Professional Documents
Culture Documents
AND INDO-EUROPEAN
by OSWALD SZEMERENYI
One of the characteristic features of the Greek verbal system is the use
of a distinct perfect participle in the system of the active voice. In histo-
rical Greek its suffix appears as -w~, -6~, -6"t'oc, -6"t'o~ etc. (e.g. e:t~w~, e:t~6"t'oc etc.)
in the masculine-neuter, that is the suffix apparently ends in a voiceless den-
tal stop. But the feminine ending -ULOC (e.g. Horn. t~ULOC, Att. e:t~IJLOC), which
clearly contains the normal feminine suffix -ya and must therefore have been
based on a participial suffix ending in -uC 1, cannot be reconciled with a
dental suffix. A suffix in -t would have yielded a feminine form ending in -t-ya
which would appear as -Cl'Cl'OC or -"t"'t'oc in the Greek dialects. The difference in
vowel between -O"t'- and -U-LOC can be more readily explained on the assump-
tion that originally the participial suffix began with a digamma, so that the
masc.-ntr. form -wot- alternated with -u-, where -u- was the regular nil-
grade form of the full-grade form -wo-.
This is as far as the historical method will take us 2. In order to find a
solution to the problem left unanswered, we may now exploit the compara-
tive method. If we turn to the cognate languages, we find the following data.
In Sanskrit, the perfect participle corresponding to Gk. e:t~w~ shows a stem-
form vid-vas-, alternating with vid-vat- in the so-called middle cases; the
feminine nom. sg. is vid-u~-i. The latter is particularly interesting as it clears
up the formation of Gk. t~uLocfe:t~ULOC: it is from *wid-us-ya, where -us-
is the nil-grade of a suffix -wes jwos-. But Sanskrit is also interesting in that
it seems to offer the variation -vas- j-vat-. A comparison of Sanskrit and Greek
3 On the distribution of these variants within the paradigm see the text further on.
grec ancien, 1933, 439; Morphologie historique du grec, 1947, 337, and 1964", 282;
Schwyzer, Griechische Gramm. I, 1939, 539 f. ; Meillet-Vendryes, Traite de grammaire
comparee des langues classiques, 19531 , 337 f.
I See M. Lejeune, Memoires de philologie mycenienne, 1958, 227 fn. 36, who suggests
ci=]8e:8Ixurexl 'assignees', 'designees '. Ever since Documents (p. 163) the word has
been commonly interpreted as 8e:81<llex)xurexl having completed their training', and
connected with didakare "*lllllex=&:M:1 ", "at school" (see Chadwick, Glotta 41, 1963,
184; Palmer, The interpretation of Mycenaean Greek texts, 1963, 47 12 5 f. 412 f.),
with no attempt to account for the appearance of the present (!) reduplication outside
its proper sphere at such an early date. The form is probably relevant to our discussion
but we can make no use of it. See also Chantraine, RPh 38, 1964, 260.
Perfect Participle Active 9
" der am besten sich versteht auf, im Gedachtnis hat ", zazus-tllmo "der im Kampf
am meisten gewonnen hat, der siegreichste" (on this Benveniste, BSL 52, 1957, 13;
Humbach, Miinchener Studien zur Sprachwissenschaft 10, 1957, 39f. with fn. 12).
11 See Wackernagel-Debrunner, ALGr. I Nachtrage, 1957, 97. The secondary cha-
a fact that is rather less surprising than the earlier assumption that it
mingled in an unheard-of manner two - which alternated between -was/
wosjus-. There is no evidence that there was, besides these variants, also
-wes- 12
We are now in a better position to assess the value of the evidence alle-
gedly supplied by the other Indo-European languages. Even with the current
confused ideas about the formation of the perfect participle, J. B. Hofmann
held that Lat. caput was rather more likely to contain a suffix -t- found in
other names of parts of the body than the perfect participle suffix -ut- 13.
I would merely add that, even if one accepts a suffix -wot-, there is absolu-
tely no trace of a variant -ut-. Lat. apud presents a much more interesting
and much more complex case. Here even Hofmann (op. cit., p. 60) is prepa-
red to accept the view, first advanced by Georges and Wharton, then inge-
niously argued by Brugmann, that apud is the neuter perfect participle of
the verbal root appearing in apiscor 14. It seems odd that it has not been noti-
ced until now that the semantic interpretation" erreicht habend" makes
the formal analysis unacceptable. For indeed, if we accept the derivation
- and it seems very attractive - we must assume that, as in the case of
aduersus and other stereotyped words, it was the masculine form of the parti-
ciple that was so often used that it eventually acquired a prepositional force 15.
In other words, the Latin preposition can only have started as the masc.
participle *ap-was. How can this primitive form be reconciled with the Clas-
sical form ?
tion with Skt. kapucchalam upheld by him (as if from *kaput-salam) is hardly tenable;
see M. Mayrhofer, ZDMG 105, 1955, 236 and Kurzgefasstes etym. Wb. des Altindischen
I, 1953-56, I56f.
U Brugmann, Berichte der Sachsischen Gesellschaft der Wissenschaften 35, Leip-
zig 1901, 108-10; he holds that apud is a ntr. acc., like aduersum, a point on which Hof-
mann, o.c. 60, does not make his position clear.
16 See on this phenomenon Brugmann, IF 27, 19IO, 234 f.; Leumann-Hofmann,
At this juncture two facts must be recalled. First, the original alterna-
tion between -wosjwosjus- led in many languages to levelling, to the genera-
lization of one variant. For all practical purposes, the Classical form of the
Greek perfect participle can be described as having the suffix -ot-, with cer-
tain morpho-phonemic variations in the nom. sg. masc., the nom. acc. sg.
ntr., and the dat. pI. In contrast to Greek, the Sanskrit neuter seems to have
extended, in some cases at least, the weak form -us- which was so overwhel-
mingly represented in the paradigm. Leumann has shown that Vedic cak$u$-
" eye" is the neuter participle of the perfect ca-k$- "to see" 16. Intrusion
of the weak stem into the strong cases is also seen in Vedic cakru$am 17. This
process is carried to its logical extreme in the Avesta where we find nom. sg.
masc. mamnus "thinking", vidus " knowing" 18 and is also found in the
Slavic ending -1> from -us, e.g. neS1> "having carried" 19. From the point
of view of Latin the Oscan form sipus " sciens " is even more important if
it really represents *sep-us 20. These examples will suffice to show that it was
possible for Latin to transform the original complex inflection into the more
regular -us jus-em etc. The oblique stem, after rhotacism, became -or- which
was then extended to the nom. We may therefore expect a Latin apor, apur
as the nom. sg. masc. of the perfect participle.
The second point concerning Lat. apud is that, as is well known, beside
the "normal" Classical form we also find apor and apur in Early Latin.
It is also known that a similar alternation between ad and ar occurs in the
same period. It is usually assumed that, since the original sound in both words
was -d, this sound under certain conditions, namely before labials (cf. apur
jinem), regularly developed into -r 21. But it is impossible to see the phonetic
ratio for this change, although one can see the advantages of a by-form ar-
before labials - once such a form is established. Now ofthe two prepositions ad
and apud, only the former has an undisputed etymon : EngI. at and other
cognates guarantee an lE *ad 22; apud, on the other hand, if it is derived
1& Leumann, Morphologische Neuerungen im altindischen Verbalsystem (Mede-
deelingen der K. Nederlandse Akademie van Wetenschappen, 15/3), 1952, 105. See also
Tedesco, Language 21, 1945, 136 f.
17 Wackernagel-Debrunner, Ai.Gr. III 297.
mann, o.c., 340. Leumann also adds Lat. memor, in his view transformed from *memnus
(: Av. mamnus) after memini, but this is rather uncertain, see Walde-Hofmann, o.c.,
Il 67f.
21 Leumann-Hofmann, o.c., 128; \Valde-Hofmann, o.c., I IIf. I ignore here the
question whether -d from an original -t, as would be the case here, could also develop
into -r. Pisani's assumption that ar apor meridies show a Faliscan rhotacism of d (Ricerche
Linguistiche 5, 1963, 60) is based on nothing.
22 An Illyrian ar "an, bei, zu ", perhaps identifiable in 'ApM-nov (cf. EPidotio) ,
'Aplh.u't"1J, ' ApM~(Xvo<; sim., is rightly explained by Krahe (IF 62, 1956, 249-50) as due
Perfect Participle Active 13
from the root of apiscor as its perf. participle, can only be traced to apur,
apor 23. It seems reasonable to conclude that the four variants derive from
the blend of two original forms, *ad and *apor, which produced ar and apud
respectively. Once the new variants gained a measure of currency, it was
natural that they should be utilized for the purpose of dealing with peculiar
problems, such as those arising out of the sequence of dental and labial 24.
Certain Celtic forms are also generally accepted as presenting lndo-Euro-
pean perfect participles with the suffix -wot-. Sommer interpreted O. Irish
bibdu "culprit, defendant" as deriving from lE *bhi-bhid-wot-s, the perfect
participle of the lE verbal root *bheid- seen in Lat. findo "split" etc. 25. While
accepting the morphological analysis, Pokorny objected to the assumed pri-
mitive form on the grounds that the resulting form would be the palatalized
*bibdiu/bibded; he therefore proposed *bhe-bhud-wot-s, from the root *bhaud-
" beat" seen in, e.g., O. Engl. beatan "beat ", which, through *bibudus,
resulted in the attested bibdu 26. But, quite apart from the meaning, even
the formal analysis is anything but certain. There is nothing to prove that
the ending was -wot-s rather than simple -ot-s, especially as simple dental
suffixes are so frequent in Celtic. What is more, even the root-part may have
to be interpreted very differently. If, instead of dividing -d-ot-, we place the
cut before -dot-, we may see in the latter the compositional form of the root
*do- "to give", i.e. "-giver", or of *dhe- "place, do", cf. Lat. sacerdot-
from *sacro-dhO-t- " sacra faciens " 27. In that case, the first part of the com-
pound, *bibi-, *bibu-, *bebi-, *bebu-, is wide open to new etymological gues-
ses 28. Even more tenuous is the case for OIrish coimdiu " lord" which, fo11o-
l.c.).
2t Sommer suggested (IF Il, 1900, 63, 66) that tenus and secus also represented
neuter perfect participles. But this view is now generally abandoned in favour of the
neuter nouns *tenos, *sekwos; see Hofmann, IF 44, 1927, 72, 74; Leumann-Hofmann,
o.c., 517 (retaining PPA for secus, but 536 rejecting it for tenus, as also Leumann-Hof-
mann- Szantyr II 248, 267), Walde-Hofmann, o.c., II 506, 667; both rejected by Ernout-
Meillet, Dictionnaire etymologique de la langue latine, 1959t , 685.
25 Sommer, Festschrift W. Stokes, 1900, 24-5.
2S Pokorny, KZ 47, 1916, 163. Pedersen did not think that Pokorny's phonological
objections were decisive (see his Etudes lituaniennes, 1933, 49), but Pokorny repeats his
derivation in his Indogermanisches Etymologisches Worterbuch Il2. R. Thurneysen
(A Grammar of Old Irish, 1946, 124), more cautiously, regards this as a mere possibility.
It would also be important to decide whether OWelsh bibid ' rei' is compatible with
-bud-.
17 On sacerdos see Schulze, KZ 28, 1887, 281 (now in Kleine Schriften, 1934, 364) ;
*Migw-i-.
Oswald Szemerenyi
U Pedersen, Vergleichende Gramm. der keltischen Sprachen Il, 19II, 102; Etu-
des litu. l.c. In Lewis-Pedersen, O.C., 175, this is repeated, but, surprisingly and without
any explanation, p. 67, the ending is said to point to -jot-s, as also in Pedersen, Vergl.
Gram. I 254.
30 See Thurneysen, o.c., 212, who, on the strength of the Gothic form, assumes that
"month ", namely Goth. menops (a consonantal stem like weitwod-), ONorse
mono Or, OEngl. monatJ, OHigh Germ. manod. Here, too, it is usually assumed
that the -t-stem (resulting in Germanic -p-) of *menop- was inherited from
Indo-European, where it was in use side by side with an -s-stem *menes- 32.
The evidence for this statement comes from Lithuanian menuo, gen. menesio
" month ", which is said to reflect an early paradigm *menotj gen. *menes-os.
But without the temptation of Gmc. *menop- one would hardly have thought
of tracing Lith. menuo to *menot rather than *menor, a development paralle-
led by sesuo " sister " from lE *s(w)esor, or *menon, paralleled by piemuo
" shepherd" from lE *poimon 33. On the other hand, Gmc. *menop- is, like
weitwod-, completely isolated within the lE family of *menes- jmens-. It seems
clear that the Germanic development must be understood from within, from
the peculiarities found within the Germanic system itself .
. Now the history of an indubitable -t-stem, lE *nepot-s "grandson ",
is apt to throw light on our problem, too. Originally it had an ablauting para-
digm, nom. *nepos, acc. *nepot-w, gen. *nept-os, etc. The ablaut-variation
was mostly eliminated, compare Lat. nepos, nepotem, etc. In Germanic we
find that the noun has been transferred to the -n- stems: nom. *nefo (OEngl.
nefa "grandson, nephew", etc.). It is agreed that this curious change is
based on the coincidence of the normal -n-stem nominative-ending -0 with
a nom. *nefo of the word for" grandson" 34. But how does this relate to the
lE form *nepos? The answer is that the inherited nominative *nepos was
at one time replaced by *nepot, emphasizing the stem-final at the expense of
the nominatival -s, which then became *nepo, *nefo.35 It will be seen from
31 See J. Schmidt, Die Pluralbildungen der idg. Neutra, 1889, 194; Brugmann.
Grundr.1 Il I. 426f . 526; Walde-Pokomy. o.c . Il 271; Specht. KZ 66. 1939. 53. and
Ursprung der idg. Dekl., 1944, 344; Walde-Hofmann, o.c.,1 Il 71; Pokomy, Idg. etym.
Wb. 73If. ; Belardi, Rivista degli Studi Orientali 23, 1948, 76f., esp. 771 ; Makajev. in :
Sravnitel'naja gram. germanskix jazykov Ill. 1963, 266. A dissentient voice from A.
Scherer, Gestirnnamen bei den idg. V6lkem, 1953, 67, supported by Nehring, Kratylos
3, 1958, 66. On the etymon see Nehring (whose suggestion seems to point to foreign
origin, but note F. A. Wood's view, CP 7, 1912, 312) ; Brandenstein, Studien zur idg.
Grundsprache, 1952, I If. ; Risch, Festschrift Sommer. 1954, 195.
33 My own view is that menuo is based on *men6Y/menesos .. moon ", which was
enlarged from *men- under the influence of the often associated *ausor/(a)uses-os
.. dawn" ; on the formation of the latter see my note in Glotta 38, 1959. II2'. A *me-
nos or *menos is favoured by Kieckers, Sprachwissenschaftliche Miszellen, 1934, 32f. ;
Endzelynas, BaIt!! kalblf garsai ir formos, 1957, 127; Palmer, Minos 5, 1957, 63; Risch.
Museum Helveticum 16, 1959, 222f. On the Baltic forms see also Otr~bski, Gram. j~zyka
litewskiego Ill, 1956, 6If. ; Vaillant, Gram. comparee II I. 1958, I75f.
U Cf . e.g . Brugmann. Grundriss' Il 2, 1261, 128; E . Kieckers, Handbuch der
in -t, instead of -s, has never heen explained satisfactorily. Brugmann, for instance,
assumed, on the strength of the Germanic declension seen in "moqn " and" grand-
16 Oswald Szemerenyi
this example that, at one time, there existed in the Germanic system the pat-
tern -osjot-~ beside -osjos-~, and that transfers from one class to the other
must have been quite easy and fairly frequent. It certainly provides the only
plausible explanation of the change from *menosj*menes-os to *menosjme-
not-~, and later *menosjmenot-~, eventually *menot/-ot-~ 36. If this is true,
the transformation of the awkward paradigm *weit-wosjwos-~j-us-os etc. to
*weit-wos- /wot- is also explained.
The result of this survey of the Western lndo-European data is that the
alleged evidence of an lE -wot-suffix is either non-existent - this is the case
in the Italo-Celtic group - or is due to secondary innovations - this is the
explanation of the single Gothic instance.
These results are confirmed by the evidence of the remaining lndo-Euro-
pean languages. The Slavic preterite participle active is of the type: m.
nom. nes'b, f. nom. nes'bsi, n. nom. nes'b. The paradigm shows the generali-
zation of the suffix-form -us- which intruded even in the m. and n. nom. : nes'b
derives from *nes-us. The fern. nes'bsi represents *nesus-ji, with -j- brought
son " , that, beside the normal sigmatic nominative, Indo-European also had an asig-
matic type, even in the animate gender; see Grundriss 2 II I, 1906, 425, 427. But soon
after he realized (IF 22, 1907, 181 1 ; Grdr.2 II 2, 19II, 128) that this was impossible. He
now suggested that all the nouns in question either were neuters originally - halip "war-
rior" and magap " maid" - or, for a time at least, became neuters - menot, under
the influence of (Goth.) sauil "sun ", and nepot under that of barn" child ". But this
involves the incredible assumption that all these nouns later reverted to their original
gender - an astonishing feat, indeed. What is more, there is no reason for assuming
that sauil was neuter in Gothic, and not feminine, as it is in ONorse and OEnglish, and
as is the variant sunna in Gothic itself; I hope to show this elsewhere. No wonder that
quite recently Specht attacked Brugmann for his change of heart and defended the
asigmatic type (Ursprung 368), but without answering Brugmann's fundamental objec_
tion. It is only fair to add that I myself succumbed to the pressure of the prevailing
doctrine when I admitted (KZ 73, 1956, 1902) that the Germanic facts warranted the
assumption of an lE nom. nepot, instead of nepos. Nevertheless, the answer is quite
different.
We must start from the evident assumption that the lE -s-, -t- and nasal stems
at first preserved in Germanic the original ablaut-alternations of the animate inflec-
tion; they had
nom. sg. -os -os -on
acc. sg. -os-'V- -ot-"l,~ -on-'l!
gen. sg. -os-os -ot-os -en-os
(or -es-os) (or -et-os)
The agreement of the -s- and -n-stems led to the -t-stems' adopting the same pattern :
long vowel + stem final consonant in the nominative. Hence *nepos beclJ.me *nepot
(or *nefop) which eventually developed into *nefo, making the change to the -n-inflec-
tion possible. Note that *fos (lE *pos) 'foot' was also replaced by *fot.
38 It will be clear that the change from -s- to -t-inflection took place before the
processes described in the preceding footnote. A further problem is the origin of the
-s-stem in the word "moon-month" but it cannot be taken up here.
Perfect Participle Active 17
in from the oblique cases such as gen. *nes-us-jas (cf. Skt. bharantyas) ; this
j was also extended to the masc. and neuter inflection as shown by m. gen.
sg. nes'bsa from *nes-us-ja, etc 37. In Baltic, too, the suffix-form -us- was gene-
ralized, but the masc. nom. sg., under the influence of the present participles,
acquired a nasal. Thus we have Lithu. nes~s, gen. nes-us-io; fern. nom. nes-
us-i, gen. nes-us-ios, etc. 38. The new Tokharian data also confirm the -s-suf-
fix, and show no trace of a -t-suffix 39.
Since these results were reached in 1959, it seems advisable to take ac-
count of two, more recent, discussions of our problem. In his " Reflexions sur
le participe parfait actif" (Charisteria F. Novotny octogenario oblata, 196z,
71-77), A. Erhart also rejects the alleged Western European evidence for an
lE alternation -wos-j-wot- but maintains lE origin of Greek -fo't"- and Indian
-vat-. But these are traced to the lE suffix -went-, so that the paradigm of
the PPA represents an amalgam of two originally distinct formations: adjec-
tives in -wos- (from wejwo+s) and in -went- (also from wejwo but with nt).
Our discussion has shown that for lE times we cannot assume the presence
of -t-forms in the paradigm of the PPA. In view of the Mycenaean evidence
- ignored by this author - we could only admit influence of the adjectives
in -fe:v't"- in the post-Mycenaean period but by that time the difference in
function between the two formations was so great that no cross-influence
can be imagined.
An entirely different lE paradigm is postulated by Puhvel (in: Myce-
naean Studies, ed. E.L. Bennett, 1964, 171-177). But his assumption of an
lE PPA with suffix -wos-jus-, except for the neuter singular nominative which
ended in a dental, is at variance with everything we know of Indo-European.
The Indian neuter in -vat is obviously a secondary transformation - within
Indian - of the original -wos (or -us), and does not justify an unheard of
lE paradigmatic variation.
The elimination of the alleged -t-suffix from the paradigm of the Indo-
European perfect participle restores to this formation its inner unity. We
are no longer faced with the question, impossible to answer, which cases
should be assigned to the suffix -t- and which to -s-. In view of the history
of our discipline, we must consider this result as one of the major contribu-
37 Cf. Vondrak, Vergleichende slavische Grammatik 12, 1924, 668; Ill, 1928, 58;
Riga 1951, 943; J. Endzelynas, BaltlJ kalblJ garsai ir formos, 1957,204; Otr((bski, Gram.
j((zyka litewskiego Ill, 1956, 257f. Pedersen's discussion (Etudes lituaniennes 47f.)
is quite wrong.
3D See A. J. v. Windekens, Morphologie comp. du tokharien, 1944, 104 f. ; Krause-
and, of course, the feminine which preserved the inherited -ULIX in most dialects
down to post-Classical times. Secondly, a major phonetic change, the aspira-
tion of intervocalic -s- and, as I believe 43, its complete loss by Mycenaean
times, presented certain problems for post-Mycenaean speakers. The Myce-
naean paradigm can be reconstructed in the following form 44 :
(( The instrumental is omitted here, as it must have been lost from the living speech
fairly soon after the Dorian invasion.
'5 Cf. Schwyzer I 249f. ; M. Lejeune, Traite de phonetique, 19561, 23If.
U Or * d8out;;? Cf. acc. eAocnOUt;;.
" To avoid the charge of " teleological" interpretation, I would like to make clear
the theoretical basis of this argument. We must obviously assume that for some time
old and new types live side by side, in our present case acc. fEt8fo(1. and fe:t8fw, gen.
fEt8Foot;; and fe:t8fout;; etc. The community cannot discard the old forms in order to
replace them by new ones, if the latter cause difficulties in the process of communica-
tion. It is during this period of coexistence, not less than a generation, that the commu-
nity decides, by trial and error, whether a form or formation is fit for its purposes.
20 Oswald Szemerenyi
U See, e.g., Risch, Wortbildung der homerischen Sprache, Berlin 1937, 54; Schwy-
zer, 1,5 83; P. Chantraine, Morphol. 78 f. but esp. Morpho1.2 83f.; Grammaire home-
rique I, 19482, 230f. ; M. Leumann, Homerische Worter, Basel 1950, 159; J. Egli, Hete-
roklisie im Griechischen, mit besonderer Berucksichtigung der Fa.lle von Gelenkhetero-
klisie, Diss. Zurich 1954, 3d., 87f. ; Pisani, Storia della lingua greca, 1959, 521; Perrot,
Les derives latins en -men et -mentum, 1961, 325'; Shipp, Essays in Mycenaean and
Homeric Greek, 1961, 5f.; Palmer, in: A companion to Homer, edd. Wace-Stubbings,
1963, II3. Of particular importance is Gallavotti's paper" 11 nome della testa ... ",
RFIC 40, 1962, 135-149.
.. J. Schmidt, Pluralbildungen 467; Chantraine, Morphol. grec 80 (cf. Morphol.2
84f.); he is less confident at Gramm. homo I 230.
60 Att. w't'- is not based on *wcr<X't'-, from an unsubstantiated *ous (Schwyzer I 348,
520, contradicting 250), nor does Dor. WC; demand *ous. Theocritus' &.[L<jlwe:c; is now con-
firmed as a precious archaism (see Sommer, Zur Geschichte der griechischen Nominal-
komposita, 1948, IIof.) by Myc. -owe. I discuss all these problems in SMEA Ill.
61 The innovation is proved, for Knossos at least, by anowoto.
62 See Schwyzer I 520f.; Pisani, Miscellanea G. Galbiati Ill, 1951, 12f.; Bolelli,
ASNP 22, 1953, 67; Perrot, o.c., 291., esp. 322f. The -t- probably has several sources.
Thus, e.g., y<x'A<xx't'- certainly received it from [Le:At-L"-, see Szemerenyi, KZ 75, 1958,
174. I also think that Thurneysen, IF 39, 1921, 192f., was right in tracing '\)[L<X't'oc; to the
- influence of vux't'6c;, but see also Kieckers, IF 41, 1923, 184. Cp. also the text further on.
Perfect Participle Active 21
zer I 578 seems to assume a diphthongal stem .Xpw(u)o;: ~)(plXf-6o;, for which, however,
Xpo:uw is no evidence. Sommer, Nominalkomposita 21, thinks that one cannot decide
between xpocr- and Xpof-, but neither of these is tenable.
22 Oswald Szemerenyi
The neuter &UXPo~ appears in the scene where Homer describes how Eumaeus
is making sandals for himself (~ 24), 't'cX(.LVCUV 8p(.Loc ~e:mv euxpoe~. .. These
compounds clearly show that the underlying noun is either a neuter *Xpoo~
or, in view of our evidence, the masc. *XpoW<; 56. This means that the original
inflection of the noun was *Xpow~/x.po6((j)oc/Xpo6((j)o~/Xpo6((j)~. It is clear
that in the nominative the sequence -ocu- was contracted to x.pw~, while in the
oblique cases either contraction with subsequent shortening took place, or
in the sequence of three vowels one 0 was lost by hyphaeresis 57.
The -s-inflection is replaced by the -t-inflection in Xpcu't'~, Xpw't'oc, which
becomes the rule in later times. As always in such cases, we must ask for
the model of this change. Schwyzer suggested that adjectives of the type -~pW<;/
-~PCU't'~ were behind the transfer. This was rightly rejected by Egli, since
the adjectives are not of the same shape and have no semantic connection
with our word 58. But his own qlW<;, qlcu't'~ will not do either 59, since the mea-
.. Already pointed out by Witte, RE VIII, 1913, 2230, 8f. and 3d. Strangely enough,
Risch, Wortbildung 119, 203, seems to think that !J.&AIXYXpOL~<; derives from XpOL~,
and this from Xpw<;, like xlXp8L'I) from x1jp. But even XpOL~ is hardly a derivative in -L'I),
but a collective *Xpoo(cr)-iX (see Risch 13). And in view of !J.E:AIXV6)(poo.:; etc. it is impos-
sible to seek in IL&AIXYXpOL~<; anything but )(pw<;. The regular -)(po~<;, seen in EU)(poe<;,
was metrically unusable and had to be lengthened; the lengthened 0 is spelt before 'I)
as Ot as in 7jYVOL'I)crE:, &.yvot~cror:cr', see Chantraine, Gram. homo 11 99, and - pace Chan-
traine - even rrvot~ which, in view of Att. rrvo~, cannot be from *ITVOfL,x (or rather *rrvo-
nIX, see M. Scheller, Die Oxytonierung der griech. Substantiva auf -tlX, 1951, 831 ). The
fact is that *rrvOfLIX would have resulted in Attic *rrvolor:/ITV61X, and only Schulze (QE
405f., followed by Meillet, RPh 53, 1927, 197; Schwyzer, GG I 469) could have hit upon
the idea that the Homeric word was different (ITVofLIX) from the Attic (ITvof-~). The
correct view (rrvOt~ metrically lengthened from rrvof~) is stated by Kiihner-Blass, Griech.
Gram. 13 382; Bechtel, Vocalcontraction 271; Risch, Wbdg. 119. Homeric )(POL~, on
the other hand, is a spelling for Xpo~, with a from 00. The neuter EU)(poe<; is not explai-
ned by saying (Risch 74) that it was formed after or:l8w<;: &.vlXt8~<; sim., since the result
would have been EU)(p~<;-)(pe<;, as it is later in ILE:AIXYXP~<; of the comic poets. On &.Xpe:LOV,
derived by Schulze from *&.)(P~<;, transformed from *&.)(po~<; 'pallid', see now Szemere-
nyi, Sprache 11, 1966, 15-17, esp. 16. The peculiar nom. sg. IL&A<xv6)(poo<; at T 246 is,
by diektasis (Chantraine, Gram. homo I 75f.), from ILE:AlXv6)(pw<; (see Sommer, Nomi-
nalkomposita 27 and 28 end), itself contracted from -)(po'l)t;; cf. &.aTU~OW-T'I)<; <-~WT'l)C;
<-~o~T'I)<;, Chantraine 82.
61 In the nom. *Xpow<; either contraction or metrical lengthening was unavoidable.
For Homeric hyphaeresis see Chantraine, Gram. homo I 73f. This explains the nom. pI.
ILE:AIXV6XpoE:C;, acc. pI. T:XILE:crLXPoor:<; siro., which stand for -)(POE:E:C; etc., as &.XA&&t; for &.x-
Ae:ee:c; etc., and resolve Sommer's difficulty, O.C., 26. Neither contraction nor hyphae-
resis would be " endangered ", even if the vowels had originally been separated by a
digamma. A stem *Xpofoo- Was thought by the decipherers to be established by Myce-
naean akorowee, interpreted as the dual/plur. nom. of *&:-)(pwf1)C; "of uniform colour"
or *&'-Xpw>"1)C; "colourless, pale" (see, e.g., Documents 76f., 387; Lejeune, RPh 84"
1958, 207 ; Doria, Parola del Passato 70, 1960, 49 '7 ) ; but the derivation has been que-
stioned by, e.g., Gallavotti, Parola del Passato 52, 1957, 9 (: Gtxpo<;+OUc;).
18 Egli, O.C., 59.
i t This was suggested by Risch, \Vortbildung 15, accepted by Egli, I.c.
Perfect Participle Active 23
ning is again very different. It seems clear that a word, or words, describing
part(s) of the body is (are) required, and I believe that the source was xpoc't"6c;;
note that (as if in exchange ?) we find the" monstrous" acc. XpOC't"1X at e 92 60.
Less important are, from our point of view, YEAW't"-, epw't"-, t~pw't"-, since,
although they are undoubtedly transformed from -s-stems, the change is
posthomeric; noteworthy is nonetheless the acc. t~pW't"1X appearing at Hes.
Erga 28961. Finally, for EUpW<;, EUpW-roc; an earlier -s-stem is proved by Horn.
EUPWEV't"IX, where -w- is from -os-w-.
These examples clearly show that an -s-stem inflection, as tending to be
obscured by inevitable contraction, was in many cases early replaced by a
-t-declension. This applies to masculines as well as neuters, whereas feminine
-s-stems are either immune (IXt~WC;) or choose a different way out (e.g. Att.
~wc;). The more interesting is the case of XcXPLC;, which, although undoubtedly
an -i-stem originally, already in Homer normally inflects as XiXpL't"- 62. One
is almost tempted to say that t performs the role of a " Hiatustilger ", but
this would rather obscure the complicated processes that result in the even-
tual word-shapes produced by refashioning.
In the light of these parallels it will be easier to understand how the
original -s-stem inflection of the perfect participle came to be replaced by
the -t-stem declension. The details will probably never be cleared up. But it
is worth pointing out that for the consolidation of the -t-stem 63 the present
participles were probably also responsible, especially if they preserved an
original variation -WV/-IX't"- (lE -~t-) down to (post- ?) Mycenaean times 6, .
0 Note that <pw~ " light" also became a -I-stem after Homer. But the poetic
b <pw~ (Egli 6of.) cannot have been the model for this ordinary household word. The source
is clearly ax6't"o~, which, originally a masc. -a-stem, became a ntr. -s-stem under the
influence of <p&o~, and formed &v ax6't"c:~ after b <pae~, axoT-e:tv6<; after <plXt~v6~, axo-
't"e:Lv6~ in its turn producing <pc.,m:~v6~ (Egli 67f.). The original inflection <pwc;, gen. <pw~
<
(<pato~), dat. <pWL (from <pii~ <pat\ under the influence of -w-) was untenable in any case.
The starting point may have been the gen. pI. ax6't"wv, producing <pw't"wv, etc. Because of
Schwyzer's curious remark (I 578), note that <PW\ does not warrant any other source
beyond <p.xo~; also that" Att ... xpw~ is not" older" than Xpo(, but transformed from
Xpor after the long vowel of Xpw~ etc., although here our view might seem to support
XPO\ from Xp06((J)~.
n See Chantraine, MorphoI. 64, Gram. homo IS 211; Schwyzer I 514; Egli 59f.
I add here that the famous &~ lpov gv-ro originally had lpo(lX) and lpo'J is not an Aeoli-
cism, but elimination of the hiatus. On the disputed question of" Aeolic" Y&AO~, ~poc;,
see Sommer, Nominalkomposita 264 , with references. On !8pw<; see also the paper refer-
red to in fn. 50 .
Risch, Wortbildung IS .
ea Schwyzer (I 540) suggests that the type ,xyvwc;/-w't"o~ may have contributed.
But the difference in the vowel-quantity between d86't"oc; and ,xyvw't"oc; is not encou-
raging.
" An especially nice example would be &At~<PIX/,xAd<pIX't"OC;, a neuter participle in
-tlt/-tl/-os, from cXAtt<pW, literally "(the) anointing (thing)", now found in Myc. arepa
Oswald Szemerenyi
PPA PresPA
pI. nom. -foec;/-folX -ov'rec;/-rx:rlX
acc. -folX<;/-folX -OV'r1X<; (-IX'rIX<;?) / -1X'r1X
gen. -fowv -1X'rWV
dat. -foO"cn -IXO"O"~ (cf. ~V'rIXO"O"t)
presented one obvious point of contact in the dat. pI. The parallelism of
-IXO"O"~/ -foO"O"~
easily led to the transformation of] the genitive to -O'rwv after
-1X'rWV. The next steps may have been either extension to plural -0'r1X<;, -onc;,
-0'r1X, or singular gen. -O'rO;, dat. -O'r~, and finally acc. sg. -0'r1X.
This scheme naturally presupposes a state of affairs when the present
participle still showed - more or less - the inherited ablaut-variations,
a problem that will be discussed elsewhere.
Before concluding, two more points should be briefly mentioned.
The Mycenaean name widuwojo, widowoijo 85 represents a f~~fo~cx; and
is certainly derived from f~~fo(O")-. On the face of it, it looks as if it were
derived fron the oblique stem of the masculine. Lejeune thinks it also possi-
ble that it started from *wid-us-yo-s which was refashioned after f~~fo(O")-;
the latter form would survive in Hesychius's 8ui:o~' !L~P'rUPeC; and Lac.
~~8u~0~ " (boy) prefects" 88. As pointed out by Schwyzer, it is, in view of the
meaning, very unlikely that this 8ui:o~ should be masculinized from ~ui:1X
nom., a1'epate instr.-dat. The nom. lXAe:t<pOCp is a late innovation as ;was surmised by
Schwyzer (I 520) on the basis of the historical data. The assumption of an old hetero-
clitic 1'/n-stem (recently by Lejeune, REG 72, 1959, 139; Ruijgh, Mnemosyne 14, 1961,
205; Gallavotti, RFIC 40, 1962, 142) is disproved by the absence of deverbative cases.
This example is certainly in favour of J. Schmidt's theory (Pluralbildungen 187f.) that
-{LOt/-(L-rot;j was re-formed under the influence of neuter participles in -/--ror;. I add
here that Hom. oM~ cannot be from 8&xvw (Boisacq s.v.; Risch 308; Schwyzer I 620,
II 491; Chantraine, Gram. homo I 250), since in that case we would have o8oy~.
Like M~, 7tU1;, it is from a nominal case: o8~ replaces o8cXO'O'(t), the old dat. pI. of
08cJ>v, the use of the dat. pI., and its archaic form, being paralleled by &yx&r; from &y-
XIXO'(t), see Risch 305; Chantraine, Studi italiani di filologia classica 27-8, 1956, 106.
And just as A&~, m)~, originally dat. pI. of ACXX-r-, 1tuy-, changed *yv&r(t) from y6vu to
yW!; so o8iO'O', too, became under their influence oM~; see Sprache Il, 1966, 20 71 A
further trace of the participial weak-grade may be the Homeric lx71-rt &X71-rt which
cannot be formed with -..fi-r- (Brugmann, IF 17, 1904, 9f. ; Schwyzer I 550') - the result
would be *excxa-rfi-r- - but, as suggested by Pisani (Rendiconti Accad. Lincei 6/7, 1931,
76f.), is metrically lengthened from (&)exii-rt, loco of (&)exwv; differently Leumann,
Homerische Warter, 1950, 25If.
la Documents 427; A. Sacconi, Rendiconti dell'Accademia dei Lincei 8/16, 1961,
276 .
.. Memoires de philologie mycenienne 224 fn. 14; 263 .
Perfect Participle Active 25
which is not attested in the meaning" witness" 67. It is more likely that
the late ~L8uLOL, ~L8EOL were modified from an earlier fL8(f)OLOL - identi-
cal with the Mycenaean form - at a time when the sequence -OLOL was
replaced by -ULOL (under the influence of -u!:ot?) and this by -ELOL 68. But
the important point is that the form *widwos-, now attested for the mascu-
line, raises, and settles, an interesting problem. Hitherto we have assumed
that in the perfect participle the root-form feL8- was used in the masc. (and
neuter), fL8- in the feminine. This view must now be revised. Some cases of
the masculine, perhaps the weak cases, had the weak form of the root: feL8fw<;
but fL8u(a)o<;, or, by later levelling, fEL8fw<;-fL8fo(a)o<; 69. This would cer-
tainly apply to the plural also.
Another problem is also concerned with the root-form of feL8fw<;. It
is generally assumed that the Indo-European perfect had a reduplicated
formation as well as a type without reduplication 70. But, as was emphasized
by Leumann a short time ago 71, the latter type is only found in the verb
*woida, which is not the' principal' representative of the type (Meillet),
but the only representative. Now in view of the well-known fact that verbs
meaning" to know" are frequently used, and as a result often deviate from
the normal development 72, it is justifiable to doubt the assumption even
for this case. In other words, if *woida lacks the reduplication, it is due to a
previous and comparable change in Early Indo-European. Now the expected
form at that stage was *wewoid- in the perfect singular, *wewid- in the dual
and plural. The participle was *we-wid-w6s/*we-wid-us-y~, with the nil-grade
necessitated by the accession of a new derivative element 73. Both forms
were obviously very frequent. Hence the perfect could easily succumb to
general tendencies, such as assimilation of *wewoid- to *wowoid- with sub-
sequent reduction in rapid speech to ' *woid-; the dual/plur. *wewid- would
87 Schwyzer I 540. For this reason I cannot accept Leumann's interpretation, Cel-
71 Compare, for instance, French sais from *sayo, instead of *sapyo; see M.K.
Pope, From Latin to Modern French, 19561, 358; H. Flasdieck, ZRPh 74, 1958, 72f.
73 Leumann's paper in Celtica 3 is important in settling this question for Greek.
In discussing Schwyzer's claim that in Greek there is no evidence of the digamma in
-wot-, Leumann (p. 246) seems to overlook MM't1X, although he convincingly shows (p. 248)
that this represents fef~3f6-ra:. Now the Mycenaean evidence settles this question too. -
Kurylowicz, Apophonie 105, regards e:t3w~ (and even &p1j?W~ etc.) as archaisms present-
ing the full-grade of the root, although he had just stated that, where the accent had
always been on the ending, the root Was weakened: we should therefore expect *wid-
w6s- even on his theory.
26 Oswald Szemerenyi
naturally follow suit, unless at that time, before the period of weakenings, -
there was only one form, *wo(wo)id-, for all numbers. In the participle the
sequence of three w-s would again easily lead to a dissimilated form *we-
(w)idwos, thus giving rise to a new stem-form *weid-. If the pluperfect and
other forms were current already in lE times, they, too, must have been
based on *we-wid-, and adopted the new stem-form *weid- (e.g. Gk. 1)-fe:L8-E~
etc.). The advantage of this theory is that it can account for the curious Greek
masc. participle fEL8-f~ 74. If this represented an lE type without redup-
lication, it would be impossible to understand how the root-syllable escaped
the consequences of the fact that it was unaccented. It is clear that the accent-
relations only permitted the variants *woid-Jwid-, there was no room for a
third variant. It is also useless to suggest that there might have been" other"
forms where the normal-grade was justified. Even if there were formations
like the subjunctive, the moods never influence the participial formations.
But if *weid- came into being in the way described, it is easy to understand
that at a later stage the unusual root-form could again be dropped in favour
of the well-established nil-grade. This is shown by Skt. vid-vas-, in contrast
to the Greek and Gothic forms.
Enough has been said to show the importance attaching to the few
perfect participle forms that have been vouchsafed to us by the Mycenaean
finds. We can now see that the alleged Indo-European parallels of a promi-
scuous -wos Jwot-suffix were a mirage. The internal origin of the Greek inflection,
on the other hand, makes it imperative to try to understand the develop-
ment of the lE legacy within the peculiar frame-work of historical Greek.
The parallels in historical and prehistoric Greek will have shown how the
details of the development must be placed in this context 75
.. The peculiar root-form of this participle has often tempted scholars to posit
*fe:f~1)w<;as its primitive form. This is of course impossible (see Schulze, KZ 27, 1885.
547f. now Kleine Schriften , 1934. 109f.; Schwyzer I 5408) ; the change took place in
Indo-European. I am glad to be in agreement on this point with V. Pisani (AGl 41.
1956. 158') but this is not" una bella e certo antica isoglossa greco-gotica ". nor is the
reduplication secondary. L.R. Palmer (Sprache 5. 1959. 131) also notices the importance
of Myc. widwoijo but thinks that this necessitates a revision of the view that the masc.
participle had the stem-form fe:~1)- ; he does not. however, offer an explanation for Horn.
Att. fe:~1)w<;. As to the fate of the Early lE *wowoida. I should here draw attention to
the fact, recently emphasized by Bulachovskij (lzvestija Akademii Nauk, Otdelenije
Jaz. i Lit., 18, 1959, Ill) that Slavic vedati with its acute accent requires an lE *void-.
This cannot represent an lE lengthened -o-grade. but could very well be from wo(w)oi-
da> w6ida.
16 After the first draft of this paper had been wtitten down in 1959. I saw that
Chadwick also assumes (The Decipherment of Linear B. 1958, 95) that Greek -wot- is
a post-Mycenaean innovation - a gratifying agreement. But his claim that -wot- is
.. not found outside Greek" contradicts appearances at least, and he has made no at-
tempt to disprove the" evidence" as has been done here. Valiant, who also reaches
this conclusion (Gram. camp. II 2, 554f.). assesses somewhat differently the evidence
of Sanskrit, Greek and Germanic, while Pisani. though aware of the Mycenaean facts,
still adheres to the old view (AGl 41, 1956, 158; Paideia 15. 1960, 227).