NPers. giriftah, haft. For the change of tt to st I do not find an example, but it seems certain because of Arm. azd 'information' = OPers. azdâ, Skt addhâ 'certainly'. Aryan ksh = Iran. khsh appears metathesized in Armenian as šχ: išχel 'govern' = Av. khshi; bašχel 'distribute' = Av. bakhsh; ašχarh 'land' = Av. khshathra. Iranian khra appears as Arm. χra in χrat 'theme' χratu 'admonition, counsel', Av. khratu, NPers. khirad; Iran. fr as Arm. hr: hra = pra, Av. fra; Iran. thr as Arm. rh: ašχarh 'land' = Av. khshathra.

6. l is not absent in Armenian; but it also occurs in all contemporary Iranian languages, so that the presence of l in Armenian would prove nothing of itself. But we will see later that Armenian is distinct from Iranian by the manner of occurrence of l ....

B. As far as the aspirated voiced stops are concerned, we might assume that the original Iranian language had already given up aspiration and merged the aspirated voiced stops with the voiced stops. Before this happened, Armenian must have separated from Iranian (if we set up a family tree); for in Armenian voiced stops and aspirated voiced stops do not fall together but rather have always been kept distinct. For while the aspirated voiced stops were shifted to voiced stops, the voiced stops developed to voiceless stops; the voiceless stops however remained unchanged or became aspirates or spirants. Accordingly the original series

undergo in Armenian a conversion to:

Examples are as follows:

a. Dental series:

Arm. d = original dh: Arm. d-ne-l = original dhâ 'set, do' ... (Other examples follow.)

Arm. t = original d: atamn (a-ta-mn) 'tooth' = dant. (Other examples follow.)

Original t was maintained when protected by neighboring consonants: astḷ 'star', dustr 'daughter'; or it developed to d: du (from túam) = NHG du; leard 'liver' = yakart; mard 'human being' = Gk. brotós; ôd 'wind' = vâta; or it developed to an aspirate: tharm 'fresh, young' = Skt taruṇa; tharšam 'wilted' (in an-tharšam 'not wilting', tharšameʒuʒanel 'wilt' trans.), root tars, Lat. torreo, tarsós; evethn 'seven' = saptan; uth 'eight' = ashtan; thandzr 'thick' root tañc; artsath 'silver' = Skt rajata; or between vowels it developed to y: hair (written hayr) 'father'; mair 'mother'; berê 'he bears' = bereti, etc.

b. Labial series:

Arm. b = original bh: band, bant 'prison', root bhandh... (Other examples follow.)

p is maintained as voiceless stop in kapel = capere; partkh (stem partu-) 'debt', Av. par (in pesha, peretha) 'involve in debt, forfeit through debt'; pšnul 'observe' = Skt paç; patmel 'narrate' pati + mâ, it was shifted to an aspirate¹⁴ in phoši 'dust' = Av. pãsnu; phetur = NHG Feder; phut 'foul' = Skt pûti 'foul, stinking', Phl. pûtak; and initially it went over to h in: hair 'father' = patar; hing 'five' = pankan; harʒanel 'question' = NPers. purs-îdan; heru 'last year' = Osset. fâre 'in the previous year', falwâre (= farfâre) 'in the second last year', Pers. pâr 'the past year'.

c. Palato-velar series: (H's term: Gutturalreihe)

Arm. g = original gh: gari 'barley' = hordeum, originally ghardha, Phl. jurdâk 'grain, barley', Baloci zurth-ânî 'a kind of grain'; mêg 'mist' = mêgha; vagr 'tiger' = Skt vyâghra.

Arm. k = original g: kov 'cow' = gâu; klanel 'devour', keri 'I ate', root gar; keal 'life', root giv; kin 'woman' = ganâ; kr'unk 'crane' = gíranos; eki 'I came', root ga, of which the present however is ga-m. gravel too does not agree with Skt grabh, Av. garb; yet the same irregular shift occurs in Goth. greipan. For further details, see below. (not included here) The voiceless stop was maintained as k in akn 'eye'; kam-il 'desire' Skt. kâma; kerp = Lat. corpus; kapel = Lat. capere; in final position it became g: erg 'song' = Skt arka; infrequently it became an aspirate: kharšel 'pull' = Av. karesh; khên 'hatred, revenge' = Av. kaêna, NPers. kîn; khandel 'destroy' (khand-el denominative ?) from Av. kan, Skt khan.

On the shift of another series (g¹ -- gh¹) see below.

This is the first sound shift of Armenian. The New Armenian of the west has undergone a second: the relationship of voiced and voiceless stops, as established after the first sound shift, is reversed, so that the original voiced aspirates are now voiceless Stops, the original voiced stops as well as a part of the original voiceless stops are now voiced stops, but the aspirates and h-sounds remained unshifted. In Armenian accordingly, voiced stops and asPirated voiced stops did not fall together as in Iranian.

C. The last point remains to be discussed, the development of spirants from original palato-velars. In this point Iranian and Balto-Slavic have much in common, so that Johannes Schmidt protested with this support against a separation of Iranian and Slavic and of Aryan and European in the early period; and he overthrew the family tree of the Indo-European languages which has been proposed up to now. For not only in the split of original k to k and k¹ = ç, s do Balto-Slavic and Aryan agree closely,¹⁵ but also in accordance with Ascoli's demonstration in that of g to g and g¹ = Iranian, Balto-Slavic ż, z and that of gh to gh and gh¹ = Iranian, Balto-Slavic ż, z. This knowledge however is not adequate for our following purpose, and in order to be able to compare the split of the palato-velars in Armenian with that in Aryan and Balto-Slavic we have to set up these series of splits completely, as I now do.

I. Split of g to g and g¹.

a. g appears in Sanskrit as g, in Avestan as g, Armenian k, Balto-Slavic g.

Note 1: Skt. gâ, gam 'go', Av. gâ in gâma, gâya, ga in gata, apagaiti, gam in ja-ghm-aṭ, aibî-gemen, Arm. eki 'I came', ek 'the stranger' (baínō, venio)... (Other examples follow.)

The g above we see developing to j in some examples; thus beside Skt. gam, Av. gâ even the root and present stem appear as jam, jim, jas, though the original g was maintained where it was protected by a consonant: jaghmaṭ ... Beside Skt. yuga : yuj we find Av. yuj. So we may also posit original g = Skt. j = Av. j = Arm. k, BaltoSlavic g; note Skt. rajas 'sphere of air, fog, darkness', = Arm. erek 'evening', Gk. Érebos, Goth. riqis. ... The complete g-series accordingly shows up as follows:

b. g¹ appears in Sanskrit as j, Avestan z, Arm. ts, Slav. z, Lith. z.

cf. Skt. aja, ajâ 'buck, goat', Av. azi, Arm. aits, Lith ożýs 'buck', aíks ... (Other examples follow.)

II. Split of the gh to gh and gh¹.

a. gh appears in Sanskrit as gh, Av. g, gh, Arm. g, Balto-Slavic g.

cf. Skt. megha = Av. maêgha 'cloud', Osset. miegha 'fog, cloud', Arm. mêg 'fog', Lith. miglá, OCS mĭgla 'fog, clouds'... (Other examples follow.)

Just as g occasionally became j in Sanskrit and j, zh in Avestan, so gh occasionally becomes h in Sanskrit, j, zh in Avestan, ž in Armenian.

Skt. druh 'vex', drogha 'insult', Av. druj, druzh 'lie, deceive' beside draogha 'deceitful', Arm. džr-el, drž-el 'deceive, miss, offend' ... (Other examples follow.)

Accordingly the gh-series shows up as:

b. gh¹ = Skt h, Av. z, Arm. z, dz, Slavic z, Lith ż.

Skt. aham 'I', Av. azem, Arm. es from ez), OCS azŭ, Lith aż (asz)... (Other examples follow.)

Some apparent anomalies must be noted here, from which the relation of g to g¹, gh to gh¹ becomes clear. We saw above that Skt. yuj = Av. yuj must go back to a root yug, the g of which must have been present in Armenian as palato-velar, as it is actually found that zuig = *yôga. Now we also find however lts-el 'hitch up in a yoke', which goes back to original yug¹ (which is not present in Sanskrit and Avestan); and accordingly we must posit for Armenian two roots, yug and yug¹, which of course were identical originally. Then the two g's are not originally different, but the one g has split in two, in part remaining g, in part becoming g¹. The same is true of Aryan g in the root gabh (and of forms with gh and k that H. cites). ... This can only mean: originally there was only one k, one g, one gh, which later split to k, k¹, g, g¹, gh, gh¹....

If one now compares the k series with the g and gh series, in accordance with the previous investigations:

		k	=	Skt	k,	Av.	k,	Arm.	k,	Balto-Slavic k
	/				|		|		|
k					c		c		?
	\
		k¹	=		ç		s		s (š, ʒ)	Slav. s, Lith. sz
		g	=		g		g		k,	g
	/				|		|		|
g					j		j, zh		k, ž
	\
		g¹	=		j		z		ts	Slav. z, Lith. ż
		gh	=		gh		g (gh)		g	g
	/				|		|		|
gh					h		j, zh		g, ž
	\
		gh¹	=		h		z		z, dz	Slav. z, Lith. ż

there is complete agreement between these series, from which it must be concluded that in the original period of the Aryans, Armenians, and Balto-Slavics speakers were in especially close contact with one another. For this common development of the palato-velars k, g, gh in two directions: to k, g, gh and k¹, g¹, gh¹ cannot be purely accidental -- or if it is, all characteristics of languages, by which we determine their conditions of relationship, must be purely accidental.

If we consider especially the relationship of Armenian to Aryan and Balto-Slavic, it turns out first of all that by its strict distinction of g (= k and ts) and gh (= g and z, dz) it is at an older stage than Balto-Slavic and Iranian, which as may be seen from the above tahle have merged both of these. This phenomenon agrees totally with the other phonological relationships of these languages. For if Sanskrit and Armenian in general maintain the distinction between voiced aspirated stops and simple voiced stops (gh-g, dh-d), which Iranian and Balto-Slavic have abandoned, then we must also expect that the two first-named languages retained the distinction between original gh¹ and g¹, and the last named language groups gave it up, i.e. they merged gh¹ and g¹ to g¹ and developed this to a spirant (Av. z = Slav. z = Lith. ż). On their part Sanskrit and Armenian are differentiated because Sanskrit, in contrast with Avestan and Balto-Slavic, merges part of the original g, gh with g¹, gh¹ (so that Sanskrit j = g and g¹, Sanskrit h = gh and gh¹); Armenian on the other hand not only continues the distinction of voiced stops and voiced aspirates, but also g, gh and g¹, gh¹, and accordingly in this respect it maintains the original phonological relationship more faithfully than Aryan and Balto-Slavic.

But we must also examine the relationship of Armenian to Aryan and Balto-Slavic in another and more important area. For the chief difference between the language families named above consists in this that Balto-Slavic at first maintains its palato-velars (g, k) unchanged, Aryan on the other hand changes them to palatals. For the Old Aryan sounds k, g, gh in part remained velars, in part also developed to the palatals: Skt c, j, h (h from jh) and Av. c, zh, j, primarily in three cases: 1. if i or y followed them originally, e.g jîv 'live'; 2. in the reduplicating syllable; 3. in root final position, when they were not protected by a following consonant, or when they were maintained unchanged before vowels in nominal derivations, as happened in part, e.g. pac 'cook', vac 'speak', but pâka and ukta. But palatalization has also taken place beyond these limits, if not widely, and in this situation k has been affected more frequetly, and g and gh less: cf. Aryan ca 'and', catvar 'four', car 'go, drive', pañcan 'five', Skt jaßhara 'belly', Skt han = Av. jan 'strike'. The agreement with which Indic and Iranian have carried out this process of palatalization provides certain proof that it took place already in the common Aryan period. And since it did not occur in this way in any other Indo-European group, this formation of palatals is particularly characteristic for Aryan. For that reason it must also serve as a test to determine whether Armenian is Iranian or not.

Now we have already noted (in a portion not given here) that: Arm. uiž baž, buž = Aryan aujas, bhaj, bhuj, possibly also žtel = Av jad, žir = Skt jîra, and držel, džrel = Skt druh, Av. druzh, iž = Av. azhi, aržani = Skt arh, Av. arej (NPers. arzân), and may because of these examples designate Armenian as Aryan. But only one thing is unclear: why do we find in uiž, baž, buž the sound ž corresponding to Aryan j, since in accordance with the sound shift we would expect c? No example at all has been found for Arm. j = Aryan jh and Arm. c = Aryan j; for this reason one must assume first of all that j and c arose only late in the separate existence of Armenian¹⁶ (accordingly ž would have arisen for j and c in Aryan times?). Arm. c = Pers. c is found frequently by the way; cf. cank, cang 'hook' = NPers. cang; caš 'eat' with its derivatives: cašel, cašak, cašakel 'taste' = NPers. cashîdan; capuk- = NPers. câbuk 'nimble'; carp 'fat' = Phl. carp; crag = NPers. cirâgh 'lamp'; vcar- 'solvere' = NPers. guzâr (from vicar); r'ocik 'support' = NPers. rûzeh; rûzî, Av. rocaṅh 'day', etc. But these examples prove too much; the agreement with Persian is here too great, for otherwise we find no trace of such agreement. And since words like crag and r'ocik are certainly borrowed (they are also found in Georgian), the above words other than these are probably also loanwords. The same is probably true of patmucan = Phl. patmucan 'dress', while the c of mucak = NPers. mûzah could only have arisen from the k of the underlying word muik 'shoe' = Phl. môk (Afghan moc-aṛ̃ah f. 'shoe') in Armenian.

But in accordance with the sound shift, the Aryan palatal c is found in Armenian as ch in: chorkh¹⁷ 'four' = catvar; gochel 'scream', kochel 'name' Lat. vocare; and it has become a dental in mrʒ-il 'battle, fight' Av. mereñc, nasalized from marc (for the meaning, cf. márnatai 'he fights' = Skt mṛrṇâti 'he crushes'), haʒ 'bread' = pac, thandzr (from thanʒr 'thick', Av. tañicish'ta), and finally in luis 'light' = Av. raocanh, NPers. rôz¹⁸. Is Armenian then Aryan?

Compare now the Armenian words eki 'I came', keal 'life', kov 'cow', kin 'woman', erek 'evening', bek 'broken', gan 'strike' with their Aryan related words as well as hing (from penkan) 'five' with Aryan pañcan and lkh-anel 'leave' with Aryan ric (erg 'song' = Skt arka, root arc and khan = Lat. quarn, khanak = quantum, Av. cvañt, NPers. cand do not come into consideration); it then turns out that precisely in those forms to which especially value is to be ascribed after Ascoli's splendid studies concerning the Aryan palatals, Armenian is decisively separated from Aryan and agrees with Balto-Slavic.

After the above remarks we can now complete the k-series set up above as follows:

and the parallelism with the g and gh series is now completely established.

From the whole preceding investigation we obtain as total result:

In accordance with its development of original palato-velars to spirants Armenian belongs to the sphere of the Aryan and Balto-Slavic languages. It agrees in part with the Aryan languages in the palatalization of the palato-velars, but in another area it also preserves palato-velars unchanged, like Balto-Slavic; for this reason it can neither be subordinated to Aryan (in the usual sense) nor be taken away from it. For this reason too it cannot be designated as Iranian, even though it like Iranian changes s to h and treats many consonants and groups of consonants (like Arm. šχ = Iran. χš = Aryan kš) in a similar or in the same way. For this reason it must be set up as an independent branch between Iranian and Balto-Slavic.

Part II.

If however Armenian stands between Iranian and Slavic, that is between Aryan and European, we must still examine its position to the special peculiarities of the European languages, through which these are sharply separated from the Aryan languages; that is, we must examine whether Armenian knows the European split of a to a and e, that of r to r and l, and whether or not it presents important points of contact in vocabulary with European. We will proceed at once to answer these questions.

1. Split of a to a and e.

Nothing is more correct than Fick's view that the most important difference between European and Aryan in vocalism consists in the split and non-split of a to a and e. It is certain that no Indic and no Iranian language knows this split. To be sure we write numerous e, i.e. short ä, in New Persian words, but short a simply becomes ä throughout (pronounced as pure a in India),¹⁹ and of a split there is no question here. Ossetic, in the Caucasus, has o and e for and beside a, but e is rare and obviously late; it appears beside ä and both beside a, which a stricter, and older, manner of speaking preserves; and also o = a seems to be only a later darkening of the a in the neighborhood of n and r,²⁰ but does not enter into consideration here at all. Accordingly Aryan does not know this split.

Armenian on the other hand splits the a completely to a, e, o in the European manner, and is thereby distinguished sharply from the Aryan related languages, even if it may not coincide in individual examples with the European. But in order to be able to undertake a comparison of the individual examples, we must first survey the vocalism of Armenian in general. The scheme for it is:

and the accentual law which controls the vocalism is as follows: the accent moved to the penultimate syllable of the word and caused the loss of the last, or the loss of its vowel, so that now the last syllable always has the accent. Short i and u as well as their lengthened forms ê and ui can be maintained only by the accent, that is, if they stand in the last syllable now; if however they lose the accent through the addition of a new syllable, then ê becomes î, ui becomes û, i and u on the other hand are lost. e.g. i: root vid 'find' = Arm. git, but in the infinitive gt-anél; original vinâç = NPers. gunâh = Arm. vnâs; lengthening: mêg 'fog', root migh; mêz 'urine', root migh¹, Skt mih, but in the genitive: mîgi, mîzi²². u: root yuj, Arm. luts, infinitive ltsel 'yoke'; lengthening: luis 'light' from raucah, root ruc, but in the genitive lûsóy. From this it is clear at once that gitém 'I know', lizem 'I lick' go back to gêt-em, lêz- em, that is, that in these verbs the lengthened present stem occurs (gêt-em = vêda + later added em = âmi, lêzem = leigh¹-âmi), except when we deal with denominative verbs, as may be true of mizel 'urinate' beside mêz 'urine'. On the other hand a, â cannot be changed: bazúm 'much' = Skt bahu; bazúk 'arm' = Skt bâhu; asél 'say', Skt. ah. The same is true of o,²³ cf. gochél 'cry', root vac; gortsél 'do' beside gorts 'work' = wérgon. e too generally remains, cf. mets 'great', Gk mégas, genitive: metsi; but it has dropped out occasionally, e.g. vtak 'rivulet' beside get 'river'; astḷ 'star' = aster; tagr 'brother-in-law' = dawer, genitive = astél, tagér. But if this e comes to stand before nasals and double consonants, it regularly goes over to i,²⁴ hing = quinque; hin 'alt [old]' = senex; sirt = Herz; and like original i, this i is elided when it loses the accent: hin 'old', hn-anal 'become old'; sirt 'heart', gen. srti. Accordingly e must be assumed everywhere in Armenian where instead of the a-vowel to be expected in accordance with the etymology, e, i or total loss of the vowel has occurred. Accordingly, if Armenian is to take part in the chief characteristic of European, we have to expect to find a) Arm. a (o) = European a (o) and b) Arm. e, i, zero = European e.

a. Arm. a (o) = European a (o). For this correspondence it is adequate to cite few examples: akn 'eye', Lith. akís, oculus; atsem = ágō; tal 'give' = dare; ail = állos; aits = aíks; hair = patḗr. gochel = vocare; chorkh 'four' = quatuor; gorts = wérgon; kov 'cow' = bow-ós; ordz 'testicle' = órkhis; orb 'orphan' = orphanós. In addition one should note for inflection that the a-stems -- apart from proper names -- went completely over to o-stems; for this reason original marta-, Gk broto- appears in Armenian as mardo-, cf. gen. dat. mardoy, instr. mardov, gen. pl. mardoʒ, instr. mardovkh. In this respect Armenian agrees with Latin and Greek.

b. Arm. e = European e. In his book, Die ehemalige spracheinheit der Indo-germanen Europas, p. 425, Fick listed the original European words to which e must be ascribed. Of these I find the following in Armenian: Arm. sirt = Eur. k¹erd 'heart'; tsnôt (= tsen-ôt) = genu 'chin'; inn (= inun = invan from envan = nevan) = nevan 'nine'; hing = penkan 'five'; mets = mega 'large'; mêj (from medyo) = medhia 'middle'; melr 'honey', melu 'bee' (mélissa) = melita 'honey'; nist (= nsit = ni-sedas) = Eur. sedas 'seat'; hin = sena 'old'; evthn = septan 'seven'; astḷ 'star' = ster; skesur 'mother-in-law' = svek¹ura, sved¹rû 'in-law'; veʒ = sveks 'six' .... For inflection e is important in nominal suffixes like ter = original tar, therefore dústr 'daughter', gen. dstér; in verbs as the stem-forming verb of the most widely distributed class: berem, beres, berê, etc.; as augment, to the extent it appears: eki 'I came', edi 'I set',²⁵ etu 'I gave', etes 'he saw', egit 'he found'....

2. Split of the r to r and l.

l is found in Armenian beside two r sounds (r' and r) and an ḷ, which is now pronounced by the western Armenians as γ, but in former days transcribes Gk. l. In European l also corresponds to the first l, while r is found in European corresponding to the r', r and ḷ.

a. l = European l: lal 'bewail, weep' root lâ... (Other examples follow.)

b. Arm. r, r', ḷ = European r:
sirt 'heart' = cor... (Other examples follow.)

3. There are various words which go back to a dffferent phonological structure in the European languages than in the Aryan.... (Of the twelve found in Armenian) the Aryan forms compare with the European in the proportion of 3 : 9 = 1 : 3.

4. The last point to be discussed here, the question about the relationship of the Armenian vocabulary to that of the European languages, I have to leave untouched for the time being, because the greater part of the Armenian words are still not yet etymologically clarified.... But in future studies numerous "European" words will be demonstrated to exist in Armenian.

Through the last part of our investigation, such a tight bond has without question been constructed between Armenian and European that it would be easier to tear Armenian from Aryan than from European. Among the European languages it stands closest to Balto-Slavic because of the spirants, with which it was also especially connected by the instrumental suffix bhi, which is common to only these two. In this situation, friends of the family tree, like Fick, will certainly be inclined to separate Armenian completely from Aryan and make it a purely European language. Against this view I might first refer to the fact that Armenian does not take part completely in the split of a and r. ...

The result of my entire investigation is accordingly as follows:

Armenian stands in the sphere of the Aryan-Balto-Slavic languages between Iranian and Balto-Slavic.

If further research makes this preliminary conclusion definitive, then the impossibility of setting up a family tree of the Indo-European languages would be strikingly demonstrated. For Armenian would be the connecting ring of both parts in the chain of the Aryan-Balto-Slavic languages, not a branch between two branches. And then too the family tree, which Johannes Schmidt's vigorous might has overturned, would remain lying forever.

But if Armenian is to be the connecting member between Iranian and Balto-Slavic, between Aryan and European, then in my opinion it must have played the role of an intermediary at a time when they were still very similar to one another, when the historical period had not yet drawn the present sharp boundary between them, but when they were still related to one another as dialects. Just as Upper Italian folk dialects with their nasal vowels, with ü instead of u and with other characteristics approach French, so that one might expect that Italian at the boundary of France is almost an intermediate between French and Italian, in the same way Armenian might once have been intermediate between the Aryan and the European dialects and therefore have taken part in the characteristics of both. For if we see that the archaic languages of the Avesta and the Veda still stand quite close to one another and that a reconstructed Iranian and an Indic Proto-language would be related to each other in the very same way as dialects, why shouldn't the European languages once have stood in this relationship to one another and Armenian as intermediate dialect between the two types? In this way the peculiar hybrid structure of Armenian can easily be explained. After the wave of the splitting of a and r had arisen in the Western dialects and that of the splitting of the palato-velars in the Eastern, the former penetrated beyond Balto-Slavic and spread further over Armenian, while the latter penetrated outside of Armenian and spread further over Balto-Slavic. The former changed, whether directly or through a subsequent effect, the a of the Indo-European numeral dakan to e in Balto-Slavic, and the latter changed the k of the same word to k¹, so that instead of the IE dakan the Balto-Slavic basic form dek¹an resulted. When later in Aryan the wave of the subsequent split of the palato-velars (to velars and palatals) arose, it was still able, whether it was of itself too weak or whether the dialects had already begun to separate more and more, to spread completely over Iranian, but only over Armenian in some offshoots, so that we find to be sure chorkh = catvar, but also hing = pañcan.

However that may be, if we must view the development of Skt c, j, h and Av. c, j from k, g, gh in the same forms as proof for the connection of the Aryan languages, then also the development in the table below must be viewed as proof for the connection of Aryan with Armenian and Balto-Slavic:

It is obvious that the minute difference between sounds in the individual languages proves nothing against this conclusion, for only the later phonological propensities²⁶ of the individual languages are responsible for their existence. For Gothic even confronts the European e with i; and the l of the individual languages which has arisen from European l is certainly not the same everywhere, as the Gk. l in Armenian transcription is always given by ḷ (etymologically = r, now = γ), never by l. If nonetheless we prove the original connection of the European languages with this e and l, we must also prove through those spirants the former connection of Aryan, Armenian and Balto-Slavic. What's sauce for the goose is sauce for the gander. And finally, as compelled as we are to derive the Aryan palatals from the pre-Aryan palato-velars, we are equally justified in deriving the series of the Aryan-Armenian-Balto-Slavic sounds k¹, g¹, gh¹, to which we take back those spirants, from the series of the Indo-European palato-velars. For a split of the palato-velars took place twice in a similar manner but in dffferent extent.

But if one might claim both series for the Indo-European original language, as Fick did with k and k¹, then our argumentation would be untenable and no closer relationship would be proved between Aryan, Armenian and Balto-Slavic; yet even so the common formation of the palatals in Aryan, in which Armenian participates, if only in part, would prove nothing for the closer connection of Indic-Iranian-Armenian:

But then, to be just and consistent, one ought also declare meaningless the few reasons for which Balto-Slavic and Germanic as well as Greek and Italic or Italo-Greek and Celtic have been generally assumed to be connected; for they are by no means of greater significance than those by which one can connect Aryan with Balto-Slavic or Armenian with both of these. Then we would arrive at a skeptical point of view, and would hold that European e and l too prove nothing: just as in the one group spirants would arise from palato-velars in certain cases without motivation, so in the other a would become e and r would become l in certain cases without motivation. Or if one assumes two basically different k for Indo-European, why not also two originally different a and r, which had fallen together in Aryan just as randomly as often happened in European with Fick's k and k¹? No one can claim that this point of view is nonsensical; only Fick will not be inclined to adopt it.

From this point of view there would be nothing further to say about Armenian than that it is an individual branch of Indo-European. And as such we will also have to view it from the other point of view, however its relationship to European and Iranian is to be conceived. Unfortunately -- and to this I'd like to point in conclusion -- the etymological investigation of Armenian is still in its beginning, and we are working with such a minute portion of the Armenian vocabulary that we cannot foresee what further investigations will uncover; and it was probably unjustified to erect at this time such bold constructions on such an uncertain basis as was done above. It is by far most important to separate the Iranian loanwords from Armenian and to arrive at pure Armenian material. Only when this has been done can one determine the more precise phonological characteristics of Armenian and thereupon loosen or tighten the bond that connects it with Iranian. But whether this bond is firm or slack, the close connection of Armenian to European remains undeniable, such as the formation of the instrumental singular with original bhi, the (partial) preservation of the palato-velars in contrast with the Aryan palatals, and the split of a to a and e, that of r to r and l demonstrate. For the time being there are not yet many deviations and agreements in the vocabulary to be given, since up to now neither Armenian nor Iranian has been investigated enough etymologically. Therefore the question of the lexical relationship of Armenian to Iranian and Slavic must still be viewed as quite open, just as we have intended to broach the question of the position of Armenian in the sphere of the Indo-European languages, not to have settled it decisively.

Notes

1. Also êi, êir, êr etc. would remain unexplained, while in this way we could assume that this imperfect of ah has gone over to the analogical influence of the other verbs; this is also true in part for the present, where ê 'is' is not explained from asti, but probably through analogy from berê 'he bears' = bereti. [return to text]
2. ʒ is emphatic ts, just as ch is emphatic tš (c). [return to text]
3. On the Old High German instrumental in u, see Braune, "Ueber die quantität der ahd. endsilben," p. 40. [return to text]
4. For further information see Spiegel, Grammar p. 345. [return to text]
5. That Old Iranian had no l is clear from the fact that the modern Iranian languages do not agree in the use of l; compare for example the following examples: Ossetic stal 'star', khalm 'crawling animal (snake)', nal 'man', malin 'die' with NPers. sitârah, kirim, nar, murdan; and conversely: NPers. gulû (gula 'throat') with Ossetic qur = ghur, Afghan ghâṛah, fem. 'throat'; NPers. talkh 'bitter' with Afghan trîχ, fem. tarχah, NPers. kulâgh = Baloci gurâgh 'crow' (Afghan kârgh*h 'crow' ?). For this reason, in spite of Oppert's objections, Revue de linguistique IV, p. 209), l will have to be denied for Old Iranian. [return to text]
6. This change is frequent: gail 'wolf' = European valka; get 'river' beside vtak, root vad; gin 'price' venum, original vasna; gini 'wine', vinum; gitel 'know', root vid; gtanel 'find', root vid, vind; gortsel 'work', Av. verez, werg-; tagr 'brother-in-law', dawer-; gochel 'cry' = vac, garun 'spring' = Av. vahra, gier 'night' = vesper, gar'n 'lamb' = warḗn, warnós. Old Persian too causes v to change to g, cf. gurg 'wolf' = Arm. gail; but otherwise in different forms than in Armenian: gul 'rose' = Arm. vard, gunâh 'pass' = vnas, etc. Both languages have carried out this change quite independently of each other. [return to text]
7. Found also in loanwords: phartham 'rich', Phl. fratum; phuriršišs 'process' = Pers. pursish; phiḷ 'elephant' = Pers. pîl. In loanwords however which had initial f rather than p, Armenian substituted the similar h, since it had no f: hrasac = Arab. farsac, Pers. farsang 'parasang'; hraman 'order' = Pers. farmân; hretak 'envoy, angel' = Pers. firishtah. The Kurds too have changed their f to h, but maintained it in loanwords, e.g. in firman (Justi, Die kurdischen Spiranten, p. 15). The Ossetes on the other hand change initial p consistently to f: farsun 'ask', fondz 'five', fathan 'broad', so that p is initial still only in loanwords, while Afghan substitutes v for f and uses f only in Arabic and Persian loanwords. [return to text]
8. The series are: original k = Skt k, Iran. k, Arm. k (kh), Balto-Slavic k; and k¹ = Skt ç, Iran s, Arm. s, Slav. s, Lith sz. Armenian agrees with Iranian and Slavic, cf. tasn 'ten'. Occasionally however š is found instead of s, as in šun 'dog' = Skt çvan, pš-nul 'observe', Skt paç, as in NPers. shâkh 'twig' instead of sâkh = çâkhâ, shustan 'purify' = çudh, an indication that the sharp s of Iranian stands very close to the sh = Skt ç. And when for that reason sz is found in Lithuanian as opposed to Slav. s = k¹, and Indic ç, now pronounced sh, as opposed to Iranian s = k', we will consider this difference irrelevant, with Johannes Schmidt against Fick. From the sole Aryan-Balto-Slavic çvan, çuni arose only late the various Skt çvan, gen. çunas, and çuni, Av. sûnô, sûni, Arm. šun, gen. šan, Old Prussian sunis, Lith. szů, gen. szùns. But whatever is valid for Slavic s = Lith. sz must also be valid for Slavic z = Lith. z, original g¹ and gh¹. [return to text]
9. Through secondary palatalization in the separate existence of Armenian there arose: jerm 'warm' = Skt gharma, NPers. garm and šeram 'silkworm' = Skt kṛmi, Phl. kirm. [return to text]
10. The secondary form khar' is related to chor- as is Av. tûirya to Av. cathware, tûirya = tvar-ya and khar' = tvar; kh = tv as in khsan 'twenty' = dvi-çanti. [return to text]
11. Add to this: Arm. ch = Skt ch = original sk in the present stem forming sk = Gk sk: Arm. can-ach-em = gignṓskō, Aor. tsaneay, and in chu 'walking' = original sku, Ascoli, Vorles. p. 189. [return to text]
12. Accordingly in this essay I have written instead of ä, as it is now pronounced in Persia itself, the older a, from which it developed. [return to text]
13. See barzond 'high', zarond 'old' = geront, zond, zund 'knowledge', zônun 'know', fondz 'five', dzorun 'speak', χor 'sun' (= svar), corun 'eat' (= svar-). How little o means here is shown by kharôn = Av. karana, which forms in the plural: Tagauric kharatthä, Digoric kharanthä. [return to text]
14. I should like to note that the quantity of vowels is not marked in the Armenian writing system; therefore elsewhere as well I have not indicated a long mark. But this is only a shortcoming of the writing system, with reference to which it must be noted that a, i, u, where they are lengthened forms of the basic vowels a, i, u, must have formerly counted as long or still do. Only e and o are always short. [return to text]
15. I write here î, as in the following û; for from mĭgí, mĭzí: mgí, mzí would have had to develop, and similarly from lŭsoy: lsóy. [return to text]
16. Petermann, Grammar p. 37: "omnium vocalium constantissima, quae fere nunquam abjicitur seu mutatur." (The most constant of all vowels, which almost never is lost or changed.) [return to text]
17. This change of e to i has been carried through completely in the modern dialect of Tiflis, in which o has also become u throughout. In older Armenian u from o = a is found more infrequently: a sure example is probably the suffix forming the decades: -sun, gen. -sni, e.g. innsun, gen. innŭsni '90' = enenḗkonta. [return to text]
18. The present is dnel = d-ne-l, which according to the sound laws must go back to de-ne-l. If one compares this with the present tal 'give', then the equation results: Arm. de 'set': ta 'give' = Gk the: do. [return to text]
19. Consider how s in Sanskrit, where it was to be voiced, develops to everything, only not z, and how jh, instead of which h shows up, is almost prohibited. It is therefore not remarkable that we do not find in Sanskrit z, to which it was completely opposed, instead of the original g¹. Notable is the preference of Iranian, Armenian and Slavic (in their oldest form) for the dental spirants over against the palatals of Sanskrit and the linguals of Lithuanian; but nothing is proved by this about the closer relationship of these languages to one another. [return to text]