By Oswald Szemerenyi, London L Old Persian £asuma- 'fourth'
The opening lines in the fifth column of Darius' great Behistun in¬
scription have given rise to varied interpretations. The main problem
presents itself at the end of the second and the beginning of the third
lines where the year must have been mentioned to which the events
related in this column refer. The text seems to be
2 .... pa tiy: a
3 mc: s Ordm
and it is obvious that the regnal year must be looked for in the remnants
between pati and Oardam. Hinz pointed out {ZDMG 93, 1939, 372) that
since Egypt is mentioned in Col. II 1.7 among the countries that rebelled
against Darius, the king would not have omitted to relate the re-conquest
of this province in Col. V, had it been accomplished by the time it was
engraved. On these grounds Hinz accepted Weissbach's restoration
pa"camäm (for s ) since Wiedemann's Geschichte Ägyptens dated
the re-conquest of Egj^it by Darius in 517 B. C. But in the latest,
detailed discussion of the passage (JNES 2, 1943, 109f.), Kent drew
attention to a statement of the Demotic Chronicle according to which
Egypt must have been under renewed Persian control already in Darius'
fourth year, i. e. 518 B. C. Moreover, he emphasised that the first letter
of the word containing the numeral was clearly s which precludes the
possibility of Weissbach's restoration. Kent rejected, therefore,
pa"camäm and proposed the reading pati avämöi sitäm Oardam 'during
that third year'.
Now sitäm is a very unlikely form to assume for Old Persian, or any
other Iranian dialect, for that matter. Kent refers to the Avestan name
drita- and MPers. dit 'second' which should attest the existence of Iran,
""flnto-'third' and *dwita-'second' . These forms — of which drita- is a
mere possibility, nothing more — cannot, however, invalidate the fact
that the ordinal is in all Iranian dialects Grittya- (cf. Av. dritya-, MP. sidtg
from sitiyaka-, MParth. hrdyg) in full agreement with Skt. tftiya-.
Moreover, by a lucky chance, we are on quite firm ground in this case :
sittyam, the neuter accusative used as adverb, is attested in Col. II
1.43, and thus shows that the Old Persian form was the same as that
presupposed by the other Iranian dialects and the later development of
Persian itself, not that reconstructed by Kent for this damaged passage.
Now it is admittedly impossible to squeeze into the existing gap the
form sitiyäm. The only possible conclusion is, then, that the third year
can safely be ruled out. Since dwitiyäm does not fit either, the only
historically admissible alternative is to try the year four.
If, with this in mind, we make a fresh approach to the text, it will be
noted that before the ordinal in question the rock has ^a. .I^mc", usuahy, and I think rightly, restored as avämc". But the end of this word remains
rather puzzling. Hinz interpreted as a defective writing for -ci and
Kent followed suit. This assumption, though not impossible, yet remains
rather doubtful, and I think we can find a better solution. In fact, I
submit that the passage is to be restored as follows:
pati avam casumäm Oaradam 'during that fourth year'^. In other words,
the stone mason inadvertently took c of the complex csumam (= cosm-
mäm) for the conjunction ca and attached it accordingly to the preceding
word, even inserting the word-divider after it. That such errors occur, is
well-known to ah Students of the textual criticism of inscriptions. In the
present case, it will suffice to point to daiy.mnrda in 1.11 of the column
under consideration which is generallj' accepted as an erroneous spelling
for viyamarda.
It will be admitted, I believe, that this restoration fits the historical
context excellently. The same can be said of the restoration itself. Kent
rightly insisted that in filling the gap, we must bear in mind that ac¬
cording to King-Thompson (The sculptures and inscription of Darius the
Oreat on the rock of Behistun in Persia, 1907, 79-) "there is room for about
five characters". It will be seen that our restoration is in cuneiform as
follows: c"«" [Mwi^am":], i. e. exactly five characters.
But here the question arises if the form supposed by us is admissible
from the point of view of philology.
The ordinal 'fourth' has not as yet been met with in Old Persian
inscriptions. However, the new data brought to light in the Turfan
documents, furnish the forms which continue the Old Persian casuma-
here supposed. Salemann had already recognized (Man. St. I 128) that
tswm 'fourth' was, in stem, the Persian form corresponding to Av. cadru- ;
he had also correctly interpreted tsb'y ,Vierfüßler' as deriving from
*ca6ru-päda-. The form cwhrwm, the vocalism of which puzled him ^ibid.
120)^, can now safely be interpreted as cohrom with umlaut from (Parth-
* For Oaradam, instead of the usual reading Oardam, see my paper in the
JAOS, vol. 70, 1950, p. 234.
2 And not only him, cf. Bartholomae, WZKM 30, 1917, 16^ and
Benveniste, JA 217, 1930, 207 f.
ian) cahrum < caßrwma-. Since then MPers. tskyrh ,viergestaltig' has been
recovered (Mir. Man. I 47).
On these grounds, the assumption of an OPers. ordinal casuma- from
Olran. *caOruma- could be regarded as well-founded. Fortunately, the
newly discovered Elamite inscriptions, published by Cameron {Per¬
sepolis Treasury Tablets, Chicago 1948), supply the last missing link.
There, p. 38, we find the following forms: sa-ää-maä '(one-) fourth',
sa-iS-Su-mas 'id.' As against these forms, attested in 18 texts altogether,
there are also found, each in one text, the forms: sa-iä-Su-iS-maS, sa-iS-
Su-iS-ma,, sa-Su-iS-ma. Since Elamite äi-is-maS (36 texts) and Si-äu-mas
(2 texts) 'third' obviously reflect an OPers. *siäva-, parahel to Av.
driSva- 'third', Cameron has been led to trace the Elamite forms sa-aä-
maS etc. to an OPers. *casuSva- corresponding to Av. cadruäva- 'fourth'. '
As, however, the most frequent Elamite forms sa-aS-mas and sa-is-Su-maS
can hardly be harmonized with this assumption (Cameron says: "The
above forms give evidence of scribal difficulties in rendering an OP word
which is assumed to be *casu-äva-"), I am of opinion that Cameron was
misled by tbe Avestan form cadruSva- and find further support for this
contention in the following Elamite forms: äS-du-maS 'one eighth', nu-
ma-u-maS 'one ninth'. Here Cameron, too, recognizes that the Elamite
forms reflect (the unattested) OPers. forms *aStania- and *navama-,
parallel to Av. aStdma- and naoma-, nauma-. Now, there can be no doubt
in this case that the Elamite names of fraction are nothing else than the
OPers. ordinals, a method widely used over the whole Indo-European
territory and also in other language groups. It would appear, therefore,
that there is no alternative but to regard Elamite sa-aS-masjsa-iS-äu-maS
as reflecting OPers. casuma- 'fourth' which is continued in MPers. tasum.,
while sa-{iS)-Su-is-ma{ä) are unsuccessful attempts at rendering the same
form.
2. Sarmatian marha
In a famous passage, Ammianus Marcellinus preserved for posterity
the battle-cry of the Sarmatians: marha marha quod est apud eos signum
bellicum (XIX 11.10), a counterpart to, and probably inspired by,
^ I. Gershevitch has suggested recently (AsiaMaior, New Series, vol. II, 1,
1951, 133') that the most common speUings of the Elamite word for "one
fourth" pointed to OPers. *6aiuva-, "derived withhaplology from *casuSuva- ;
this would seem to indicate that s sounded very much like S even to Persian
ears." The latter conclusion is, however, unjustified since the two sounds
were never mixed up in the entire history of Persian. As tho the question
of haplology, we should realize that the OPers. spelling conceals a pronunci¬
ation *&asuSva- which would hardly be susceptible of haplology in any case.
Tacitus' recording the barritus of the Germans. After being interpreted
by various scholars as Iran, marka- 'death', the word has more recently
been dealt with by Benveniste (JA 221,1932,135—8) who reached the
conclusion that it represented Iran, marta- 'man'.
In offering this solution, Benveniste was led by considerations of two
orders. First, he found a Persian correspondence in the phrase mard u
mard preserved in Arabic chroniclers; secondly, he referred to a feature
of Middle Iranian dialectology according to which Olr. -rt- appears as
-rd- in the South West but as -hr- and -hl/lh- respectively in the North
West and North East. Surprisingly enough, he inferred from this dis¬
tribution that Ammianus' marha represented not the North Western
mahr but the North Eastern malh 'avec une adaptation de 1 en r due ä
Ammien ou aux copistes' (1. c. 138).
This assumption seems rather unwarranted. If Ammianus or his mid¬
dlemen heard mnlha, there was nothing that could conceivably have
induced them to transform this into marha. Further, the dialectological
foundation on which Benveniste based his dialectal attribution, is not
borne out by the facts. He contends that Olr. -rt- becomes -rd- in the
South West but -hr/hl- in the North West/East. Unfortunately, this
theory is at variance with the dialectal forms of the very word marta-
'man' and there can be little doubt that one or two years later, after the
publication of Mitteliranische Manichaica I — III, he would have refrained
from advancing such an opinion. Here we find — what is common
knowledge today — that the North West had the same form for 'man' as
the South West, namely mard ; this is also true of the North East for which
cf. Sogd. Chr. mrt, Buddh. mrt, Yaghn. märti. It is not insignificant that
for the alleged NW/NE sound change Benveniste could only quote the
names of the primeval twins, Olr. Martya and Martyäni, the dialectal
forms of which had been restored by Schaeder (Studien zum antiken
Synkretismus, 1926, p. 226 note) as SW murdyagjmurdyärmg, NW
mahrehjmahryänih and NE malhijmalhyänä. But herewith we touch on
a vexed problem of Iranian philology the latest treatment of which is
due to the late E. Herzfeld (Ap. I. 249—251).
As is well-known, our Avestan text exhibits numerous spellings with
the letter usually transcribed s for the original cluster -rt-:asa- •—- arta-
etc. Andreas discovered that s represented a forgotten ligature of -whr- ;
most philologists seem to be agreed today on -hr- rather. Yet there can
be no doubt that, e. g., kdsa- 'done' was never pronounced as kahra but
successively as k^ta- > kirt > kird. As Hbrzfeld says: ,,e3 gibt kein Bei¬
spiel für den Lautwandel rt > hr und kein Wort, wo das s-Zeichen für
die Gruppe rt zu recht stünde, außer den 5, schon von Hübschmann
aufgeführten, rehgiösen Wörtern: 1. asa- - 'rta- und asavan- > ahrav;
2. amdsa. spanta- > amuhraspand; 3. asdmaoga- > ahramök; 4:. fravasi-
> fravardin, fravahr; 5. masya- > mahrya-, Berüni malM. Die Schrei¬
bung desselben gath. asa- mit im Komp. ßtarta- wäre allein für die
Unechtheit des s beweisend. Bei allen 5 Wörtern ist ursprüngliches rt
und Wandel in hr gesichert und bei allen vocal. /• wahrscheinlich".
Hebzfeld then goes on to say that a soimd change rt>ä is unknown to
both Persian and Median', while some traces of a change rt> hrjhl can
be discerned in what he calls Middle Parthian. He thus concludes that
the five words enumerated above, in which a real change of rt to hr can
be ascertained, reflect the Parthian of the 2nd century B. C, taken over
into the Avesta as fixed by Volagases around 50 A.D.
Whatever be one's opinion regarding the positive part of Hebzfeld's
theory, one fact emerges clearly from the picture as presented by him:
a soimd-change rt>hr is unknown to Iranian; for the limited number of
cases where it is found, a special explanation must be sought after.
Before applying, however, this result to the question of Sarmatian marha,
I may be allowed to try and give a new explanation of the problematic
'5 words'.
To my mind, Hebzfeld's one or two unconvincing examples cannot
establish an otherwise unheard-of Parthian sound change rt>hrjhl and
so the question is stih open. In my opinion, Hübschmann {Pers. Stud.
195) was right in laying stress on the fact that all five words are specific
Parsee terms ; this, however, is in itself no explanation since the core of
the problem, shifted on to the Parsee community, remains untouched.
Though fully aware of the difficulties involved, I venture to submit the
following suggestions.
From the philological point of view, it seems plain that if the Parsees
established these forms they could do so only because on the one hand
these forms were based on a real Iranian development, on the other they
served the useful purpose of evading some 'homonymies fächeuses'
scandalizing the religious mind.
For the first statement, a case in point is supplied by the names
Mahrl Mahriyäna. As is well-known, Andbeas referred already in 1911
to the cmious fact that the Turfan fragment T III 260 gives the names
of the primeval parents as gyhmwrd and mwrdy'ng, i.e. Gehmurd and
Murdiyanag (see Andeeas-Wackeenagel, Die vierte Gatha, GON 1911,
p. 3 note, and now Andbeas-Henning, Mir. Man. 1, 1932, 48). Obviously
the forms (Geh) Murd and Murdiyänag constitute an independent tra-
^ Hebzfeld does not mention Junkeb's paper in Ungarische Jahrbücher
5, 1925, 49ff. ; if this is to be taken as a silent criticism, I am inclined to agree with it since Junkeb's , mittelafghanisch ä' can hardly be harmonized with the rest of our knowledge.
dition, probably prior to the Parsee forms. Now while these forms are in
full agreement with the expected Iranian development rt>rd, they also
give a hint as to the cause of the Parsee deviation. In fact, Murd{iyänag)
can only be derived from Olr. mft-. But if this is true, then the Parsee
forms must continue Olr. mrdya- and mrdyäni, i. e. derivatives of mrt-
in which the original cluster -rty- regularly became -rQy- and later rhy-
>rhlhr. In other words, the Parsee tradition does not require an otherwise
unknown phonetic development; what it reflects is simply the Parsees'
tendency to divorce their 'nomina sacra' from everyday language, a
tendency which turned to good account the existence of normally develop¬
ed but not workaday terms.
It would seem that the same reasoning could apply to the case of
amahraspand and fravahr. The former should be traced to Olr. amj-Oya-
spanta- ; the latter, though continuing Olr. fra-vftiS, will be influenced by a derivative /ra?;fÖ2/a-, cf. MPT. fravahren ,ätherisch''.
As to the other principle, the tendency to avoid unwanted homonymies
seems to have been at work in the case of Pahi. ahlav. Henning (TPS
1942, 502) rightly pointed out that in contrast to Zor. Pahi. ahlav, Man.
'hlw, the genuine linguistic forms were OPers. artävan (= Skt. ftävan-)
giving MPers. and Parth. ardäv, Pähl, ardä and Sogd. artäv. I think that
in this case the Parsees were actually confused and embarrassed by the
existence of the word ahl<arh< Olr. *ar-da- = Skt. artha-, meaning
'advantage, possession'. This assumption implies that the living language
developed Olr. *ar6a- along different lines from Av. ardda- petrified in the
Yasnas and Yasts ; but the assumption seems to find support in NPers.
xer, usually derived from Olr. ardiya- (see Nyberg, Hilfsbuch des Pehl.
II 106). The Parsees thus visualized the asa of their religion as 'the
advantageous possession', just as they conceived of 'paradise' as the
supreme good or the best life.
For Pahi. ahramög = Av. ahmaoya- I can offer no solution if the
second member is really to be connected with Skt. mogha-. However,
if we are justified in taking it as a Middle Iranian form with -y- repre¬
senting Olr. -k-, then the obvious analysis is ahr-amök '(he who professes
the) evil doctrine', with ahr of AhramanyuS and amok (MPers. hmwg =
Tokh. amok, cf. Hansen, ZDMG 1940,1431'.; Henning, Sogdical) from Olr.
ham-oka- 'training, teaching' for which see Bailey TPS 1936,98—101^.
1 For fravrliS, however, the parallel case of NPers. puhl 'bridge' from
OP. prtu- should be borne in mind which also generalized the spirant
6 legitimate in some case forms only, as e.g. the Gen. Sg. prOva.
^ I cannot accept Hansen's suggestion, 1. c, (found also — from the
formal point of view — ap. Ghilain 63) that hamok etc. derive from a root
*mauk- = Lith. mokyti 'teach', as the latter necessarily continues IE
Returning to the battle-cry of the Sarmatians, we can then safely
conclude that there is no possibility of deriving it from Olr. marta-. This
conclusion receives further support from the fact that Ammianus' words,
if not in so many words, exclude an interpretation in Benveniste's
sense, according to whom they represent the accepted Iranian phrase
for a challenge between leading personalities of two opposing armies,
singling out one of the enemies and inviting him to a duel. As is weh
known, the battle-cry is preserved in a passage describing the moment
when Emperor Constantins was about to deliver an address to the
barbarians foregathered on his invitation before the tribunal; suddenly
and most unexpectedly, one of the Sarmatians "furore percitus truci,
calceo suo in tribunal contorto, marha marha exclamavit, eumque secuta
incondita multitude, vexiho elato repente barbarico, ululans ferum, in
ipsum principem ferebatur". It is clear from this description that the
leader of the Sarmatians could not possibly have in mind a challenge
like that supposed by Benveniste since his battle-cry actually prompted
the whole mass of his compatriots to rush on the Romans — without the
least sign of chivalry.
On the other hand, it would appear that the meaning of 'death' as¬
sumed by former scholars, fits the situation described by Ammianus
excehently. However, all we know about early Iranian dialects forbids
us to regard marha as a possible dialectal development of Olr. marka- :
mark or marg are the only likely continuations'.
Now Armenian has a word marh, mah 'death' which was listed by
Hübschmann (AG 472) as belonging to the genuine Armenian vocabu¬
lary. Against this. Bailey (BSOS 6, 1930, 62; cf. also Benveniste, 1. c,
135^) pointed out that the Armenian word, an w-stem, was borrowed from
a Middle Iranian form continuing Olr. mrOyuS 'death', attested in OPers.
hvä-mfäyuä and recognized by Henning also in Sogd. my^ry 'death'
{BSOAS 11, 1945, 484^). Bailey also pointed out that there must have
been a North Iranian dialect in which Olr. f was continued as -ar-. Indeed,
Armenian loanwords like vard 'rose', marg 'bird' demand such a develop¬
ment which is actually attested in Ossetic, cf. mary 'bird' (1. c. 60 f.)-.
*mäk-; but I am glad to see that he, too, derives Av. aigmaoya- from
ahramauka-.
It will be noted that in two cases at least, namely amahraspand and
Mahrth, Olr. r is continued as -ar-, not, as is the Persian rule, as -ur-;
for this I refer to my discussion of Arm. mah in the text further on.
1 I cannot, therefore, agree with Altheim's view adhering to this deriva¬
tion. Lit. und Gesellschaft I, 1948, 268*^ with Nachtragp. 330; 11,1950, 141.
" The same seems to be true, as Bailey observed, of Sogd. mry which is
never spelt mwry. To Arm. öanaparh 'path' discussed by Bailey, I would
add Arm. pailipan , Beschützer' as compared with NPers. puStbän .Stütze,
To sum up. From the evidence adduced above we can safely conclude
that in North West Iranian there existed a word marh 'death'. There can
be no doubt that it is this word that has been handed down to posterity
by Ammianus. The Sarmatian battle-cry thus reveals itself as the Iranian
nomads' bloodthirsty and murderous shout: 'death, death'.
It also foUows that this word does not support Benveniste's con¬
clusion that Sarmatian belonged to the North Eastern group of Iranian.
3. Sogd. wtSnyy 'old'
Sogd. wtSnyy, wtSnyh, apparently first attested in M 14 R1 9 and V 27,
was rather hesitatingly translated by Lentz (Waldschmidt-Lentz,
Manichäische Dogmatik 547 f. and 571) as 'old'. With more material at
his disposal, Henning {Bet- und Beichtbuch 62) proved this to be a lucky
hit. He added, however, that the meaning of the word was not ,alt an
Jahren', but ,abgenutzt, vergangen, früher'. In my opinion, this state¬
ment is not borne out by the facts. Indeed, Henning referred to f9 and
38 as proving his contention. Now these texts read as follows : f 8 (BBB 49)
'ty cn Ss' (9) zng'n 'kt'nyy 'ty cn wtSnyh (10) ynd'k 'kty' 'pstwyy „und
sich von den zehnfachen Sünden und den früheren Handlungen lossagt" ;
f 38 (BBB 50) 'rty ms wtSnyy Syrkty' (39) cn wny rw'nyy wjxstyyh ,,ja,
sogar seine früheren guten Handlungen werden von seiner Seele ab¬
getrennt".
However, these passages, together with M 801, 516 (BBB 33): nwyy
pxw'q wtSnyh „ein neues Stück oder ein altes", obviously do not admit
of any closer interpretation than that established for M 14; in other
words, the only meaning found in these texts as well as in the few more
adduced by Henning, BBB 62, is 'old' as wih be admitted by anyone
familiar with the biblical phrase^ of the old man and the new man etc.
Now the word also occurs with the spelling wtcnyy and it would seem
that in suggesting an etymology, Henning was influenced by this spel¬
ling. In fact, Henning submits as etymon (BBB 62) „vitacina- (vgl. av.
Beschützer' where Hübschmann {AO I 221) tried to explain the difference
in the root vowel by the assumption of an Iranian alternation parätij
priti, unknown otherwise; it is now clear that the Armenian loanword
simply continues an Iranian dialectal form where r became ar while a
Persian form puätepän was in its turn borrowed into Armenian as
p'uMipan.
It may be noted here that Oss. arc 'Bayonnet, Lanze, Pike' (Milleb-
Freimann, Wb. I 48) obviously derives from Olr. rSti- 'spear'. This
equation, hesitatingly put forward already by Miller, Ossetisch 40, in
the Grundriß, is of some interest as here again Arm. aStejaSteay 'spear' is
borrowed from a dialect where Olr. rMi- became a{r)&ti-, in contrast to
Pers. {h)iH.
taciria-) zu vi-\-tac ,laufen, fließen', vgl. auch av. vitäöina- ; man darf wohl
ruhig einem *vitacina- die Bedeutung ,verflossen' geben, wenn auch sonst
vi-^tac meist nur .schmelzen' heißt". I for one, still feel rather uneasy
about the semantic development involved and propose to give a dif¬
ferent solution.
It is quite natural for the Iranian scholar that, given the consonantal
skeleton of a word as wtSny, he should look immediately for the preverb
vi- as hidden behind w-. Nevertheless, the meaning of the word being
established as 'old', it is quite as plain to the Indo-European scholar that
wtSny should be read as wat{u)sane, i. e. a precious relict in Iranian of
the well-known Indo-European word *wetuso- 'old', from *wet- 'year',
attested in OSlav. vetixi 'old', OLithu. vetusas 'old, aged, stricken in
years' and also Lat. vetus, veteris^. This adjective became watuSa- in
Iranian. In Sogdian, the form was enlarged by the suffix -ne, giving
watusane and, by syncope, watSane, phonetically wacane; it is at this
stage, represented in our texts, that the word can be spelt either wtSny or
wtcny both intending the same pronunciation.
4. NPers. pöyidan
NPers. pöyidan 'run, trot' was explained by Fe. Müller (WZKM 9,
1895, 379) as a denominative from pöy 'haste' continuing Olr. *pauda-
and the latter equated with Greek ctttouSy). For the relation spud: pud
he referred to the analogous one existing between Goth, stut: Skt. tud,
in other words, he assumed what we would call today an 's mobile' at
the beginning of the root *(s)peud-. Nybeeg (MO 25, 1931, 189f.)
objected to this etymology on the grounds that the original meaning of
*speud- was , stoßen, drängen' and that this particular root seemed to be
unknown in Aryan anyway. He preferred (p. 190) to interpret pöy- as
1 Cf. Waldb-Pokoeny I 251. By some authorities Lat. vetv^s is identified
with Greek FStoi; instead, so quite recently by Benveniste, Revue de
Philologie 74, 1948, 124—6. According to current opinion (see Leumann-
HOFMANN, Lateinische Orammatik 22, 245f., 459) there are two cases in
Latin where a substantive came to be used (also) as an adjective, namely
über and vetus. I shall show in another paper that über is a good IE adjective
that has nothing to do with über 'udder'. With the disappearance of this
only parallel, the semantic exceptions taken to the identification of Greek
F^TO? and Latin vetus, and most ably put forward by Pebsson, Olotta 6,
1914, 88—91, regain their full weight in spite of Benveniste's endeavours
to eliminate them who, incidentally, does not seem to be aware of the
earlier protracted debate and of his main argument having already been
used to the same purpose. Lat. vetus is, then, syncopated from IE *wetw(o)s
while the oblique cases continue vetero- from wetuso-, cf. socerl from
swekuro- and for the transfer to the 3d declension, pauper from pawo-paro-.
contracted from 'i)ayuS<paty-uSa, the nil-grade of the root mS-', the
South Western correspondence to Av. vaz-, IE. *wegh-, (sich) bewegen'.
This explanation, involving the improbable assumption of the nil-
grade of a root which otherwise is well-attested but only in the normal
and lengthened vowel grades väd/väz (cf. andarväyjandarväz) and also
that of an unlikely contraction of -ayu- to -Ö-', could hardly be accepted
even in 1931. But in 1937 Henning published in his list of Middle
Persian and Parthian words {BSOS 9,87) the Parthian form pwd 'walk,
run' which renders Nybeeg's position altogether untenable. The Turfan
documents have thus brought decisive proof for Müller's suggestion
from the Iranian point of view: pöy- really continues Olr. *paud-. I
think, however, that the same applies to his further Indo-European con¬
nections. Indeed, (jTreuSfo from *s-peud- suits very nicely both from the
phonetic and the semantic point of view.
It seems that today we can advance a step further in om analysis.
One of the hymns pubhshed by Waldschmidt-Lentz in Die Stellung
Jesu im Manichäismus (APAW, 1926, p. 113) contains the fohowing
verse :
1 a 'wt 'br gryw brm'm kw 'g bwxs'n 'c hw
lb 'wt 'cd'md'd'n 'sp'w ky 'yw byd'n x'z'ynd.
The editors translated: ,,Und über das Ich will ich weinen: Möchte ich
erlöst werden daraus und aus den ... wilden Tieren, von den die einen
die anderen auffressen". The gap has been supplemented by Henning
in his List (p. 81) who determined 'sp'w as 'terror, fright'^. I submit now
^ There seems to exist one instance only where the history of an Olr.
-ayu- can be traced with certainty down to New Persian. I have in mind
OPers. mayüxa 'doorknob' which became NPers. mix 'nail, peg'. (If we
interpret the OPers. cuneiform as intending mayüxa so as to equate it
directly with Skt. mayükha- ,peg', the word — attracting the accent onto
the long penultimate — would have remained mayüx). This shows that
mdyüxa followed the regular pattern and lost the unaccented, because
short, second syllable (cf. hist from vtsatl and duvest from duvi-sate),
yielding mayx > mex > mix, seen also in Sogd. myyk 'pointe' in SOB
(= mexk, continued in this form to the present day in Yaghn. mexk 'nail',
see Klimcickij, Zap. Inst. Vost. 6, 1937, 21). There is further a possibility
that Av. täyüri- (sic!) can be connected with MP. ter in this manner, cf.
Herzfeld, Ap. I. 259.
The same principle was at work in the case of -avi- as is shown by
NPers. «0^ from (Av.) taviSi, see Hübschmann, IF Anz. 10, 1899, 24. These
clear instances ought to have cautioned Nyberg against putting forward
his analysis ; the cases invoked in support of his thesis are none of them
convincing.
2 Chr. Sogd. 'wib' .Schrecken' and its correspondence in Buddh. Sogd.
wzp-ywn'k, wzp-ynch 'd'aspect effrayant' (Benveniste, JA 228, 1936. 223
and Henning. BSOS 8. 1936. 585'' and 11. 1946. 723) is obviously a
that this word represents Olr. *us-päwa- from an IE root *peu- 'to run',
and that *(s)petid- is an enlargement of this root. The basic meaning,
preserved in (s)pevd-, was 'to run' which developped into 'scare, fright'
in the same way as the IE variants tremjtresltrep all combine both
meanings, the concrete being the more primitive; cf. for tres-, e.g., the
Homeric i'Tpeaaav S'aXXuSit; aXXv) 'they ran in different directions' while
Pers. tarsidan has confined its meaning to 'fear'.
This root accounts also for Lat. paveö and its family. Walde-Pokoeny
(vgl. Wb. II 76 f.) rather hesitatingly posited a root *pou- ,sich ängstigen' ;
as, however, the only reliable offshoots seem to be Lat. paveo and Greek
TCTOia, Scheu, Furcht' with the verb TtToieco, I believe that both are bet¬
ter explained as continuing IE *peu- 'run' (Lat. paveo from the iterative
*po'weyöy-.
5. NPers. raftan 'to go'
NP. raftan, ravam 'to go' is traced to Olr. rab-, cf. Hoen, Grdr. I 2,137
note 3; Htjeschmann, P. St. 67 (Salemann's root rap-, ibid. 1, 297, is
out of the question). It does not seem to have been noticed that this
Iranian root is connected with the following words deriving from IE
*rehh-: Germanic reh- ,in heftiger Bewegung sein' as attested in MHG.
reben ,sich bewegen, rühren', NHG. (Bavarian) rehisch ,munter'; Irish
reh , Spiel, Tücke' (<*rehä), rehrad ,Kinderspiel', rebaigim ,ich spiele'.
Walde-Pokoeny, II 370, posit for this root a meaning ,sich bewegen,
spielen'; it would appear, however, that the meaning ,spielen', found in
different word. Were Sogd. ß'w 'danger, crainte' (Benveniste, JRAS 1933,
33) secured, it could be explained from Olr. bäya- 'fear' (to bayate = Skt.
bhayate) if we could assume that this became bä which could be spelt with
-w, cf. MPers. nasävjnasäy, both representing nasä as pointed out by
Henning on .several occasions (BSOS 9, 1939, 829 sq., and most recently
ibid. 11, 1946, 732) and also the clear case of ps' in ps's (Salemann,
Manichaica IV 46) continuing OPers. pasäva. But cf. Henning's note in
BBB 68 ad 546. — May I add here that the MPers. conjunctives Savay,
abaräy, baväy, with -'y which puzzled Barb, BSOS 8, 1936, 402, are mere
spellings for -Ö, the correct continuation of Olr. -ähi > äh > ä, and thus
identical with the NW forms in -'h.
1 Walde-Pokobny, II 659, suggest that a more primitive meaning
, drängen, drücken' of the root "'spend- can be found in Greek ottouSo^'
d>eTpißav6i; Hes., which would denote the pestle as ,Zerdrücker'. I prefer
to connect this word with the other root *peu- , schlagen', ibid. II 76. —
I add here that according to Benveniste (TPS 1945, 73) Iran, pod- is
also found in the Armenian loanwords hrapoyr-k' 'attrait, entrainement,
seeluction, sollicitation', hrapuret .entrainer, s6duire' ; I still feel that the
meanings can not easily be reconciled which is rather exceptional in the
case of Armenian loanwords.
the Irish group only, is secondarily developed from an earlier 'romp, run
about'', and this meaning alone is warranted by the clear derivation of
*rebh- from IE *er 'go, run, flow'.
Iranian *mh- from IE *rebh- appears in MParth. as raf- ,angreifen,
kämpfen' (Mir. Man. Ill), raf 'elan', rafay 'assaihant' (Ghilain 56). In
former days, the solution to this sort of variations was to place them on
the Indo-European plane. It can, however, hardly be defended today
that against the unanimous testimony of the rest of the Indo-European
languages, Iranian is sufficient to justify, e. g., a rootform *neph- as
alternating with *nebh- 'navel, kin, people' (as Walde-Pokoeny, I 130).
Today we are more inclined to look for an explanation within Iranian
itself I very much doubt, however, that the latest attempt by S. Vikan-
DEE (Vayu I 59ff ) gets us any nearer to the truth. The grounds on which
he establishes a dialect of the Fry ana-tribe (where -bh- becomes -/-) are
too slender to support his contention. Moreover, in the case of Aryan
näbh-, Iranian näf is found in Sogd. n'f (and accordingly Av. näfa-), but
also in the West, witness the Armenian nahapet and nähbaS, the name
of the kinglets of Gurgan and Dahistan (Vikander 62); fmther, the
Kusan name Nahapäna appearing in India is obviously also derived from
Iran, näfa-päna. Thus the geographic extent of the territory where the
sound change -b->-f- can be observed, militates in itself against the
alleged Fryäna-tribe.
A perfect analogy is offered by MParth. kaf- 'fall' to which belong
MParth. kfwn = MPers. qhwn = NPers. kahan (= Khot. kuharn) 'old'
(cf. Germ. hinfäUig), Khot. kas-, PPP kasta 'fall' < kaf-s- and Osset.
ID. xdun xaud 'to fah' (see Bailey, TPS 1945, 30)2. Already at the end
of the last centmy, when practically nothing was yet known of Middle
Iranian, Justi (ap. Hoen, Np. Et. 282) was led to suspect that Pahi.
kaft .gefahen' belonged to a group of Indo-European words which are
now conveniently listed in Walde-Pokoeny, I 457, under the heading
*qob- ,sich gut fügen, passen, gelingen' :OIrish cob ,Sieg', OE gehaep
,passend', Swed. hampasik ,sich ereignen' = Norw. happa 'id.', heppen
,glücklich, günstig', Olcel. happ , Glück' (whence NEngl. hap, mishap,
happen etc.), OSlav. kobb ''^^yyi, genius, Schutzgeist', ChSlav. kobb 'ol-
covoCTXoTita, augmium', OCzech pokobiti sc. ,gelingen', koba , Erfolg'.
Today, with a vastly increased Iranian material, we can confidently
assert that the primitive meaning of the IE root was 'to fah' which, in
1 In view of the numerous semantic parallels, I suggest that Russian
rebenok 'child', rebjata 'children' derives from this *rebh-, not from IE
*orbho- (if contained) in OChSlav. rabh , Knecht' etc., see Walde-Pokobny, I 184.
^ For Khot. kas- see now Bailey, AsiaMaior II 1, 1951, 31.
Europe, partly evolved into ,passen, gelingen', partly into , Glück, Er¬
folg, Sieg'. A pertinent parallel to the latter change is found in Oss. xai
'share' <xävya- <Iran. *hib- (not kaf-, as Bailey 1. c. 9).
Now in these two cases the existence of the PPP raft and kaft points
rather to -/- having been generalized from forms where -6-, standing
before voiceless stops, was unvoiced. An unquestionable instance of the
process here assumed is furnished by NPers. bäftan 'weave' where the
IE root *webh- became vaf-, first, wc must assume, in the PPP, and
ultimately superseded the original present stem vab-, leaving bäf- instead
of *bäv-. This explanation seems to be corroborated by two further cases
of a different order. For NPers. les- 'lick', Sogd. rys (Yaghn. les, see
Klimcickij, Zap. Inst. Vost. 6, 1937, 22), noone has thought so far of
positing IE Heikh- or a similar primitive form to account for -s instead
of the expected -z from IE Heigh- ; the same applies to MPers. dys- 'to
build', dys , Gestalt' (for further forms cf. Henning, BBB 56), which
is, besides the lonely Olr. *pari-daiza-, the Iranian correspondence to Skt.
deha-^. Similarly, the form näf - must have originated in compounds, say,
e. g., näba-pati- which became näfpat and reacted upon the root word
näb-.
Whatever the case may be, there can hardly be any doubt that
MParth. raf- is an Iranian peculiarity, possibly extending to other dialects
as well, but not contraverting the etymology proposed for Pers. rav-^.
In support of my thesis I adduce Oss. räväg 'leger, rapide' which was
connected by Benveniste with MParth. ray 'rapide (ment)' from (Av.)
rayu-, attested also in the Armenian loanword aragjerag 'rapide' (JA 228,
1936, 200—1). As, however, Olr. *ragu- could only yield Oss. *räyug<
*raguka- or *räy-ag, if the suffix -ogr was added at the Middle Iranian
1 The same change is to be found in Av. zafarjn-, NPors. dahän as
compared, e. g., with the evidently related Lith. i'ebiu .langsam essen',
OSlav. zobati , essen'. Walde-Pokobny, I 570—1, posit an IE root *gep(h):
gebh, but while tho form gebh- is established by the Balto-Slavonic, Germa¬
nic and Celtic data, gcp(h) is only assumed for the sake of the Iranian word
and the Greek gloss yaTzeXstv • äfxeXeCv ,,wenn .apathisch, schläfrig sein'".
The latter can be dismissed without further coment as clearly not belonging
to our group. Thus Iran, zaf- is the only form contrasting with IE *gebh-.
I would suggest, therefore, that / is generalized from cases like the Olr.
gen. zafnah etc. where the primitive 6 was unvoiced before n just as zn
gives snjSn (Av. i/asna-/NPers. jaSn) revealing the same tendency to
unvoicing; for this tendency an instructive case has been found by Hen¬
ning (BSOS 8. 1936. 583) where he derives bainped from baSn -< baiin
thus showing that the tendency was fairly strong even at a late date.
" If Henning is right in aligning Sogd. rnß 'attack' with the Parth.
root raf- and Khot. rraph- (BSOAS 11. 1946. 719). the formation is to
be compared with that of OSlav. sfdp 'sit' < *send- at the side of *secZ-
and similar cases.
14 ZDMG 101
stage, I have no doubts that the Ossetic word continues Olr. *rabalca- 'running, swift' from IE *robhoko- to the root *rebh- 'run'.
^ This explanation seems to throw light on the name of a Saka tribe
which has been much discussed lately. It is well-known that ancient
historians have preserved the memory of an Iranian tribe, variously cal¬
led Saraucae CaxapauXot Sacaraucae, which played a not inconsiderable
role in Eastern Iran in the second half of the second century B.C. Alt-
HEiM restored the name as Saka Ravaka and interpreted this as 'swift
Sakas' connecting Ravaka with Av. rava- < rayu- (Weltgeschichte Asiens
I, 1947, 11 with note 86).' In a detailed philological note (ap. Altheim
p. 24), JuNKEE found this interpretation very attractive; he suggested,
however, that the unattested form ravaka- could be explained as deriving
from *rayvaka-. Without going into details, it will be clear that this form
is impossible from the point of view of Iranian word formation, the form
should have been *rayuka-. I submit therefore that Ravaka- is Olr.
*rabaka- 'running, swift' discussed in the preceding paragraph and sur¬
viving in Oss. räväg 'rapide', which confirms Altheim's interpretation
from the point of view of meaning.
I may be allowed to discuss in this connexion also the names of the
K Saka and the Scythians, since in a recent paper on "Les noms des Saces
et des Scythes" (Beiträge zur Namenforschung 1, 1949, 98—102) van
Windekens proposed a solution for these names which I f ind unacceptable.
As to the name of the Saka, van Windekens takes up the explanation
proposed by H. Sköld (Bulletin de la Societe des Lettres de Lund, 1931—2,
1—5). According to Sköld, a clay cylinder from Essarhaddon's time
mentions "Ispakai from Askuza in Eastern Manna". Although the As-
syriologists take Ispakai for the name of a prince, Sköld argues that,
as is shown by the parallelism between Gimhrai and Ispakai, this is
wrong, and concludes that Ispakai is the name of a people, the oldest
form of the name of the Saka, namely *S'paka. The phonetic difference
between ""Spaka and Saka is easily explained by the well-known fact
that in KLhotanese, the language of the Saka, Iranian sp (from Aryan
sv) appears as ss, cf. aspa ■—' assa, vispa ■—' bissa. As the name of the Saka
appears as such in Darius' Behistun inscription, Sköld draws the further
inference that the sound change sp > ss must have taken place between
679 B . C . and the date of the inscription, adding thus a valuable point to our
scanty knowledge of early Iranian dialectology — if his result holds good.
Sköld's paper, published posthumously and therefore possibly not
completed by the author, does not contain any hint as to the meaning
of this name. Windekens, accepting the whole of Sköld's argument,
' A different Iranian pre-form is now reconstructed for Saka Ravaka by
Altheim, Aus Spätantike und Christentum, 195L 95.
i-ounds it out by concluding that Saka, or rather Saka as in Sanskrit,
— Saka being the Old Persian form with substitution of s-Jfor s- non¬
existent in the phonetic pattern — is the Saka form of the well-known
Iranian word spaka 'dog' (Pers. sag) showing that the Saka designated
themselves as 'chiens, Hundeartigen' meaning 'watchdogs (of the cattle)'.
It would have been possible, of course, to hit on the idea of equating
Saka with Pers. sag independent of Essarhaddon's tablet. But as it hap¬
pened, the idea was suggested by the name Ispakai occurring in his
prism, representing, according to Sköld, Spaka which became Saka in
the century or so preceding Darius' reign. Now there are two consi¬
derations militating against this conclusion. Eirst, by some mishap, the
Khotanese representative of Iranian *span- 'dog' is not ssan- but svän-
or s'sMwäw-.WiNDEKBNs' efforts (p. 99) to explain away this fact by taking
u{v) for a secondary adjustment to the oblique stem *sun-, wih hardly
convince anyone since the Saka form obviously continues an Olr. form
*suvan- corresponding to Skt. suvan- and Greek xüwv, with full vowel
between s and v, for which the IE doublets *duw5ujdwöu canbe compared.
Second, if a primitive Iranian *swan- reached the stage *sj)an as
supposed by ISpakai — then sp could not possibly change to ss as is
proved by the fact that, e. g., Iran. *spas 'look, watch' has been preserved in Khot. spässa- 'see' < *spas-ya- ; it is only from the stage *sw- that we
can get at ss- or -ss-. To be sure, Windekbns seems to be aware of this
possible objection to his explanation and tries to obviate it by having
recomse to an other Saka dialect, different from Khotanese in this
respect. But when he refers us to the Pamir dialects, and even Afghan,
as Saka dialects showing sp from sw, he simply begs the question.
Besides, the form Spaka appearing on the frontiers of the Assyrian
empire, we are compehed to extend the territory of this 'allotrop' Saka
dialect to cover the whole area between Pamir and Media, a conclusion
which will appeal to very few scholars.
But from the semantic point of view, too, the strongest objections
must be raised against the assumption that the word 'dog' might have
been used as a tribal name. Windekens insists that "cette appellation
n'aurait pas eu le sens d'une depreciation, au contraire, puisque chez
eux cet animal domestique etait entome d'un respect peu ordinaire,
voire meme d'une veneration religieuse" (p. 100), but he is hard put to
it to find an actual instance of such usage in our tradition. The name
CTiaSaxT)? of a Pontic Scythian, quoted by him after Coman (p. 101), has
nothing to do with spaka, but must be derived from *spädaka — cf. our
name Trooper and similar names of military origin in other languages —
and the case of Cyrus' nmse Spakö throws light, if any, only on the not
quite flattering views of the Iranians regarding their women.
14»
For these reasons I cannot accept Windekens' explanation of the
name Saka. Instead, I submit that Saka is the adjective/agent noun
derived from the Iranian root *sak- 'go, flow, run' found in the morpho¬
logically still unexplained OPers. Oakatä, a date formula preserved down
to MParth. sxt and Sogd. syty' (see BBB 115 and 134). This root belongs
to Lith. Sokti 'to jump' (which cannot be reconciled with Slav, skok-)
and underlies also Skt. sakatam 'wagon' — cf. Lat. currus from cnrro =
Celt, carrus — possibly reflected in the Hungarian loanword szekir
'wagon', see B. Munkäcsi, Ärja es kaukdzusi elemek a finn-magyar
nyelvekben, Budapest 1901, index s. v. The name Saka thus means
'running, swift' or 'vagrant, nomadic', either of them an appropriate
description of the swift horsemen of the steppe, cf. also the designation
Ravaka discussed above.
Since the pan-Iranian representative of this word is Saka, it is only
natural that this name should appear in this form also in Darius' ins¬
cription. As regards the Sanskrit form Saka, it should be borne in mind
that by the time the Saka appeared in India, that is to say in the second
century B. C, the inhabitants of "Hindu" must, through their intimate
contacts with the powerful Western brethren stretching over several
centuries, have become aware of, and familiar with, the fact that in the
conqueror's language s corresponded in many cases to their s'. Thus it
was only to be expected that they should adapt the name Saka to their
own pattern in the form Saka.
In his paper Windekbns proposed a new solution also for the name
of the Scythians. According to him (p. 101 f.), *8kuda- (sic!) derives from
the root *{s)keu- 's' appercevoir, veiller etc' ("le role de -da dans *skvßa- ne m'est pas clair", p. 102) and means "gardeur(s) (de troupcaux)". The
same notion is found in the related tribal name CxoXotoi and the personal
name CxüXt)?, formed from the same root with an -I- suffix. It should be
added, for the sake of completeness, that the name Skunxa appearing
in the Behistun inscription, and Kusäna, contain the same voot*{s)ku-,
forming with the preceding names a coherent pattern.
As my own paper on the name of the Scythians, published in 1947
in the Hungarian periodical Magyar Nyelv ("Hungarian Language"),
seems to have escaped Windekens' notice and, owing to the language
barrier, not too many Western scholars can be expected to be familiar
with my results, it might be not out of place to summarize them here.
Before doing so, however, I would sum up my criticism of Windekens'
suggestions.
^ It should be remembered that of all Middle Indian dialects it was the
North Western group that stuck most tenaciously to s while the rest
developed this sound into a.
Iranian *Shußa- can represent a derivative of the IE root *sheu- ; it
is, however, impossible that CxoXotoi and CxuXt)? should represent a
formation with an -I- suffix since the Iranian phonological pattern has
no I sound and consequently no I suffix'. But the main objection to be
raised against Windekens is, that he takes no notice of either the labours
of modern Western scholars or the relevant data of antiquity. It will be
as well, then, to proceed to the presentation of my own view.
The name of the Scythians is well-known in its Greek form : Oxii&ai,the
Iranian form of which could indeed most readUy be reconstructed as
*skuOa-. There is, however, an independent line of tradition that contra¬
dicts this conclusion. The Scythians, ousting the Kimmerians from their
earlier home at the end of the Vlllth century B. C, were not contented
with the wide expanse of the Pontic region, but made several inroads in
Asia Minor and even the Near East, following, it would appear, everywhere
in the footsteps of the Kimmerians. To counteract the threatening at¬
tacks of the Kimmerians and their allies, Essarhaddon of Assyria was
anxious to conclude a treaty of alliance with the hordes in their rear who
appear in the cuneiform tablets as ASguzai Iskuzai (cf. Minns in Cam¬
bridge Ancient History, vol. Ill, 1929, p. 188). It is obvious that these
are the Scythians and the phonetic structure of the name Is established
beyond any reasonable doubt by the evident emendation of the biblical
name 'Sknz (Gen. 10, 4) into 'Skwz.
We thus have to try and establish a common Iranian form which can
appear as Skytha- in Greek, and Askuza- in Assyrian-Hebrew. It is
plain that Iranian -0-, while accounting for Greek cannot be reconciled
with Semitic -z- ; therefore Windekens' explanation must be ruled out.
Another view was proposed by Maekwart (Philologus, Supplement-
band X, 112) and adopted by Herzfbld (e.g., AMI 1, 1929, 104';
Ap. I. 307) who suggested an Iranian basic form *Skuca. It cannot be
denied that in early Greek transcriptions of Iranian names !> seems to
represent Iranian c as, e. g., in 'Acttckö-ivtji; = OPers. Aspacanä, Ts'kt;x7)i;
= OPers. Ciäpiä. On the other hand, it does not seem to have been
1 The samo exception must be taken to the personal name KoXdt^ai? being
interpreted as kula-xajas 'rex generis' (Windekens p. 101, following the
dilettante Coman); 'genus' should bo "Scythian" *kara-. Venturing on this
slippery terrain, I would suggest to interpret kola- as Scythian kala- from
*karya- 'army', of. OPers. kära- and Greek xotpavo?, Germ. *harjis from
IE *koryos, and for the phonetic change -ry- > -I- the name of the Alans,
continuing Aryänäm. (Note that Scythian had a "secondary" I from ry
and from yet another source, see my explanation of the name Skolotoi in
tho text further on). Accordingly, the name Kala-xsaya- (in Greek guise
Kolaxais) means ,army lord'. For -ry- > -I-, cf. now also Szemerenyi ap.
Altheim, Geschichte der lateinischen Sprache, 1951, 75 fn. 1.
noticed yet that this derivation is precluded by the Semitic form of the
name. Iranian c should, indeed, appear in Akkadian transcriptions as
§, or possibly s, see Eiders, OLZ 1935, 212 and ZDMO 94, 1940, 215';
the same can be inferred from Talmudic loanwords which similarly show
5 01 s for the Iranian surd affricate, see Teleqdi, JA 226, 1935, 205;
hence it follows that Semitic z cannot correspond to Iranian c. Nor does
the semantic solution, proposed by Markwart at a later date (Caucasica
6 I, 1930, 58 f.): Sku-ca- 'scalp hunter', seem more satisfactory. The root
*sku- ,das Pell abziehen', based on Khotanese, cannot be regarded but
as an unwarranted inference from certain usages in one text group, not
supported by other Indo-European data. Furthermore, if there was a
name of this character, it surely must be looked for in Strabo's CapdcTtapai;
olov xecpaXoTOfAou? (XI 14,14) where xecpaXr) clearly reflects Iran. *sarah-
and the second compositional element can also be explained from
Iranian'.
It would seem that the real solution had already been given implicitly
by Justi when stating (Grdr. II 441): ,,Das Volk, welches die Hellenen
mit dem skythischen Worte Skythen (Schützen, luTioTo^oTai, Hdt.) be¬
nannten, bezeichnet sich selbst mit dem Namen Skoloten". I submit
indeed that the name of the Scythians represents Olran. *Skuda- 'archer'.
It is well-known that the chief weapon of the nomadic horsemen, and
so also of the Scythians, was the bow and arrow. This is amply illustrated
by the remains of ancient Persian and Scythian art, but perhaps even
more strikingly by a passage of Herodotus where it is described that at
Marathon the Persians thought the Greeks out of their senses opcovxe?
auToüi; eovTa? oX'.youc, xai toutou? Sp6[i.co l7TeLyo[j,£vou(;, oCts tTrirou
'j7rap)^ou(Tif)i; ff^fi gute To^eufxaTCov (VI 112). In other words, the Persians
were struck by the absence of what was natural to the Iranian: the
archer riding on his flying horse. In another passage, cited by Justi,
Herodotus says: lolai yap [xy,Te acTTsa [ay^te tei/^e« f) exTt(T(i.eva, aXXa
tpepeoixoi eovTE? ttocvtec scocri, mTroTo^oTai, . . . xw? oux av sl'vjtTav outoi
afxaxoi xtX (IV 46). That the Scythians were, in the eyes of the
Greek, in particular of the Athenians, essentiahy riding archers, is shown
by the remarkable fact that tho name of the Athenian police was CxüDai
or To^oTai, thus showing that the names were identical in meaning^.
And the fact that the basic word of the latter name, Greek to Eov, is a
' Hebzfeld's suggestion of a connection between Sku a and OPers.
akau-di- (Ap. I. 307) is ruled out by Semitic -z-.
^ Cf. also the following well-contrasted passage in Xenophon's Anabasis
(III 4, 15): ZTzei . . . oi 'P6Sioi Ea9Ev86vv]CTav xai ol CxuS^at TO^öxai ExÄ^Euaav, xai oüSsli; YifAapTavsv avSp6i; xtX., where the accentuation of the "national"
weapons will be noted,
Scythian loanword as was cogently demonstrated hy Benveniste {Mdl.
Boisacq I, 1937, 37 f.), is an additional proof of the thesis that the
Scythians' distinctive weapon was the how and arrow.
It is thus plain that from the historical point of view the interpretation
submitted above offers a very suitable explanation of the name. We now
shall investigate the philological presuppositions of this etymology.
Iranian does not seem to know the root *skeud-. In Sanskrit skud- is
attested in the Dhätupäthas as being used "äplavane", not a very suitable
meaning it would seem. In Germanic, however, the following closely
knit group is found: OHG. scio^^an, OS. skeotan, OEngl. sceotan, Olcel.
skjöta 'shoot'; the word is also attested among the meagre remnants
of the vocabulary of the Crimean Goths in the gloss Schieten 'mittere
sagittam'. From this root are formed OHG scutto (giving NHG. Schütze,
cf. Justi above), OS. schütte, OEngl. scytta, Olcel. skyti 'sagittarius' ; the
interpretamentum of the last word is of special importance from our
point of view.
But we can advance a step further and nearer to the Iranian orbit.
The Germanic group quoted above has been variously connected with the
following Balto-Slavonic words: Lithu. Sduju Soviau Sduti ,Brot in den
Backofen schieben; schießen', East Latv. saut ,schießen', Latv. Saünu
äävu Saut ,schieben; schießen'; OChSl. sovajg sovati ,werfen, schleudern',
Bulg. sövam ,hin- und herbewegen', Russ. süju and suju, sovdtb ,rücken,
schieben, einstecken, stoßen', and with an w-formans OChSl isungti
'ex/seiv', Slov. sunem suniti ,Stoß versetzen', ORuss. sunuti ,schleudern
(z. B. Lanze)'. This group has been equated with Goth, skewjan ,gehn',
Olcel. sksewa ,sich vorwärts bewegen'. East Fris. schdjen ,sich umlier^
treiben' < *skaujan, see Trautmann, Baltisch- Slavisches Wörterbuch,
p. 300, and also with the Germanic root *skut- 'shoot'. This equation can,
however, only be upheld if we assume that Balto-Slavonic §- continues
IE *sk- with a palatal k and that this group gave Balto-Slavonic S.
Although this view has recently been voiced by no less an authority than
Endzelin {Zeitschrift für Slav. Phil. 16, 1939, 107—115, esp. 114), I
cannot but agree with Stang {Das Slavische und Baltische Verbum, 1942,
137 f.) that this thesis has not been proved'. In my view, however, there
is positive proof against this theory. It does not seem to have been
noticed that Lithu. skaust ,es tut weh' must be traced to IE *skoud-, the
o-grade of the root *skeud- under question, and shows the interestmg
semantic change of the meaning ,es schießt' (cf. German Hexenschuß)
evolving into the general sense of pain. In other words, Balto-Slavonic
has the IE root *skeud- in its expected form and thus BSI. Sawä- cannot
1 1 also agree with Stang, 1. c. 135i, that M. Leumann's view of a change
IE sK > BSI. st (IF 58, 1942, 126f.) is highly improbable.
be connected with it, at least not directly. It would appear, therefore,
that the latter should be equated with Av. spä- 'throw', both continuing
IE *lc{e)wä- as suggested alternatively by Teautmann, 1. c, and Kury-
towicz Etudes IE I, 1935, 20'.
With IE *skeud- 'shoot' thus secured for Balto-Slavonic, it seems
reasonably safe to assume the existence of the same root for Iranian.
What is more, Skt. skud- 'äplavane' would also seem to fit now into tho
general pattern as shown by the semantic development seen in many
Germanic languages, cf. Danish skynde ,antreiben, beschleunigen',
skynde sig ,sich beeilen', Germ, dahinschießen, Engl, to shoot^.
An Iranian representative of the basic root *skeu- seems to be preserved
in Oss. wxsun sexsin ,schießen, fangen, treffen'. Miller suggested (Os¬
setisch § 33,8) Skt. aksati 'to reach' ,, obgleich iranische Parallelen mir
unbekannt". I am of opinion that the word continues Olr. *skau- 'shoot', possibly found also in Sogd. nSk'w 'tirer, faire sortir', cf. Benveniste-
Gauthiot, Gramm. Sogd. II 19 and Henning, JRAS 1944, 138 with
140*. I maintain therefore that Old Iranian had the IE roots *skeu- and
*skeud- 'shoot, haste' and that the name of the Scythians is an agent
noun derived from the latter form: Skuda- 'archer'. This view seems to
find support in the legend handed down by Herodotus according to
which (IV 9sq.) of the three sons born of Heracles' marriage it was the
youngest, Skythes by name, who was alone able to tend his father's bow
and by his skilful handling of this weapon deserved to become the
ancestor of his people.
Some remarks must now be added on the Semitic and Greek forms of
this name. Semitic aSkuza causes no difficulties. The prothetic vowel as
well as ä for s can be due either to Semitic adaptation or Scythian
dialectic variants, these phenomena being well-known in Iranian. Sound
substitution seems also to explain -z- for the Iranian spirant -d-; as
KÖNIG says {Der falsche Bardija, 1938,192): „In dieser Gegend (Assyria)
wechselt in den Schreibungen z mit d."
1 Endzelin, 1. o. p. 109, derives Lithu. sok-, through metathesis, from
kSäk- < skäk- in order to equate it with Slav, skakati ~ sko&iti. In my
opinion, Lithu. Sok- belongs to Iran, sak- 'run', see above, and must be
kept apart from Slav. skok-. I admit, however, that the latter root might
ultimately represent IE s-Uek- if, by Meillet's depalatalization law, sK-
became sh- as maintained by Kurylowicz, 1. o. p. 19f. In that case, skeu-
'shoot, move rapidly' could also represent an enlargement with s mobile
of the root *Uewä- 'throw' established above.
^ Since Lithu. has, by the side of skaust, also Saudyti 'shoot' which,
according to our results reached above, must continue IE *Keud-, I sug¬
gest that Sogd. swd swdn 'haste' (cf. Henning, BSOS 8, 1938, 585'')
continues the same root and thus reveals the same enlargement as Lithu.
Saudyti compared with Sduti,
In the Greek form, only one point requires to be elucidated: -8 forlr.
d. It is well-known that the Greek dialects of the VII — ^Vth centuries B.
C. had no interdental spirants; the question is thus reduced to the more
narrow one as to what Greek sound is likely to substitute for the voiced
spirant. A case in point is the name of the river Don whose Iranian name
Dänus gave rise to the Greek form TavaCc, cf. Kebtschmee, Glotta 24,
1936, 4: ,,die Griechen gaben das Iran, d, weil es ungefähr in der Mitte
zwischen ihrem S und t lag und sie nicht von der Schrift, sondern nur
von ihrem Ohr geleitet wurden, mit t wieder"'. It would appear now
that the "breathing" element of the interdental spirant invited substi¬
tution by an aspirate sound. That languages lacking the spirants in
their phonological pattern, are bound to take this course, is well shown
by the case of OPers. v^dana- 'hamlet, town' which was borrowed into
Sanskrit as vardhana- 'town', see Wackeenagel, KZ 67, 1942, 168—9.
Herodotus informs us (IV 6) that GU[nza.m elvai ouvo|jia GxoXotouc; toü
ßacnXecoi; £7rwvu(i,t7)V. CxiBai; Ss "EXXtjvs? ouvofiacrav. While the second
part of this statement can by no means warrant aninference that the name of
the Scythians was of Greek origin, the first part gives rise to the question
as to what relation, if any, there is between the two names, Skolotoi and
Sky thai. It appears plain from the outset that the form Skythai must
owe its origin to the first, superficial contacts of the Greek merchants
and adventurers with the Iranian nomads or their neighbours on the
Northern shore of the Black Sea, while Skolotoi reflects a later stage of
more intimate relations between the settlers of the Greek colonies and
their ,hinterland'2.
The name Skolotoi has been variously connected with the name of
the youngest son of the primeval king of the Scythians, namely Kolaxais.
K. RÖNNOW says {Le Monde Oriental 30, 1936, published 1944, p. 151):
"Kolaxais is Iran. Kolo-xSaya 'the king of the (S)kolota8"'*. Since,
however, the name Skolo- is well attested in the name of king Scolo-pitus
preserved by Justinus (cf. Maekwaet, Phil. Suppl. X 78 sq.) and also in
the name of king Cx'jXyjc (Hdt. IV 76), there can be no question of con¬
necting forms with and without the initial s ; it is incredible either that
1 A curious sound substitution is found in Etruscan ziumide for Greek
AiofXTjST]? (ScHWYZEB, Griech. Gramm. 154); here the initial S is represented
by z while the same sound in the middle of the word appears as 0.
" There is no foundation whatsoever for Maekwart's assertion (Gauc.
1. o. p. 56) that the name Scythians was ,, sicher ein dem Volk selbst unbe¬
kannter Schimpfname".
^ In view of fn. 1, p. 213 above, it should be added here that RöNNOW
does not shrink back from interpreting the word skolo- as 'snake' and to
connect it with tho Indian tribal name Kuliita which should all testify to
a snake-totemism.
the Scythians should have used both forms as variants or that the
Greeks should have dropped it in one form.
It is much more plausible therefore when Mabkwaet (Cauc. 1. c. p. 60)
derives the name Skolo- from Iran. (u)skarya- ,jagend, Jäger', assuming
the sound change -ry- > -I-, for which see also p. 213 fn. 1. But as
wih be seen from the basic form posited by him, he interprets this word
as derived from the root ^har- with the preverb uz-. Now while it is pos¬
sible that an Iranian name beginning with a consonantal cluster should
receive in a borrowing language a prothetic vowel (cf. Askuza), it is
impossible that the borrowing language should drop the initial vowel of
Iranian. And although Maekwaet might have started from the well at¬
tested Iran, verbal root skar- 'lead, push, hunt', avoiding these dif¬
ficulties, I am of opinion that the real solution is much nearer to hand.
We know from Herodotus' narrative (IV 6) that the aforementioned
king Kolaxais became the ancestor of the royal house of the llapaXaTai.
This name was, obviously correctly, equated by Vasmee (RVO XII 241)
with the Avestan word *paradäta, appearing as an attribute of prince
Hausyaha, but in later tradition designating the primeval royal house
of the Pesdat. Vasmee already pointed out that the phonetic change
d>l, implied in this derivation, was well-known in Sogdian and the
Pamir dialects. Today we can go even further and state that this phonetic
change, though not general, sporadically appears over the whole of the
Northern Iranian territory, last but not least in Ossetic. I suggest
therefore that Skolo- is nothing else than the dialectic variant of the
name Skuda-, with the same phonetic interchange.
It wih be clear that this coherent interpretation of the two names
of the Scythians is recommended by internal reasons ; for the explanation
of Skolo- it should be added that, since I cannot continue an IE I, it
must needs be of secondary origin. If it comes now to a choice between
Olr. skarya- and skuda-, the scales are tipped in favour of the latter by
the following point : the name of king Skyles, obviously an eponym of
his people, shows that the Iranian vovel was u.
Both observations point to a geographical and chronological dif¬
ference in the two borrowings. The Greek form Skythes shows that it
was borrowed at a date when Greek u still had the value u; the form
Skolotoi was borrowed at a later date when Greek u already had the
value Ü and thus could not substitute for Iran, u which had to be re¬
presented by the nearest Greek sound, namely o. This chronological dif¬
ference has to be supplemented with a geographical factor : in my opinion,
there is a high degree of probability in favour of the assumption that the
dialectal change d>l was a distinctive feature of the language of the
central region of the Scythian expanse, the territory where their royal
house and hs tribe had their abodes, while the coastal regions preserved the original spirant.
MParth. abgäm
In Henning's often quoted List (p. 80) we find the fohowing Middle
Parthian words :
'bg'm, abgäm 'torment, agony'
'bj'm'dn, abzämädan 'torture, agonize' 'bf'mySn, abzämiän 'agony'.
It is obvious that these words have nothing to do with the weU-known
group of Iranian gam-jzam- 'to go', and so Ghilain is at a loss as to their
etymology (Verbs Parthe 73). However, there can be little doubt that
this group belongs to the IE root *gem- .greifen, fassen, zusammen¬
drücken, pressen' (Walde-Pokorny I 572—4): it wih suffice to quote
from among its proliferous derivatives Arm. cmkm .(zusammen) drük-
ken', OSlav. ibmg z^i 'CT9ÜYYSIV, comprimere\
It is of some importance for Indo-European dialectology that whUe
the South Indo-European representatives of this root show the meaning
jgreifen, fassen' (cf. Greek yevxo 'he seized' etc., Lat. gema'* to be full
>sigh'), it is in the geographically adjoining territories of the Slavs and
Armenians that we find the closest semantic parallels to the development
seen in the Iranian group ; the original meaning of Olr. ahi-gama-f
abijama- was 'oppress'.
The same grouping of isoglosses, significant from the point of view
of Indo-European linguistic geography, has now been observed in the
roots *rab- 'go, run', *kab- 'fall' and *skeud- 'shoot' which thus should
be added, together with *watuäa- 'old', to the list of Abntz' Sprachliche
Beziehungen zwischen Arisch und Balto-Slavisch, 1933, p. 35 ff.
Von Christian Rempis, Tübingen
Die Dichterbeschreibungen [Tadkirat) führen gemeinhin als erste per¬
sische Dichter nach der arabischen Eroberung Persiens den Grammatiker
und Philologen Hakim Abü Hafs-e Sogdi und den Sprachgelehrten
Hakim 'Abbäs-e Marw! an, von denen der erste bereits im ersten Jahr-
huYidert d. H. gelebt und gedichtet habe.
1
Von ihm wird folgende Doppelzeile überliefert :
4.J_^.^>- jl(J jl) jjljU j\ O-ia J^^^^i^yt,\
Das muß metrisch wohl wie folgt gelesen und berichtigt werden:
„Ähü-je köM andar DaSt cigün daw'dä ?
Ö ne därad Jar, ahe Jär cigün baw'dä ?
= Wie läuft die Berggazelle wohl in (ebner) Heide ?
Sie (er) hat keinen Freund; wie mag sie (er) ohne Freund sein?"
Das ist noch ein rein persisch empfundener Zwölfsilbler, in dem —
wie noch heute in den mundartlichen Dichtungen' — nur die Silben ge¬
zählt und noch nicht deren Längen und Kürzen wie in der arabischen
Dichtung beachtet werden:
->--| I II
was aber später überliefert ist, als ob zu messen wäre :
Deshalb kann die von Sams-e Qeis in seinem Kitäb al-Mu'aggam ft
Ma'äjjir AS'är al-'Agam an anderer Stelle gebotene Angabe nicht stim¬
men, daß Abü Hafs-e Sogdi im Jahr 300 d. H. gelebt habe. Er muß, der
sprachlichen Gestalt der ihm zugeschriebenen Dichtung nach zu ur¬
teilen, wirklich noch in einer Zeit gelebt haben, da die Überlieferung aus
der Zeit der mittelpersischen Dichtung noch nicht abgebrochen war. Da
1 z. B. ,,Haftsad Taräne az Taränehä-je rüstä'l-je Irän . . . gerd äwarde-je
Hosein Kühi Kermäni", Teheran, Ausgabe 1327 H. s. (1948); vgl. auch
Salemann-Shukovskis „Persische Grammatik", S. 102.