Sie sind auf Seite 1von 24

On the Indo-Iranian Accusative Plural of Consonant Stems Author(s): Hans Henrich Hock Source: Journal of the American Oriental

Society, Vol. 94, No. 1 (Jan. - Mar., 1974), pp. 73-95 Published by: American Oriental Society Stable URL: http://www.jstor.org/stable/599732 . Accessed: 22/01/2011 15:43
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at . http://www.jstor.org/action/showPublisher?publisherCode=aos. . Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

American Oriental Society is collaborating with JSTOR to digitize, preserve and extend access to Journal of the American Oriental Society.

http://www.jstor.org

ON THE INDO-IRANIAN

ACCUSATIVE

PLURAL

OF CONSONANT

STEMS

tIANS HENRICH HOCK


UTNIVERSITY OF ILLINOIS

The predesinential 0-grade of the Indo-Iranian accusative plural of consonant stems has by most earlier linguists been considered inherited from Proto-Indo-European. In more recent publications, however, the view prevails that it must be of secondary origin. The purpose of this paper is (a) to show that the latter view is more likely to be correct and (b) to give a more plausible morphological motivation for the secondary nature of the predesinential 0-grade than offered in previous publications.

1.1 As Is WELLKNOWN,Indo-Iranian1 regularly forms the accusative plural of the consonant stems with a predesinential 0-grade, as can be seen from the following partial paradigms:
N A N A sg. sg. pl. pl. rajd rfjanam rajanas rajnas pita pitdram pitdras pitrn.

to be more widely accepted, gives historical priority to the Greek full grade; cf. 3.1-9 below.2 1.4 The purpose of this paper is to show that the second of these two views is more likely to be correct and to give a more plausible motivation than previously offered for the secondary nature of the predesinential 0-grade in the accusative plural of Indo-Iranian. 2.1 The first linguist to have attempted to account for the predesinential 0-grade of IndoIranian seems to have been Pott. According to him, the Indo-Iranian state of affairs arose in Proto-Indo-European as a result of the 'heavy' ("gewichtvoll," "schwerfillig") phonetic nature of the accusative plural ending -ns (1833: 52) or -mas (1836: 15). Unfortunately, however, Pott's view on the subject cannot be considered consistent, for in a different passage (1833: 61) he claimed that the 0-grade of the accusative plural serves to distinguish that case from the nominative plural with its predesinential full grade. In addition, Pott does not seem to have tried to account for the Greek state of affairs.
2

1.2 This contrasts with the situation in Greek, where in the only class of consonant stems which paradigmatic preserve the Proto-Indo-European ablaut alternation between extended, full, and 0-grade, namely the kinship r-stems, the accusative plural regularly is formed with a predesinential full grade, as in the following partial paradigm:
N A N A sg. sg. pl. pl. pater patera pateres patdras

1.3 Concerning this discrepancy between IndoIranian and Greek, two major views have been advanced. According to one, espoused mainly by earlier writers on this subject, Indo-Iranian preserves the original state of affairs; cf. sections 2.1-12 below. A second view, which appears currently 1 Although what will be said in the discussion of this paper would seem to hold true both for Sanskrit and for the ancient Iranian dialects, examples will be drawn almost exclusively from (Rig-Vedic) Sanskrit, since in that dialect they are more numerousand readily available and since the problem of the marking (or fate) of length in Avestan, the better-attested ancient Iranian dialect, would render the data from that dialect not sufficiently reliable for some of the more important arguments of this paper. 73

There is yet a third attitude toward the mentioned

discrepancy between Indo-Iranian and Greek, an attitude

which however can only be described as equivocating on


the issue of historical priority. This attitude can be dis-

cerned in Hirt 1927: 68-9. It is found even as recently as in Szemerenyi 1970, where on p. 147 the accusative plural is consideredan originalstrong case, while on p. 151
and 154 the original accusative plural of the n- and ntstems is given both with predesinential full grade and with predesinential 0-grade ("dnt-ns (dont-zns?)" and "-(e)n-ns"), and on p. 156 the original accusative plural of the r-stems is given only with predesinential 0-grade ("-r-ns").

74

Journal of the American Oriental Society 94.1 (1974) (ibid.). Osthoff saw further confirmation of his interpretatioii of the Sanskrit evidence in the Rig-Vedic coexistence of the accusative plural forms uksnds and uksdnas:
"jenes, ukshn-ds, ist die altere, vom nom. plur. noch erheblich abstehende, dieses, ukshdn-as ..., die dem nomin. ved. nkshdc.-as vollig gleich gemachte form." (37).

Because of his failure to account for Greek, because of his inconsistency, and especially because his views on the phonetic origination of the 0grade can no longer be maintained in view of our present knowledge about Proto-Indo-European ablaut, Pott's views on the subject cannot be It should however be considered satisfactory. noted that his claim that the 0-grade found in the Indo-Iranian accusative plural served to distinguish that case from the nominative plural will turn out to be quite plausible; cf. 2.3, 2.4, 3.1, 3.8 below. But, as the discussion of this paper will show, this differentiation must have taken place in Indo-Iranian, not in Indo-European. 2.2 Apparently without knowing of Pott's views, also Osthoff (1876) claimed that the Indo-Iranian predesinential 0-grade is inherited from ProtoIndo-European. Unlike Pott, however, he did try to account for the Greek predesinential full grade, as well as for isolated (Rig-Vedic) Sanskrit accusative plural forms with predesinential full grade (such as uksd.nas), as secondary replacements of the original 0-grade formations. In addition, he offered a more plausible phonetic motivation for the (origination of the) predesinential 0-grade; for according to him, it is intimately tied up with the fact that in the regular formation of the Sanskrit accusative plural, it is the case ending -as which carries the accent. That is, both in terms of accentuation and in terms of stem formation, the Sanskrit accusative plural regularly is a weak, not a strong case (35-6). As for exceptions like Skt. vacas and Gk. opas, with accent on the predesinential element, rather than on the ending, Osthoff believed that they can owe their existence to a general Indo-European tendency to shift the accent forward, away from the ending. An additional motivation for this particular instance of accent retraction lay, in his opinion, in the common influence of the nominative on the accusative (36). The possible alternative proposal that the 0-grade of the accusative plural, as well as its accentuation on the ending, may be a secondary phenomenon, motivated by an attempt to differentiate nominative and accusative plural,3 must in Osthoff's view be rejected, since such a differentiation would be without precedent
3 This is, of course, what Pott had proposed (for Proto-

Concerning the predesinential full grade in forms like Gk. pateras, Osthoff believed that one must assume that the attempt to make the accusative plural more similar to the corresponding nominative must have taken place quite early in 'European', resulting in the fact that in common 'European' the accusative plural was a strong case
(37).4

Osthoff found corroborating evidence for the correctness of his hypothesis in the fact that it permitted him to account for the 'European' difference between the vocalism of the nominative plural ending (-es) and that of the accusative plural ending (Gk. -as), a difference which he believed would otherwise be inexplicable; for the absence of any trace of a nasal in the accusative plural ending -as of Sanskrit, "wo doch dergleichen spuren durchaus zu erwarten wdren," makes the (then) usual assumption of a Proto-Indo-European ending -ans unlikely. In Osthoff's opinion, both the nominative and the accusative plural
endings are to be derived from a Proto-Indo-Euroof a to e in pean ending -as, with "weakening" the unaccented plural, ending of the nominative but with that weakening blocked in the accented ( ) plural (38-9). ending of the accusative 4 "Zwei ganz vereinzelte, aber auch ganz unsichere spuren" of the original state of affairs may, in Osthoff's view, be seen in the Greek accusative plural forms kunas and drnas (38; cf. also 75-77). In view of the fact that also the corresponding nominative plural and accusative singular formations have predesinential 0-grade (cf. kines, drnes and kina, drna) and that in these formations, which beyond any doubt were strong cases in ProtoIndo-European, the predesinential O-grade must be secondary, generalized from the oblique forms of the paradigm, Osthoff was certainly correct in considering kdinas and drnas very uncertain evidence for the assumption that the accusative plural originally was a weak case. For like the predesinential 0-grade of the nominative plural and accusative singular, that of the accusative plural may be secondary.

Indo-European). But as pointed out earlier, Osthoff apparently was unaware of Pott's views.

HOCK: Indo-lranian 2.3. Osthoff's views were attacked in the same year by Brugmann (1876) who raised the following, generally quite valid, objections. (1) In Vedic, the ablauting nouns frequently show the accusative plural in the strong form, both in respect to accentuation and in respect to ablaut; cf. dpas beside apds, ukscinas beside uksnds, vfsanas beside vtfsnas. In nonablauting monosyllabic nouns, Vedic sometimes accentuates the accusative plural as a strong case, sometimes as a weak case; cf. vacas beside vacds (289). "Nun entspricht dem ved. ukshdnas haarscharf got. Wenn auhsans und vakas ... eben so dem gr. 6pas . demnach ALLEidg. Sprachen den ace. plur. als starken Casus kennen und nur das Arische neben dem allVerhaltniss noch ein anderes, gemein indogermanischen ihm eigentiimlichesaufweist, so ist von vorn herein die Wahrscheinlichkeit daftir,dass diese rein arische und in den arischen Sprachen nicht einmal regelmassig auftretende Bildung die unursprtingliche ist." (ibid.). (2) It is unlikely, and without any Greek precedent, that the accent on the accusative plural ending *-as should have prevented a "weakening" of a to e. As a matter of fact, the "weakening" of a to o in the genitive singular ending *-as, where both Sanskrit and Greek agree on having the accent on the ending, as seen in Skt. vacds: Gk. op6s, clearly militates against Osthoff's view (28990). (3) The ending of the accusative plural must originally have been -ams, that is, the accusative singular ending -am plus a plural marker -s. This is especially clearly shown by the Greek evidence (290). (4) Beside accent retraction from the ending to the stem, which Osthoff had considered to be such a common tendency that it would plausibly account for forms like Gk. 6pas, one can also encounter accent shifts in the opposite direction, as can be seen by Greek proper names like Dexamenos.
"Es liegt hier sehr nahe die Annahme, das Altindische

Accusative Plural

75

(5) Finally, Brugmann did not consider very cogent Osthoff's argument that the common tendency is to make the accusative similar to the nominative and that therefore it is unlikely that the predesinential 0-grade owes its origin to differentiation of the accusative plural from the nominative plural. For other circumstances, which to be sure are not at all clear as yet, may have favored the assimilation of the accusative plural to the weak-grade oblique forms of the paradigm (292). 2.4 To Brugmann's arguments against Osthoff's hypothesis may be added the following. (1) Osthoff's claim that an attempt to differentiate nominative and accusative (plural) would be unprecedented cannot be accepted. Witness the well-known Slavic differentiation of accusative and nominative in the animates by the replacement of the accusative by the genitive, a process which began in the o-stems where phonological developments had led to the formal identity of nominative and accusative. Compare Kurylowicz 1964: 221-5 for some of the details and the syntactic motivation of this development. Surely, if phonological merger of nominative and accusative (singular) can lead to a remaking of the latter category in Slavic, so as to differentiate nominative and accusative, then an analogous phonological merger of nominative and accusative plural in the Indo-Iranian consonant stems may well have led to an analogous differentiation (although, of course, the details of the phonological and morphological developments in the two language groups may differ considerably). (2) The fact that accusative plural forms like Sunds (from Suvdn-) show the accent on the case ending is not necessarily probative. For whether the weak-grade accusative plural be original or secondary, in the accusative plural of most characterized consonant stems with underlying stemfinal accent (such as suvun-), there is no other place for the accent to go but on the case ending, since the stem-forming suffix (-n-) appears in consonantal shape and thus is not capable of bearing the accent. And it is only in the characterized consonant stems that the case ending regularly carries the accent. In the (nonablauting) root nouns, the accent can in Vedic be found both on the root and on the ending; cf. Brugmann's examples. As a matter of fact, in nonablauting root nouns, the accent regularly falls on the root; cf. de Saussure's view in the following section of

habe, nachdem durch den lautgesetzlich notwendigen Wegfall des Nasals [which is the main subject of Brugmann's famous paper] die Form des ace. plur. und die des nom. plur. zusammengefallenwaren, den
Accent im ace. plur. nach der Analogie anderer casus obliqui auf die Schlussilbe geworfen, um dadurch eine Differenzierung herbeizufiihren, wie doch wol auch [Gk.] Dexamends gegeniuber [pple.] dexdmenos auf einer solchen Differenzierung beruht." (291).

76

Journal of the American Oriental Society 94.1 (1974)


si ... (l)a syllabe pred6sinentielle etait AFFAIBLIE
..." (208).

this paper, as well as the discussion in Kurylowicz 1968: 27, for which see section 3.8 below.5 What is especially important is the fact, indicated in Debrunner-Wackernagel 1930: 24-5, that where in the nonablauting root nouns "neben den stammbetonten Formen auch endungsbetonte vorkommen, sind diese gew6hnlich seltener und
SPATER

MALGRE L'ACCENT [de Saussure's

emphasis]

[my emphasis]

. . ."

In some cases,

such

as RV hrutas (6.4.5) and srdcas (1.144 1), the forms with accent on the ending first occur in AtharvaVeda; cf. hrutds (AV 6.51.1) and srucds (AV 18.4.2). Moreover, according to Debrunner-Wackernagel (ibid.), PSanini prescribes end accent for forms like niSds. That is, as the extant historical evidence of (Vedic) Sanskrit shows, the root-accented forms clearly are archaisms in the language, archaisms which slowly are replaced by the synchronically more regular forms with end accent (which in addition to being more regular have the virtue of being differentiated from the corresponding nominative forms).6 2.5 Brugmann's view in turn soon came under attack by de Saussure (1879: 208-9) who agreed with Brugmann on only one point, namely that the Sanskrit accentuation of the accusative plural ending -as is most likely to be secondary, considering that in the Rig-Veda, the accusative plural forms of root nouns are more frequently accented on the root than on the ending, and that Greek always has root accent in the accusative plural of root nouns. Concerning the question whether the accusative plural originally was a strong or a weak case, however, de Saussure asked "Comment cet affaiblissementsystematique de toutes les espices de themes sanskrits a l'accusatif plur. seraitil dfi au hasard d'un remaniement secondaire? Comment, en particulier, expliquer la forme des themes a

As for the n-stems, de Saussure asserted that their accusative plural should originally have been uks.ns or perhaps rather uksnnns. And the latter form may well be reflected in Vedic uksdnas. 'En tous cas uksnds n'est pas un type pur.' (208-9). 2.6 Although de Saussure's agreement with Brugmann's view that the case ending was not originally accented in the accusative plural must be welcomed as a step in the right direction, the remainder of his argument is open to serious objections. First of all, it is difficult to see why the ProtoIndo-European accusative plural of the n-stems should have been *uksns or rather *uksinnns, if that of the r-stems is taken to have been *patfns. Should one not equally expect an n-stem accusative plural *ukshns, with nonsyllabic, rather than syllabic, accusative plural ending *-ns? Secondly, de Saussure's claims notwithstanding, it is quite possible to consider Skt. pitfns and similar Iranian forms (such as Av. niatdrFs) to be secondary replacements of an earlier type *pitrds, which in turn might be an innovation for an earlier *pitdras (whether along Brugmann's or some other lines). For, as Bartholomae (1894: 123) has shown, a new accusative plural in *-r-ns could be created on the model of the i- and u-stems:
I pl. L pl. A p. *-i-bhig : *-u-bhi ,: *-r-bhi ' : *r-Su : *-u-su *-i-su : *-u-n? : X = *-r-nS. *-i-nS

In Indic, but not in Iranian, the accusative plural formations of the i-, u-, and r-stems additionally came under the influence of the a- and i-stems, leading to the lengthening of the stem-forming suffix and to the secondary gender distinction masculine *-i-ns, *-u-ns, *-r-ns and fembetween ... ne se forme Cette concoit qu'en liquides, pitFn? inine *-i-S proposal Dans la supposi,*-, --.-Bartholomae's partant de l'indo-eur. pAtr-ns ... tion de M. Brugman on ne pourrait attendre en san- appears quite attractive, especially considering the following facts which Bartholomae failed to skrit que 'pitrds' (pour '*pitdras', '*pitdrns'). Ainadduce in favor of his hypothesis. (1) The r5 Note in this respect that the accusative plural pitrif stems share with the i- and u-stems the feature of actually shows the accent on the stem-formingsuffix, and being the only synchronic resonant stems, withl not on the ending. But, of course, here as in the case of regular surface alternation between vocalic and forms like gunds, the placement of the accent is merely nonvocalic resonants, while the n-stems, with their a consequence of the surface phonetic structure of San- surface alternation between a and n, no longer follow this pattern. (2) Beside the i and u of the skrit. 6 For the very different conclusions reached by De- i- and u-stems, the (a)r of the r-stems is the only brunner and Wackernagel concerning this chronological inherited stem-forming suffix which conditions the change of a following s to s (in accordance distribution of forms, cf. fn. 12 below.

HOCK: Indo-Iranian

Accusative Plural

77

with the ruki-rule). (3) Additional evidence for this close affinity between the r-, i-, and u-stems can be seen in the later, exclusively Indic, transfer to the r-stems of the genitive plural ending -narm from the i- and u-stems (where it was in turn secondarily introduced from the a-stems). That this transfer was rather late is seen by the existence in Rig-Veda of the relics nardm and svusram and by the fact that the ending -nam is lacking in the Iranian inflection of the r-stems.7 De Saussure's argument that the type pitfns must be inherited and that the type uksnas must be secondary thus cannot be considered cogent. 2.7 By 1892, Brugmann had changed his mind about the original shape of the accusative plural of consonant stems. He now believed that the predesinential 0-grade of Sanskrit is more archaic than the full grade of Greek, as seen by correspondences like Skt. catuiras = Li. keturis (A pl.), Gk. Aeol. pisuras. "Doch ist diese Frage vorlaufig noch eine offene." (671-2). Brugmann reiterated this view in 1904: 294, fn. 1, and 1906: 32.8 In the latter publication, Brugmann became quite explicit about the fact that he believed Skt. caturas, Li. keturis, Gk. Hom. pisuras to be relic formations. And apparently because of this belief, he now presented without any of the previous qualifications his view that the accusative plural originally was a weak case. 2.8 Brugmann's view that the set caturas, kIeturis, pisuras is a set of relics and thus proves that the Proto-Indo-European accusative plural of consonant stems was a weak case cannot be accepted. In Sanskrit, caturas simply follows the regular synchronic rule that the accusative plural is made from the weak grade of the stem. And if it can be shown that that general rule is an IndoIranian innovation, then catiras is an innovation, just like rajnas and pitfn. That is, caturas cannot be considered any more probative than any
7 On this transfer of the ending -nam to the genitive

other weak-grade accusative plural of consonant stems. It cannot be accorded the status of INDEPENDENT evidence in favor of the view that the accusative plural originally was a weak case. Considering that in the Lithuanian inflection of the numeral 'four', the stem ketur- occurs throughout the whole paradigm, even in the nominative (plural) keturi (m.), keturios, it must be

concluded that in pre-Lithuanian there was leveling of the 0-grade stem alternant throughout the paradigm. The accusative (plural) masculine keturis therefore need not be any more original than the corresponding nominative (which definitely is not). Even in Greek, it is not at all certain that pisuras must be considered an archaism. For in addition to pisuras (Hom. I1. 15.680, 23.171, 24.233; Od. 22.111), there is also a nominative pisures (Od. 5.70, 16.249).9 It is true, the nominative is restricted to the later Odyssey, and it might therefore be claimed that it is a later, artificial formation built by an Ionic poet on the model Ion. tessaras: tessares = 'Aeol.' pisuras: X. However, the restriction to the Odyssey may well have been accidental, for in relation to the occurrences of the usual Homeric stem tessar-, the occurrences of the stem pisur- are extremely limited in the Homeric poems. Apparently the latter stem was favored only under certain metrical conditions; cf. Chantraine 1942: 260.10 In addition, as long as we do not precisely know what the nominative corresponding to pisuras was at the time of the
composition of the

to tell whether or not the pisur-dialect had leveled the weak stem throughout the paradigm, just like Lithuanian and, more importantly, just like all the other Greek dialects (with their uniform tessar-, tesser-, or tettar- depending on the dialect). The form pisuras thus cannot be considered reliable evidence for or against Brugmann's view.
9 Other forms of the stem pisur- are restricted to later

Iliad,

it is simply

impossible

plural of the r-stems, cf. Debrunner-Wackernagel 209-11 with references to earlier literature.

1930:

8 For some unexplained reason, Brugmann in these

two publications credits Osthoff 1907: 117-8 (fn.) with having pointed out the probative nature of forms like Skt. catdras, Li. keturis, Hom. pisuras. However, the passage in Osthoff referred to does not make any such statement, nor did Osthoff make any such statement elsewhere in his paper.

their existence to metrical considerations. That is, it


could be claimed that possibly none of these occurrences stem pisur- in the "Aeolic" source dialect.

the nominative pisures of the Odyssey is a poetic artifact, it might be claimed that ALL Homeric occurrences of the stem pisur-, not only those in the Odyssey, are poetic artifacts, part of the Homeric 'Kunstsprache,' and owe

(epic) poetry. 10 As a matter of fact, instead of merely claiming that

may reflect the original paradigmatic distribution of the

78

Journal of the American Oriental Society 94.1 (1974) That is, if the usual view is correct that in ablauting morphemesl3 there is a regular correlation between accentuation and full grade, then the accusative plural would as a matter of fact seem to be more likely an original strong case than a weak one. However, in light of the fact that the genitive/ablative singular of the r-stems shows the form Skt. pituir, from *palts, with accented 0-grade of the ablauting suffix *-(e)r-, this argument is perhaps not as powerful as it might appear at first blush.l4 Still, in light of the strong evidence in favor of an original predesinential accentuation it cannot be considered warranted to posit an original accent on the case ending and to consider the predesinential 0-grade a consequence of the original lack of accentuation on the predesinential element.'5 2.10.2 Leaving aside the accusative plural forms vfs(a)nas and uksd.nas, uksnds, which will be dealt with in section 2.10.3, (Debrunner-)Wackernagel's evidence for a prevocalic 0-grade alternant (*-nn>) -an- even in the simple n-stems, occurring after a single consonant, preceded by a long vowel or even by a short vowel, is very slim indeed. omdnai really is not a simple n-stem, but rather a man-stem (from av-); cf. Mayrhofer 1956: 133. It thus follows the overwhelming Rig-Vedic pattern of man-stem inflection, with prevocalic *-mnn- > -man- after long vowel in the 0-grade forms of the paradigm.l6 The form therefore
"weil sich der Akk. nur durch den Akzent vom Nom.

2.9 Also Debrunner and Wackernagel (1930: 20, 60, 276) believed that the accusative plural originally was a weak case.ll Unlike de Saussure, however, they were of the opinion that it was the ending, rather than the predesinential element, which carried the accent. Apparently the reason for their belief that the accent originally was on the ending is the fact that this would motivate the preceding 0-grade (in terms of the by now established usual correlation of 0-grade and lack of accentuation). However, as Debrunner and Wackernagel themselves pointed out, what is disturbing is the fact that the case ending -ns appears in the 0-grade, in spite of its being accented. As for forms like Vedic uksdinas beside uksnds, Debrunner and Wackernagel do not consider them valid counterevidence to their hypothesis. For in their view they are to be explained as reflecting an earlier *-nn- (> -an-). In support of this view they referred to Wackernagel 1896: 11, where the crucial environment for the occurrence of *-nn-> -an- in these forms is given as follows:
"Hinter Konsonantengruppen und sonstiger langer Silbe: kl[assischl regelmaissig im ersteren Fall in den schwachen Kasus der Stiamme auf -van- -man- . v[edisch] oft auch hinter langem Vokal z.B. rtav(a)n-e, auch bei Suffix -an-: raj(a)n-d omdn-a ... sowie vereinzeltes an (auch bloss n geschrieben) in der v[edischen] Flexion derer auf -an: v.rsan-as vrs(a)n-as aryam(a).n-e mah(a)n-a ..." 2.10 Concerning the hypothesis advocated Debrunner and Wackernagel, the following jections must be raised. by ob-

P1. unterschied..." (1930: 24). However, as the discussion in section 2.4 above (which incidentally was based on relevant data assembled in Debrunner-Wackernagel 1930: 24-5, the same passage as the one referred to above) has shown, the extant historical evidence of (Vedic) Sanskrit indicates that the root accentuation is an archaism, rather than an innovation. The alleged earlier stage with end accentuation is, as Kurylowicz (1964: 201) correctly observed, "fictitious". 13 In nonablauting morphemes, syllabic resonants (i.e. "0-grades") do not infrequently occur accented; cf. PIE *wjkwos, *septrm. 14 For further discussion of the type pitfir, cf. section 3.7 below. 15 For additional arguments against the view that the usual desinential accentuation of Sanskrit (characterized) consonant stems must be an archaism, cf. section 2.4 above. 16 Only damdn- 'gift; giver' consistently has prevocalic -mn-; cf. damna (5x). The stems premdn- and loman-

2.10.1 The view that the usual desinential accentuation in the accusative plural of Sanskrit characterized consonant stems is inherited cannot be considered acceptable. For, as shown by the regular Rig-Vedic root accentuation in the accusative plural of nonablauting root nouns, the accusative plural must originally have had predesinential accentuation; cf. sections 2.4 and 2.5 above, as well as the arguments in Kurylowicz 1968: 27, for which cf. section 3.8 below.l2
l1 This view was accepted by Schwyzer (1939: 567-8). 12 Debrunner and Wackernagel were aware of the regular root accentuation of the nonablauting root nouns. However, they considered this accentuation innovated. In their opinion, the nonablauting roots gave up their original accentuation on the ending in favor of the root accentuation of the corresponding nominative plural forms

HOCK: Indo-Iranian

Accusative Plural

79

cannot be considered evidence for simple n-stem inflection. Wackernagel's aryam(a).ne (RV 4.3.5) may well be an unnecessary and unwarranted emendation for the aryamner actually found, a form which recurs twice in Rig-Veda under metrically unFor Arnold (1905: 300, ambiguous conditions. ad locum) plausibly suggested that the line in which it occurs, namely RV 4.3.5a, is a Gautami verse, which would accommodate the unamended form quite well, perhaps better than the emended form would fit into a Tristubh verse. In addition, the fact that there are two (or perhaps three) other instances of Gautami verse in this hymn would seem to furnish ample precedence for the assumption that also this line is in Gautami if the emendation to aryamane were meter.-Even to be accepted, however, the form cannot be considered prime evidence for the inflection of the simple n-stems. For whatever its historical origin -and neither of the two proposed origins involves a simple n-stem-,17 synchronically the stem aryaman- would no doubt have been interpreted as a man-stem. As a man-stem form, however, aryamane would be comparable to Ab/G mahimcinas (RV 2x), with sporadic generalization of the prevocalic 0-grade alternant -man- even to the environment after short vowel. And the relationship
between between aryamane mahimdnas (lx) and aryamne (2x), ary-

amnds (6x) would thus mirror the relationship


(2x) and unremade mahimnds

(2x), mahimne (lx), mahimna (lx), mahina (35x). Wackernagel's mah(a)na is attested twice in Rig-Veda (4.2.1 and 10.6.7) and might, as a genuine n-stem form, thus be considered to strongalso show only -mn- (>-n-) in the forms prena and lomno-

lomnas;but each of these forms is attested only in a single passage. The stems (-)naman- and bhumdn-have forms
both in -mn- and in -man-; cf. namnd (Ix), nam[a]nd (lx;

with metrically unambiguously required emendation of


-mnd to -mana), dhindmndm (Ix), suhdvitundm[a]ne (lx);

bhamdnd(lx), bhimna (2x). All other -man-stems with a root-final long vowel show only forms in -man-; cf.
omdnd (lx), tramane (lx), d'mane 'giving' (Ix), -manas (lx), dam[a]nas 'strick, rope' (lx), dhamand (2x), -mane (5x), -manas (2x), bhumand (lx), -manas (4x), yamanas (lx), Satdydm[a]nd (lx), dnusraydm[a]ne (lx), pravddydmand (lx), viomanas (lx), sam[a]nd 'song' (lx), -m[a]nas (3x), susam[a]ne (3x), sam[a]ne 'Erwerb' (2x), sydmand

ly argue in favor of Wackernagel's hypothesis. However, both these attestations are metrical emendations for actually attested mahna, a form which is attested elsewhere 32 times in metrically unambiguous environments. In the case of the attestation in RV 10.6.7(a), the emendation to mahana actually seems to be unnecessary and unwarranted; for as Arnold (1905: 318) observed, the line in which the form occurs is most likely to be a Pentad verse which would well accommodate the attested mahna. The occurrence of eleven other Pentad verses in this hymn, including line 7b, the line immediately following the line in which mahna occurs, furnishes ample precedence for the assumption that also line 7a is a Pentad verse. As for the occurrence in 4.2.1, it is true that metrical emendation to a trisyllabic form with the pattern uul - is inevitable. However, it is an a priori assumption that the result of that emendation must be an otherwise unattested mahana. Considering the semantic closeness of (the instrumentals of) the two stems nahan- and mahimdn- and considering the actual and frequent attestations of the instrumental mahina (cf. above) which would fit the metrical requirements perfectly, Arnold's emendation to mahina (1905: 300, ad locum) would seem highly preferable to the nonceform emendation mahana. Finally, as for Wackernagel's raj(a)nd (RV 10:97:22b; cadence of Epic Anustubh), written I accept Lanman's (1877:525) interrajfia,' pretation that metrical emendation is not necessary, since unemended rajnia would be acceptable in the cadence of a cateletic verse. There would be ample precedent for such a catelectic interpretation, for, as Arnold (1905:323) showed, there are three other catelectic verses in this hymn (14c, 15c, 19c), in addition to two further instances of (less typical) heptasyllabic verses (13a, 20a). 2.10.3 As for the accusative plural forms uksdnas (10.86.13) beside uksncs (ibid. 14,10.28. 11) and vftanas (4.2.2)19 beside vfsnas (8.7.33,8.20.20). the following two points are of importance. 18 Beside this form, there are eleven attestations of the expected, regular weak stem rdjii-. However, none of
these happens to be an instrumental singular. 19 According to Grassmann (1872: s.v.) there is also an accusative plural vfr'nas (8.7.33) which is to be read

(Ix), hemdnd(lx). That is, out of 49 (or 50) occurrences of relevant forms, 38 show prevocalic -man-. 17 On this matter, cf. section 2.10.3 below.

as vfsanas. However, an alternativeandmetricallypreferable emendation of this line in Arnold 1905: 311, ad locum, renders Grassmann'sreading unlikely.

80

Journal of the American Oriental Society 94.1 (1974) (2) What is equally important in the evaluation of the accusative plural forms of uksdn- and vtsan- is the fact that these two stems are aberrant not only in the formation of the accusative plural, but also in the formation of the accusative singular and nominative plural, categories about whose status as original full-grade formations there cannot be any doubt. The stems uksdn- and vfsan- belong to a small group of n-stems which form their accusative singular and nominative plural in a synchronically aberrant fashion, by having guna, rather than the usual vrddhi, in the stem-forming suffix. It is true, just as alternative forms with synchronically regular predesinential 0-grade can be found in the accusative plural of uksdn- and vtsan-, so also in the words belonging to the small aberrant class of n-stems under discussion (including uksdn- and vfsan-) there are occasional alternative accusative singular forms with the synchronically regular vrddhi of the stem-forming suffix. However, except for one stem, namely anarvdn- (which may well not belong into this group to begin with, cf. below), there are significantly no such alternative vrddhi forms occurring in the paradigmatically more pivotal nominative plural. Before discussing the significance of these facts, it will be good to have a look at the stems belonging to this class and at the relevant attestations of these stems. anarvdn-: A sg. anaruvdnam (10.92.14, Jagati cadence)
beside anarva.nam (1.37.1, opening of Gayatri; 1.51.12, opening of Jagati; 1.136.5, opening of Jagati verse of

(1) None of the other synchronic weak-grade forms of these two nouns shows -an- in prevocalic environment. The stem vfsan- shows the following forms: G/Ab. vfsnas (40x), I vfsnd (2x), D vfsne (34x). Similarly, the stem uksan- shows the genitive singular uksnds, not uksdnas*. This distribution is of paramount significance. For if the accusative plural of these forms really had originally been a weak case and if the -anforms of the accusative plural owed their existence to (generalization of) the effects of Sievers-Edgerton's Law after heavy syllable, one should surely expect to find similar traces of the effect of SieversEdgerton's Law in (some of) the forms of the other weak-grade cases. The absence of such forms and the fact that at least in vfsan-, with its light syllable preceding the stem-forming suffix, there would not have been any motivation for the application of Sievers-Edgerton's Law, would seem to indicate that the accusative plural originally was different from the (other) weak-grade cases. And this difference presumably and most likely consisted in the fact that the accusative plural originally was a full-grade case, especially considering the evidence adduced in sections 2.4 and 2.5 above in favor of the view that the accusative plural originally had predesinential accentuation. The forms uksnds and vfsnas would then have to be considered remakings of the more archaic uksdnas and uvsanas, so as to conform to the productive synchronic pattern according to which the accusative plural is a weak-grade formation, with accent on the ending (except for the r-stems, which accent the stem-forming suffix). It might be objected that the sole occurrence of uksdnas is in a rather late hymn (10.86.13) and that the form therefore can hardly qualify as an archaism. However, also the alternative form, uksnds, occurs only in late hymns (10.28.11 and 10.86.14), one of which is identical with that in which uksdnas is attested. That is, we can at least reach the negative conclusion that uksnds does not have any greater claim to being archaic than uksdnas. In the case of vfsan-, however, the attestation of the full-grade form vfsanas occurs in a quite early hymn (4.2.2). And considering the synchronic aberrancy of this full-grade, predesinential-accent accusative plural form of a characterized consonant stem, as well as the arguments in favor of original root accentuation in the accusative plural of the nonablauting root nouns, this early attestation of vfsanas can hardly be anything but an archaism.

Atyasti; 1.190.1, opening of Tristubh; 2.6.5, cadence of Gayatri; 6.48.15, opening of Jagati; 7.97.5, break of Tristubh and first syllable of cadence; 8.92.8, cadence of Gayatri)-N pi. anarvdaas (5.51.11, cadence of Jagati; 8.31.12, cadence of Gayatri)20beside anarvanias(1.190.6 opening of Tristubh; 8.18.2, opening of Gayatri in Us0ih)-G pl. anarvdinm (10.65.3).-Arnold proposed to emend three occurrencesof the accusative anarvanamto anarvdnam,namely those in the Gayatri cadences (2.6.5 and 8.92.8) and the occurrencein the break of a Tristubh
(7.97.5); cf. Arnold 1905: 297, 310, 315, ad loca. The acproposed emendations would of course make these cusative plural forms fit better into the regular metrical Howpatterns expected in the respective environments. the for be to seem necessary, not do strictly ever, they
20

These two attestations of the nominative plural are not listed in Grassmann 1872. They are listed in Debrunner-Wackernagel 1930: 267.

HOCK: Indo-Iranian unemended forms would fit metrical patterns which, though rarer and irregular (cf. Arnold 1905: 160 and 165), do occur elsewhere and which are characteristically, although not exclusively, found in early hymns. And, as Arnold himself pointed out (p. 165), one of these hymns, namely RV 2.6, is a hymn of the archaic period. Leaving aside these three forms, we find that of the remaining ten occurences of accusative singular and nominative plural forms, five attestations fit the regular metrical patterns expected in the environments in which they occur; that is, they may possibly have been secondarily adapted to fit the (regular) meter. These occurrences are: anarvdnam (10.92.14) and anarvcdnas (5.51.11 and 8.31.12) in Jagati and Gayatri cadences, anarvanam (1.37.1) and anarvanas (8.18.2) in Gayatri openings. Note that this set of forms which may possibly consist of metrical adaptations includes all of the three relevant guna forms of the accusative singular and nominative plural. The remaining five attestations, namely anarvanam (1.51.12, 1.136.5, 1.190.1, 6.48.15) and anarvdnas (1.190.6) occur in Tristubh or Jagati openings. What is significant about that environment, however, is the fact that the third syllable of the opening is metrically 'free', both -and u being possible. The fact that this free environment shows only predesinential vrddhi forms would seem to make it quite certain that the stem anarvcn- really regularly has predesinential vrddhi and that the guna found in anarvadnam and in the two occurrences of anarvdnas owes its existence to metrical considerations. (As for the genitive plural anarvdnam, it merely shows the usual Sievers-Edgerton's Law alternant -van- after consonant.) aryamdn-: A sg. aryamdnam (15x), D sg. aryamne (2x), aryamane (lx), G sg. aryamnds (6x), N/A/V du. aryamdna (lx), N pi. aryamdnas (lx). No forms with predesinential vrddhi occur in Rig-Veda; some occurrences of guna forms are found in metrically free environments, such as aryamdnam (4.2.4, opening of Tristubh); for the dative form aryamane, cf. section 2.10.2 above. In addition, notice that even in post-Rig-Vedic times, relevant guna forms occur, such as aryamdnam (AV 19.1.17); cf. Lanman 1877: 532. There can thus be no doubt about the fact that aryamdn- regularly has guna in the stem-forming suffix of the strong cases. Since de Saussure (1879: 220, fn. 1), this has generally been explained as due to the fact that aryamdn- is a compound of arya- plus the root man-, and that the inflection of the compound follows the (guna) root noun inflection exemplified by -han-, for which see below in this section. Compare DebrunnerWackernagel 1930: 267 and Kurylowicz 1956: 56 for reiterations of this view. A different interpretation, however, was offered by Thieme (1938: 135-41) who convincingly demonstrated that aryamdn- can just as well be considered an original man-stem. This view was accepted as prefer-

Accusative

Plural

81

able to de Saussure's view in Debrunner 1954: 4. In addition, as pointed out earlier (section 2.10.2 above), synchronically the stem aryamdn- was no doubt interpreted as a man-stem, whatever its origin may have been. After all, it is hard to believe that, even if the word were an original compound, speakers of the language would still have been aware of its original compound character. There is thus no reason against believing that, at least synchronically, aryamdn is a bona-fide member of the set of n-stems with aberrant (guna) inflection in the accusative singular and nominative plural. uksdn-: A sg. uksanam (1.164.43) beside uksdnam (4x), G sg. uksnds, N pl. uksdnas (llx), A pl. uksdnas (lx) beside uksnds (2x). The accusative plural forms, as well as the (genuine) prevocalic 0-grade form of the genitive singular, have already been discussed earlier in this section. Since the full- and 0-grade forms of the accusative plural are metrically incommensurate, because of the difference in the number of syllables, it is impossible to determine whether one or the other of these forms may owe its existence to metrical considerations. However, in the case of the vrddhi and guna forms in the synchronically strong cases, it is possible to show that the vrddhi form of the accusative uksanam (1.164.43) may well be metrically motivated, while the guna forms uksdinam and uksd.nas are genuine. For not only does uksanam occur in a late, 'popular' hymn (cf. Arnold 1905: 273), more importantly, it occurs in the opening of a Tristubh verse and its predesinential vrddhi conforms perfectly to the length expected in the second syllable of that metrical environment.21 On the other hand, there is at least one attestation of a guna form (ddhoksdno, 8.1.33) with the predesinential guna occurring in a metrically free position, namely in the third syllable of a Tristubh opening. The stem uksdn- therefore can be included in the set of aberrant n-stems with regular guna inflection in the accusative singular and nominative plural. Notice that, as in the case of aryamdn-, this exceptionality continues into postRig-Vedic times, as can be seen by uksdnam (AV 3.11.18) beside uksanam (VS 28.32, Ait. Br. 1.15), N pl. uksdnas (AV 4.24.4) beside uksanas (Katy. gr. Sitr. 23.4.7, 8, 11), and even A pl. uksd.nas (Bhag. Piir. 10.83.13); cf. Lanman 1877: 532, 538, 540. -ksan- (?) in rbhu-ksan-: A sg. rbhu-klsdnam (4x), N/V pl. -ksdnas (12x). Considering that the nominative (and vocative ) singular of this noun appears as rbhuksas (17x) and that, as the discussion in Debrunner-Wackernagel 1930: 309 shows, the n-inflection outside the nominative/vocative singular is suspect of being modeled on the inflection of the rhyming uksdn-, it is doubtful whether the word should be included in the set of aberrant n-stems. However, if the n-inflection of rbhuksdnis indeed modeled on that of uksdn-, then the pervasive

82

Journal

of the American

Oriental

Society

94.1 (1974)

(v'san vrsadnamn).In both environments, a long syllable is definitely preferred.21 On the other hand, guna forms can be found in environments where a short syllable, though not imipossible, is definitely less regular than a long syllable, such as vrsanas (1.177.2), with the guna stem-forming suffix in the fourth syllable of a (delayed) Tristubh opening (ye te vfrsano). The stem Vl'f'Sta-therefore can be included in the set of aberrant n-stems. Again, like the other bona-fide memibers of this set, it shows guna forns in post-Rig-Vedic timies. Compare N/A/V it would seem du. v.sanda (AV 7.73.1. 2; 7.110.2) beside v.rsunsla (AV Wortern .. ." Under these circumstances, dubious to include jeman- in the set of aberrant n-stems. 19.13.1), and even an accusative plural form, v'sanas pusdn-: A sg. puisdnam (21x), I sg. piusrni (Ix), D sg. (AV 11.2.22); cf. Lanman 1877: 524 and 537. The only post-Rig-Vedic accusative singular and nominative plural puisne (4x), G sg. pa.snds (5x), N/A/V du. pusadnc (lx), forms listed in Lanman 1877: 532 and 538, however, all somapuscand(u) (3x), indrdpusdnd (lx), Npl. pisasdias (lx; or is this form nominative singular of pCusdna- ?). No have predesinential vrddhi, cf. v.r.sdnam(VS 20.40; Sat. forms with predesinential vrddhi are attested. Forms Br. 1.2.5(15)) and vrsdanas(Sat. Br. 13.3.3(7)). -lian- in vrtra-hdn- and 27 other compounds: A sg. like G sg. pusnds show that the word is a (regular) ablauting n-stem. And attestations like pu.sdnam in RV -hdnam (17x), I sg. -ghnil (2x), D sg. -ghln (6x), G sg. 1.186.10, with the guna stem-forming suffix in the metri- -ghnds (3x), N/A/V du. -hdnd(L) (8x), N pl. -lhnas (3x). No forms with predesinential vrddhi are attested; and cally free third syllable of a Tristubh opening, show that there are occurrences of guna forms with the stemthe guna forms are genuine. The stem pusdn- therefore is a bona-fide member of the set of aberrant n-stems. forming suffix occurring in metrical environments which And as in the case of the other bona-fide members of are either free or where a long a would actually be preferthis set, so also with puisdn- one finds post-Rig-Vedic at- red, such as vrtralzcinam(1.59.6), early break of a Trislubll verse (plus first syllable of cadence). Though not in testations of guna forms, namely puisd.nam (AV 11.6.3, 18.2.53) and N/A/V du. indrdpusadna (AV 6.3.1); cf. origin an n-stem, but rather a root noun, -han- has been entirely integrated into the n-stem inflection (of the aberLanman 1877: 532 and 537. nrant variety), as can be seen by the fact that the nominative of this forms No other N (8x). pl. yosd.nas yosdn-: singular ends in-ha (94x); cf. the similar conclusions reachstem are attested; but compare the related noun stems ed in Debrunner-Wackernagel 1930: 238. The fact that the that originality ydos-, y6sand-, yosddn-. Considering the inflection of -han- only shows guina forms in the acof the n-stem nominative plural still is a matter of dispute confirms the (cf. Mayrhofer 1964: 27), it must at present remain an cusative singular and nominative plural in (pre-)Rign-inflection of an aberrant aberexistence the in be included can guna whether yosdnopen question Vedic. And as of Rig-Vedic times, -harn- must definitely rant ni-stems. v.san-: A sg. vrsacamn(9.34.3, 10.89.9) beside v.rsanam be considered a member of the set of aberrant n-stems discussion. Interestingly, as in the case of the (55x), I sg. vfsnad (2x), D sg. v.rn.e (34x), G/Ab. vrsnas under imembers of this set, so also in the case bona-fide other (38x), vfrsanas du. N/V pl. (57x), vfsana(u) N/A/V (40x), of accusative The -han-, post-Rig-Vedic gulia forms are attested; cf. A A pl. vrsa.nas (lx) beside vrsnas (2x). raksohdcnan, marnohdnzam, sapatnahdnlam, each attested O-grade as the sg. as well prevocalic (genuine) plural forms, once in Atharva-Veda; N/A/V du. vrtrahdCnd(AV 2x); forms of the singular oblique, have already been discussed N/V pl. matsyahdnas (Sat. Br. 13.4.3(12)), yajrahdcnas earlier in this section. As in the case of uksdn-, the full(TS 3.5.4.); cf. Lanman 1877: 480-1. Even more interand 0-grade forms of the accusative plural are metrically esting and important for the present discussion is the incommensurate, making it impossible to determine fact that there are post-Rig-Vedic guna attestations of its owe forms of these may other whether one or the existence to metrical considerations. However, just as the accusative plural; cf. raksohdno valagahdno vaisnlavuna (TS 1.3.2; 5x) beside saiaghnds (AV 3.15.5); cf. Lanman in the case of uksdn-, it is possible to show that the vrddhi 1877: 481. (According to Debrunner-Wackernagel 1930: forms of the accusative singular may well be metrically are 238, accusative plural forms in -hdnas are also found in motivated, while the guna forms of the strong cases the and 10.89.9) Vajasaneyi Samhita.) Notice, however, that there is of vfranam (9.34.3 genuine. The attestations have the vrddhi stem-forming suffix occurring in the 21 Also Lanman (1877: 532) thought that the d of vrsabhir) second syllable of a Gayatri opening (vursalaQn uksadoam and vrsa.nam "may be ascribed to the metre." and in the fourth syllable of a (delayed) Tristubh opening

guna of the attested forms of rbhuksdn-, some of which occur in metrically free environment (cf. rbhuksdcinam 1.111.4, opening of Jagati), would seem to lend further support to the view that in Rig-Veda, the strong cases of uksan- regularly have guna of the stem-forming suffix. jdman- (?): du. jdmana (?) (10.106.6). The hapax occurs in a stanza of extremely difficult interpretation which Geldner (1923) leaves untranslated and on which he cormments "Die Kunst des Dichters besteht ... in eineii Aufwand von unverstandlichen und seltsam klingenden

HOCK: Indo-Iranian also one post-Rig-Vedic occurrence of a guiia form in a genuine weak-grade case, namely G sg. brahmahanas (Manu 11.101.128), where Lanman suspects that the anomaly may be "due to the metre" (ibid.: 478); for -ghnas occurs in Manu 8.89. Elsewhere in post-RigVedic, the genuine weak-grade forms only show the 0grade of the stem-forming suffix; ce. I sg. vrtraghna (AV 3.6.2), D sg. makhaghne (TS 3.2.4(1-3)), bhrtinaghnle(TBr. 3.9.15(3)), G sg. vrtraghinds (AV 4.24.1; 6.82.1); cf. Lanman 1877: 480. Considering the late date of the colmposition of the Laws of M'lanu,the possibility of metrical adaptation, as well as the overwhelmling evidence of Rig-Vedic and post-Rig-Vedic literature in favor of a genitive singular -ghnds, it seems best to consider Manu's brahmahanas a late aberration, of no significance for the discussion of the Vedic inflection of -han-. As for the accusative plural forms of the Taittiriya Samhita, however, the rather early occurrence in a (repeated) FORMULA would seem to make such an interpretation highly unlikely. It rather seems that we are here dealing with genuine formations, comparable to the (post-)RigVedic accusative plural forms uksd.nas and vr.salias. And just as the latter coexist with the presumably innovated 0-grade forms uksnds and vrsnas, so beside -ldnas one also finds -ghnds. Debrunner and Wackernagel, to be sure, suggest that the accusative plural forms in -hdnas are formed "nach dem Vorbild der vielen unveranderlichen Nomina agentis"(1930: 238), i.e., according to the model N pl. -dfsas: -hdnas = A pl. -dfsas: X. However, the fact that -hdn- is not "unveranderlich", but rather is ablauting, the fact that the model of the nonablauting root nouns, with their regular 0-grade of the root, should more likely have brought about a generalization of the 0-grade than of the full grade,22 and the undeniable synchronic membership of -hdn- in the (ablauting) nstems (cf. N sg. -ha) render the view of Debrunner and Wackernagel unlikely. We may then safely conclude that, from the synchronic point of view, -hdn- is a bona-fide member of the aberrant n-stems and that, like two other members of this set, it shows traces of a presumably more archaic accusative plural with predesinential guna (coexisting with innovated, synchronically more regular 0-grade accusative plural forms).

Accusative Plural

83

(uksadn-, vtsan-, and -hdn-) show the additional aberrancy of having accusative plural forms with guna of the stem-forming suffix, beside synchronically more regular forms with predesinential 0-grade. The other two members of the set unfortunately have no accusative plural forms attested. However, even in the absence of that evidence, it would seem reasonable to suppose that the two mentioned aberrancies are not just linked up by accident, but rather that they are intimately connected and that both aberrancies are ultimately to be explained in the same fashion. In the case of the accusative plural formations of uksdn-, vfsan-, and (implicitly) -hdn-, it has been argued above that the synchronically aberrant guna accusative plural forms are more likely to be archaisms than the corresponding 0-grade forms, since the latter can be accounted for as innovations, conforming to the synchronically productive pattern of ablauting consonant stem inflection. Similarly, in addition to the metrical arguments offered so far, it is also possible to argue on purely linguistic grounds that the predesinential guna in the accusative singular and nominative plural of the aberrant n-stems must be an archaism, compared to the occasional vrddhi found in the accusative singular. For not only is it easier to account for the vrddhi forms as secondary, modeled on the inflection of the regular n-stems.23 What is equally significant is the fact that in the bonafide members of the aberrant n-stems, the nominative plural, the paradigmatically more pivotal form, never shows any forms with predesinential vrddhi. However, before we can be completely certain that the aberrant behavior of the n-stems under discussion is indeed an archaism, it is necessary to show that there is a MOTIVATION for these words to retain an archaic type of inflection. According to Kurylowicz (1956: 56 and especially 62), the difference in formation of the accusative
23 As a matter of fact, in some of the metrical environments in which the vrddhi accusative singular forms oc-

There are then five bona-fide members of the aberrant set of n-stems with guna in the accusative singular and nominative plural, namely aryamdn-, ukscsn-, pusdn-, vfsan-, and -hdn-. Three of these Such a secondary nominative plural with 0-grade of the root on the model of the 0-grade root nouns is actually attested once in Rig-Veda,namely divas (9.108.11) for regular dyavas.
22

flection may have been stronger than metrical considerations in motivating the predesinentialvrddhi. Compare
v.sdnam (10.89.9) and vrsanas (1.177.2) whose stemforming suffixes both occur in the fourth syllable of a (delayed) Tristubh opening where length is definitely

cur, the LINGUISTIC pressure of the regular n-stem in-

preferred,but where a short syllable evidently is possible also.

84

Journal of the American Oriental Society 94.1 (1974) vable on-stems like PIE *k(u)won-. Kurylowicz's suggestion that they may contain a "primitive" o, which I can only interpret to mean a nonapophonic o, cannot be considered acceptable. For as Skt. pdti- and dvi- from PIE *poti- and *owi(the latter from earlier *Oewi-) show, an original nonablauting o does not undergo Brugmann'sLaw lengthening in open syllable, while the *o of *k(u)won- does (cf. A sg. s(u)vdnam). A more likely explanation would seem to lie in the assumption that in Proto-Indo-Iranian times the inflection of the Proto-Indo-European on-stems began to become generalized. The motivation for this generalization presumably lay in the fact that in the original full-grade cases of the accusative singular and nominative plural (for the accusative plural, cf. section 4.2 below), PIE *o had become a by Brugmann's Law and that the resulting vrddhi of the stem-forming suffix was more marked than the guna of the original en-stems. The generalization of the more marked vrddhi of the original on-stems then led to the effective transfer of most original en-stems to the inflection of the original on-stems, since after the merger of nominative *-e(n) and *-o(n) into a the two classes of inflection had already merged in all the other forms of the paradigm.
with secondary -ud in the nominative singular, Gk. poimen is no doubt derived from *poAy-, a pseudo-root abstracted from the *peA-derivative *poA-yu- 'protector; protection, etc.' (cf. Skt. pdyu- (m.) 'protector,' Gk. poii (n.) 'herd') which was reinterpreted as *poAy-u-. (For the difference in meaning, accent, and gender between the Sanskrit and Greek reflexes of *poA-yu-, cf. the wellknown difference between Skt. brdhman- (n.) 'formulation, etc.' and brahmdn- (m.) 'formulator, etc.'.) Incidentally, the phonological development of *poAy-men- = [poAimen-] to *poymen- via *poimen-, with loss of intervocalic laryngeal, would to my mind most convincingly illustrate the impossibility of deriving the preconsonantal alternants of the so-called Schulzian bases, as in Skt. patra- 'drinking vessel', Lat. pdculum 'drink', PIE *po-tlo- (from the root *po(y)/pi-), from original forms of the type *peOy-tlo-. For poimenlpiemu6 clearly shows that such a sequence would yield Skt. petra-*, Lat. piiculum*. Compare also the similar interpretation of the *poy- in Gk. poimen advanced by Kurylowicz (1968: 218-20) who, however, seemed unaware of the fact that this obvious DERIVATION of Gk. poimen from secondary *poy- is in conflict with his view of 1956 that the ProtoIndo-European ancestors of the Sanskrit man-stems were (derived I) mon-stems, and not men-stems.

singular and nominative plural24 between the aberrant n-stems and the regular n-stems is due to the fact that the former are Proto-Indo-European en-stems, while the latter reflect Proto-Indo-EuroIn Proto-Indo-European, howpean on-stems. ever, this difference between en- and on-stems followed a definite pattern. While the en-stems were synchronically "unmotivated," i.e. nonderived, the on-stems were "motivated," i.e. derived. And, "si l'on fait abstraction des theses hasarddes (p. ex. ursen- < *uers- 'arroser'), aucun de ces mots [i.e. of the aberrant n-stems] ne se laisse deriver d'une forme-base connue par ailleurs." On the other hand, original on-stem forms like Skt. "tdksa, tdiksanam< tksati ..." are clearly derived.
"Dans certains cas la motivation nous 6chappe, soit qu'il s'agit d'un vocalisme o primitif soit que le motbase a disparu: sgv, svanam ([Gk.] ktin) . . . En revanche, tous les derives en -man-, -van- allongent la voyelle suffixale aux cas forts en garantissant un vocalisme o, qui est confirme par le grec (-mon, -won)."

Though Kurylowicz's view that the difference between predesinential guna and vrddhi reflects a Proto-Indo-European difference between en- and on-stem inflection is no doubt correct, his argument that in turn the difference between en- and on-stems was in Proto-Indo-European motivated by the fact that the on-stems were derived, while the en-stems were nonderived, must be considered dubious. For even if the usual, and in my opinion not all that "daring," etymologies of vrsan- and uksdn- as derived, via 'emit semen, inseminate', from *wers- '(let) rain' (cf. above, as well as Mayrhofer 1970: 251) and *ug-, *uks- 'sprinkle, moisten' (cf. Mayrhofer 1956: 98) may ultimately prove to be wrong, the fact that aryamin- most likely is a derived man-stem (and Kurylowicz claimed that the derived nature of the on-stems persisted into Sanskrit) not only is in conflict with Kurylowicz's claim that the aberrant nstems are nonderived, but also with the claim that the Sanskrit man- and van-stems guarantee ovocalism for the ancestral Proto-Indo-European suffixes; for the lack of vrddhi in the strong forms of aryamcin- clearly points to a Proto-Indo-European suffix *-men-. Moreover, even Greek shows at least one men-stem about whose derived nature there cannot be any doubt, namely poimgn 'shepherd'.25 Finally, note that there still remain underi24 In this book, Kurylowicz did not try to account for the aberrant accusative plural forms. 25 Like Lith. piemuo, G sg. piemeiis 'shepherd boy',

HOCK: Indo-Iranian What preserved the aberrant n-stems, which in light of their pervasive guna no doubt are original en-stems, from the transfer to the inflection of the original on-stems must then have been not their alleged nonderivability (since at least one of them clearly is derivable), but rather their meanings: In the case of the religious terms aryamdnand pusdn-, and in the case of -hdn- with its strong legal and religious connotations, it is the general conservatism of religious and legal terminology which can be considered responsible for the preservation of the original en-stem inflection in the accusative singular and nominative plural. As for uksdn- and vtsan-, there are of course also certain religious connotations in the Vedic hymns. However, I believe that the key factor in their preservation of en-stem inflection lies in the fact that in the cattle-raising society of the early Indo-Iranians, words referring to the male of the species, especially the bull, belonged to the central, basic part of the vocabulary which notoriously is most resistent to analogical remaking.
No doubt the same
SEMANTIC

Accusative Plural

85

adumbrated by Brugmann (1876: 290, cf. sections 2.3 and 3.1 of this paper) and Burrow (1955: 235 and 1965: 235, cf. section 3.6 below), that the accusative plural, being evidently derived from the full-grade (I) accusative singular by the addition of the plural marker -s, should be expected to have predesinential full grade too, is valuable and lends further support to the arguments of sections 2.4 and 2.10 in favor of the hypothesis that the accusative plural originally was a predesinentially accented, full-grade case. 3.1 The first linguist to have suggested that the Proto-Indo-European accusative plural originally was a strong case and that the usual Sanskrit predesinential 0-grade is secondary seems to have been Brugmann (1876). His positive arguments in favor of this view, which were already included in the evaluation of Osthoff's view of 1876 (cf. section 2.3 above), can be summarized as follows: In Vedic, the ablauting nouns frequently show the accusative plural in the strong form, both in terms of ablaut and in terms of accent; cf. dpas beside apcis, uksdnas beside uksnds. Also in the other Indo-European languages, the accusative plural is a strong case; cf. Gk. 6pas, Go. auhsans. That is, strong forms of the accusative plural are found in all Indo-European languages, while weak forms are limited to Indo-Iranian and are therefore suspect of being a regional innovation. The motivation for the Indo-Iranian remaking of the accusative plural must have consisted in the fact that by regular sound change, nominative and accusative plural had become synonymous. The remaking of the accusative plural thus led to a differentiation of nominative and accusative plural, although the precise circumstances which brought about this remaking are as yet unknown. (An additional argument in favor of the view that, like the accusative singular, the accusative plural must originally have been a strong case, was only adumbrated by Brugmann when he stated that the original accusative plural ending was evidently that of the accusative singular plus the plural marker -s; for the obvious corollary to the assumption that there was an original identity in the endings of the two cases would seem to be that also the preceding stems originally were identical.) 3.2 Brugmann's arguments certainly are generally quite cogent, although the claim that the accusative plural of ablauting consonant stems is a strong case in ALL the non-Indo-Iranian Indo-

motivation

ac-

counts for the aberrant accusative plural forms of uksdn-, vfsan, and -hdn-, although the FORMAL explanation of the difference between productive 0-grade and archaic full grade (guna) in the stemforming suffix of the accusative plural, for which compare section 4.2 below, will turn out to be quite different from that of the difference between vrddhi and guna in the accusative singular and nominative plural. 2.11 The most recent reiteration of the view that the accusative plural originally was a weak case seems to be found in Bailey 1969: 81-2, with reference to Osthoff 1876. Although accepting the view, Bailey was clearly, and I believe justly, troubled by the fact that "it is not clear why the original uter [= masculine/feminine] accusative plural formation had a zero stem vowel, while the corresponding singular had a normal [ = full] grade stem vowel", considering that the accusative plural ending evidently is that of the accusative singular plus plural marker -s. 2.12 The view that the accusative plural originally was a weak case must be considered quite dubious, considering the evidence in favor of original predesinential accentuation, which might very well entail a predesinential full grade, as well as the direct evidence for a predesinential full grade in the otherwise archaic, aberrant nstems. However, Bailey's observation, already

86

Journal of the American Oriental Society 94.1 (1974) replaced by *-ns > -as which, however, retained the end accentuation [and presumably the predesinential 0-grade] of the ending *-ens. Considering the absence of any unambiguous evidence for an ending *-ens in any of the (other) Indo-European languages,26 Bartholomae's hypothesis must be considered to be without any foundation and therefore to be rejected. 3.5 Also Thumb (1905: 150) expressed doubt that the predesinential 0-grade of the Sanskrit accusative nlural is original, since "das Griechische und andere Sprachen bilden den Ace. Pl. von der Hochstufe." This view was repeated in subsequent editions of his book; cf. for instance ThumbHauschild 1959: 5-6. However, no motivation or model for the secondary origin of the predesinential 0-grade is offered. In addition, note that, as already pointed out in section 3.2 above, outside of Greek no (non-Indo-Iranian) language shows convincing evidence for the assumption of an original predesinential full grade in the accusative plural of consonant stems. 3.6 A more substantial argument in favor of considering the accusative plural an original strong case was presented by Burrow (1955: 235 and 1965: 235). According to Burrow,
"The acc. pl. is a weak case in Sanskrit, that is to say the termination is accented and the stem appears in its weak form. This is in contradiction with the fact that the termination itself appears in the weak grade and it is therefore in all probability an innovation. If IE -ns in this case is derived from -ms the form can have arisen by the addition of the plural sign -s to the ace. sg."

European languages must be considered exaggerated. Only in the Greek inflection of the kinship r-stems do we find convincing evidence for the claim that the accusative plural of the consonant stems is a strong case (as opposed to the weak cases of the genitive and dative singular and plural). Elsewhere, levelings of various kinds have led to the identity of the accusative plural stem with that of (most of) the other forms of the paradigm, as in Gk. poimenas = poimenes, poimenon, etc. As a consequence, the argument that the IndoIranian predesinential 0-grade, being limited to that dialect area, must be a regional innovation, loses its force. Other evidence in favor of considering the Vedic strong forms of the accusative plural to be archaisms must be found (along the lines of the discussion in sections 2.4 and 2.10 above). Finally, Brugmann did not succeed in finding a plausible model for the presumably secondary 0-grade forms of the accusative plural, a fact which considerably weakens his hypothesis. 3.3 Hillebrandt (1878: 314-6) essentially agreed with Brugmann's view of 1876. However, he believed that it is the original "length" of the accusative plural ending -ans (vs. the "short" ending -as of the nominative plural) which was responsible for a shift of the accent from the predesinential
element to the ending in HISTORICAL
TIMES,

change which was not yet completed in Vedic, as can be seen by the coexistence of forms like A pl. dpas and apas. Considering that in Vedic, where this change allegedly was still taking place, the ending of the accusative plural of consonant stems -as is not any "longer" than the nominative plural ending -as, Hillebrandt's view must be rejected even in the framework of Proto-Indo-European reconstructions which he employed. (Needless to say, today's reconstruction of the accusative plural ending as *-ns, which is just as short as the nominative plural ending -es, renders Hillebrandt's views entirely untenable). 3.4 Following a tentative suggestion of Brugmann (1876: 292, fn. 7 [on p. 293]), Bartholomae (1894: 131) attempted to motivate the discrepancy between the Indo-Iranian predesinential 0-grade and the (Sanskrit) accent on the case ending by postulating two Proto-Indo-European accusative plural endings, namely *-ens [presumably with predesinential 0-grade] and *-ns [presumably with full grade]. In Indo-Iranian, the ending *-ens was

3.7 Like almost all the other linguists who argued in favor of the view that the Indo-Iranian predesinential O-grade is secondary, Burrow unfortunately failed to provide for a motivation and model accounting for the secondary nature of the O-grade. In addition, one might raise the following objection to his view that the double 0-grade (of predesinential element and ending) found in the
26 The Latin accusative plural ending -es is of course ambiguous, derivable just as well from the *-ns indicated by the other Indo-European languages as it is derivable from the alleged *-ens. Being ambiguous, it cannot in good conscience be considered probative evidence for the reconstruction of an otherwise unattested *-r7s.

HOCK: Indo-Iranian regular Indo-Iranian accusative plural must be secondary: As is shown by the genitive/ablative singular of the r-stems (cf. Skt. pitur < *-r-s), such double 0-grades, though rare, do occur elsewhere in nominal inflection. It is true, Burrow considered also the latter double 0-grade to be secondary. He believed that *-r-s most likely was a replacement of earlier *-er-s which in his view is found also in Lithuanian G sg. -ers, as well as in
Av. nars, zaotars, sastarg. According to Burrow,

Accusative Plural

87

this replacement was brought about by the generalization of the 0-grade of the stem-forming suffix (*-r/r-) from the other oblique forms of the paradigm. (1955: 243, 1965: 243). This explanation of the double 0-grade of the genitive/ablative singular, however, must be considered doubtful. For the Lithuanian genitive singular ending -ers reflects an earlier -eres, attested in Old Lithuanian (cf. Senn 1966: 140-1 for references). As for the Avestan forms, it must be kept in mind that an Avestan ar(9) is not always and necessarily a good indication of an original full-grade er (or or), since, as is well known, ar(9) (rather than "regular" ard) appears also in (other) 0-grade formations, whether this be due to phonological merger or some other process; cf. most notably the -ta-participles of the type darasta-: Skt. drstd- beside barata-: bhrtd-, as well as the fossilized and therefore even more valuable asa- : rtd-. In this respect note especially that beside the narg cited by Burrow, there is also a Gathic nardS. Thus, the best thing that can be said about the Avestan forms is that they are ambiguous, permitting both a reconstruction *-r-s and a reconstruction *-er-s, the latter however being without any (direct) support from any other Indo-European language. On the other hand, note that Old Norse ofQur and OE fador seem to furnish quite convincing non-Indo-Iranian evidence for an original ending *-r-s. It is true, on the evidence of n-stem genitive singular forms like Av. xvTng, ddng (< *suHen-s, *dem-s) and Old Irish anm(a)e 'of a name' (< *n-men-s), it might still be considered possible to reconstruct also original rstem genitive singular forms in *-er-s. In that case, Burrow's proposed Indo-Iranian replacement of this *-er-s by a form with the 0-grade of the rest of the oblique would probably have to be assumed to have taken place already in (dialectal) Proto-Indo-European, with subsequent remaking of the anomalous *-r-s to the more usual *-r-e/os in the majority of Indo-European dialects. However, that reconstruction would seem far too specu-

lative to be made the basis of an argument in favor of considering the double 0-grade of the Indo-Iranian accusative plurals unprecedented and therefore no doubt innovated. More convincing is Burrow's adumbration of the view that as a pluralized accusative singular, the Proto-Indo-European accusative plural should be expected to have the same predesinential ablaut grade as the accusative singular; and there can be no doubt that the Proto-Indo-European accusative singular was a strong case. This view had already been adumbrated by Brugmann (1876: 290); cf. section 2.3 above. It was clearly stated in Bailey 1969: 81-2, although Bailey nevertheless accepted the view that the accusative plural originally was a weak case; cf. section 2.11 above. 3.8 According to Kurylowicz (1964: 200-1 and 1968: 26-7), the predesinential 0-grade and the usual concomitant end accentuation of the IndoIranian consonant stem accusative plural is an innovation, motivated by the attempt to differentiate the accusative plural from the nominative plural with which it had become identical through the merger of *-ns and *-es to -as. Relics of the older accusative plural with predesinential full grade are found in forms like RV A pl. uksd.nas. In Kurylowicz's view, the innovated predesinential 0-grade of the Indo-Iranian r-stems came about on the following model of "closed" i- and u-stem inflection.
i-stems: N pl. -ay-as : A pl. i-stems: -av-as: r-stems: *-ar-as27: -i-ns -u-ns X = -r-ns.

The accusative plural forms of the remaining consonant stems came about on the model of an alternative pattern of "open" i- and u-stem inflection:
i-stens: N pl. -ay-as : A pl. u-stems: -av-as : n-sterns (etc.): *-an-as27: -y-as -v-as X = -n-as etc.

The concomitant end accentuation of the accusative plural then first spread to the ablauting root nouns "bei denen er [i.e. the end accent], wie in den eben angefiihrten Stimmnen,einen Wurzelablaut begleitete . . .
27

I have put the asterisks into these formulae to

indicate that in order for the model to have the proper phonological motivation (-dy-as : -dv-as : -dr-as, not -dras*), it must be assumed that Brugmann's Law had not yet brought about predesinential vrddhi in the (majority of) r- and n-stems.

88

Journal of the American Oriental Society 94.1 (1974) Finally, and most importantly, there is evidence that the "open" i- and u-stem inflection originally had predesinential 0-grade not only in the accusative plural (and the singular oblique), but also in the nominative plural; cf. Skt. RV N pl. dryas (16x), mddh(u)vas (4x), as well as Gk. Hom. It is oles (< *ow-y-es) and huies (* <suy-w-es). true, there are (i- and) u-stems of the "open" type which have attestations of nominative plural forms in (-ayas,) -avas; cf. N pl. krdtavas (5x) vs. D sg. krdtve (llx), G/Ab. sg. krdtvas (2x). However, also other cases of the paradigm show forms which formally agree with corresponding forms of the "closed" type of inflection; cf. D sg. krdtave (lx), ConG/Ab. sg. krdtos (lx), A pl. krdtIn (lx). sidering that the forms in -ave, -os, -avas, -un follow the synchronically regular pattern of ("closed") i- and u-stem inflection and that the "open" iand u-stem inflection is moribund, it is quite likely that all of the "closed" forms of krdtu- (and of other "open" i- and u-stems) are innovations, with generalization of the synchronically regular predesinential guna (except in the accusative plural, where it is merely the shape of the ending ending which is innovated). It is possible that Kurylowicz felt that as far as the last-mentioned point is concerned, he was on safe ground. For the predesinential 0-grade of dryas, mddh(u)vas, oies, hules has frequently been considered secondary; cf. for instance Brugmann 1911: 215, 216, 222, as well as the attempts in Debrunner-Wackernagel 1930: 159 (with earlier literature) to explain away (most of) the RigVedic nominative plural forms in question as genitive singular masculine or neuter. However, there are also many linguists who consider the nominative plural forms in question archaic and original. The reason for this belief, if any is given, lies in the close direct and indirect agreement between (Rig-Vedic) Sanskrit and (Homeric) Greek, especially between the "open" inflection of RV G/Ab. sg. dvyas (13x) and the pervasive "open" inflection of Gk. ois, N pl. oies, etc.; cf. for instance Kretschmer 1892: 327-8, Schwyzer 1939: 573-4, Burrow 1955: 244-5 and 1965 244-5, Szemerenyi 1956: 172-4 and 1970: 163. Of these two views, the second one would seem to be eminently preferable, not so much because of the agreement between Sanskrit and Greek (for that agreement could conceivably be due to parallel independent innovations), but rather for the following reasons.

In Wurzelstimmen ohne Ablaut ist dagegen die Barytonese des Akk. P1. noch ziemlich gut bezeugt, vgl. RV
VI, 4, 5 Akk. P1. hrrftah 'Feinde' ..., AV schon hrutdl . .. hat." (1968: 27). wogegen der

3.9 Kurylowicz's argumentation represents a definite improvement over earlier views that the predesinential 0-grade is an innovation. For he furnished both a motivation and a model for the origination of the secondary 0-grade. In addition, the observation that the spread of end accentuation to the root nouns was tied up with whether the root undergoes paradigmatic ablaut or not will turn out to be of considerable interest; cf. 4.2 below. However, even Kurylowicz's hypothesis is open to a number of objections. First of all, it would have been desirable to indicate, as I have done in section 2.10 above, on what grounds forms like uksdnas can and must be considered archaisms, rather than aberrant formations with secondary generalization of the Sievers-Edgerton's-Law alternant -an- of the stemforming suffix in its prevocalic 0-grade shape. Secondly, it is difficult to see why one type of consonant stems, namely the r-stems, should have been remade on the model of the "closed" i- and u-stem inflection, while all the other consonant stems were remade on the model of the "open" inflection. One would expect at least some fluctuation between the two types of inflection, in at least some consonant stems. Witness especially the fluctuation actually found in the u-stem accusative plural pacsn (RV 5x) beside pasvds (5x). However, no such fluctuation is found in any of the consonant stems.28 Thirdly, considering that the "open" type of i- and u-stem inflection is clearly moribund in Indo-Iranian, it is difficult to see how it could have been the model for the remaking of the MAJORITY (I) of the consonant stems.
28 It is true, this particular objection could be met by claiming that the remaking of the consonant stems took place only on the model of the "open" i- and u-stem inflection and that the accusative plural *-r(n)? owes its existence to a later remaking on the model of the productive (!) "closed" i- and u-stem inflection along the lines suggested by Bartholomae (1894: 123), cf. section 2.6 above. However, I believe the next two points of the discussion in this section will show that this modification of Kurylowicz's hypothesis would not be sufficient to salvage it.

HOCK:

Indo-Iranian

Accusative

Plural

89

of the nominative (1) A remarking plural on the model of the accusative plural is a priori unlikely, since the former is the "forme de fondation" in relation to the latter. And, as Kurylowicz (1945: 164) correctly pointed out, in morphological remaking it is the "forme fond6e" which is remade on the model of the "forme de fondation", not vice versa. (2) Considering that in the consonant stems, Indo-Iranian secondarily differentiated (rather than and accusative identified) nominative plural, it is quite unlikely that the two cases should have been identified in the "open" i- and usecondarily stems, at least in those of Indo-Iranian. (3) As in the case of the members of the aberrant n-stems, so also in the case of the nouns with bona-fide "open" i- and u-stem inflection, it is possible to show that their meaning makes it likely that (all of) the aberrant inflection of these nouns is an archaism. In Rig-Veda, the following forms can be found:29 dri-: D sg. draye (lx), G/Ab. sg. dr(i)yas (39x), N pl. dryas (16x), A pl. dryas (8x). dvi-: G/Ab. sg. dvyas (13x). pdti-: D sg. pdtye (8x), G/Ab. sg. pates (lx) beside pdtyur (lx), N pl. pdtayas (16x). In light of the wellknown influence of the kinship r-stems on the inflection of the stem pdti- in the kinship meaning 'husband' (cf. the genitive/ablative singular pdtyur) and considering that all eight occurrences of the dative singular pctye are instances of pdti- 'husband', the possibility that the predesinential 0-grade of this form is modeled on the predesinential 0-grade of the kinship r-stems cannot be excluded. The stem pdti- therefore cannot be considered a bona-fide member of the i-stems with "open" inflection rayi-: D sg. raye (62x), G/Ab. sg. rdyds (105x) beside rayas (3x) whose accentuation is probably modeled on the (secondary) "root" accentuation of the nominative plural, N pl. rayas (20x), evidently with secondary "root" accentuation on the model of the (ablauting) root nouns, A pl. rayas (4x) beside rdyds (22x) where again the root accentuation is probably secondary. For a discussion of the derivation of this paradigm from a stem *reE-iwith "open" inflection, with predesinential 0-grade Feminines, as well as the rare occurrences of root nouns without the usual t-extension (exemplified in stems like ji-t-), are excluded from this list; the former because of the well-known mutual influence between i- and i-, u- and a-stems, the latter because of their tendency to follow the usual nonablauting pattern of 0-grade root nouns. 29

throughout the paradigm, cf. Szemerenyi 1956: 170-5, with earlier literature. However, note that on purely phonological grounds, the dative singular and nominative plural of this stem could also be reconstructed with predesinential full grade, since both *reE-ey-ey, *reE-ey-es and *reE-y-ey, *reE-y-es will equally yield Sanskrit rdye and rayas. sdkhi-: D sg. sdkhye (9x), G/Ab. sg. sdkh(i)yur, etc. As in the case of pdti-, the inflection of this word has been influenced by the kinship r-stems. The prevocalic 0-grade of the dative singular therefore cannot be considered probative evidence for an original "open" inflection (in the singular). *krtu-: Attested only in the fossilized accusative plural krtvas 'times; Germ. mal' (2x). krdtu-: D sg. krdtve (llx) beside krdtave (lx), G/Ab. sg. krdtvas (2x) beside krdtos (lx), N pl. kratdvas (5x), A pl. krdtun (lx). nau-: G/Ab. sg. ndvds (lx), N pl. navas, with secondary "root" accentuation comparable to that found in N pl. rayas. For a discussion of the derivation of this paradigm from a stem *neA-u- with predesinential 0-grade throughout the paradigm, cf. Szemerenyi 1956: 185-6, with earlier literature. Notice however that the nominative plural is phonologically ambiguous and can be derived both from *neA-ew-es and *neA-w-es. pasd-: D sg. pasve (3x) beside pasdve (lx), G/Ab. sg. pasvds (10x), N pl. paddvas (lx), A pl. pasvds (5x) beside pasun (5x). pitu-: G/Ab. sg. pitvas (9x). mddhu-: D sg. mddhune (lx), G/Ab. sg. mddh(u)vas (83x) beside mddhunas (lOx) and mddhos (14x), N pl. mddh(u)vas (4x). madhydyd-: N pl. madhydydvas (1.173.10). Considering that -yu-derivatives regularly have "closed" u-inflection and that the word is a hapax legomenon, occurring in the generally late first book of the Rig-Veda, in a group of hymns which Geldner (1923: vol. 1, p. 237) characterized as containing "Viele seltene Worter oder hip[axJ leg[omena]", it is more likely that the form under discussion is a hyperarchaism than a precious relic of original "open" inflection. mitrdyd-: N pl. mitrdyivas (1.173.10). Like the preceding form, with which it cooccurs in the same stanza, also this hapax legomenon is most likely a hyperarchaism. sisu-: D sg. sisve (lx) beside sisave (2x), G/Ab. sg. sisvas (lx) beside gisos (lx), N pl. SiSavas (lx). sramayzi-: N pl. sramayuvas (1.72.2). Like the other two -yu-forms in this list, this form is a hapax legomenon and occurs in the first book of Rig-Veda. The group of hymns in which it occurs was described by Geldner (1923: vol. 1, p. 86) as being characterized by "die Haufung der oft gesuchten und abstrakten Vergleiche, eine uber-

90

Journal of the American Oriental Society 94.1 (1974)


likely that the Indo-Iranian "closed" inflection of the words in question is innovated. This is essentially the view of Burrow (1955 and 1965: 244, 247). Ultimately, however, the extraparadigmatic evidence for an original oblique stem in *-e/ow- assembled in Anttila 1969: 117, 122, 151, 159 would seem to favor the view that the IndoIranian type of inflection is more archaic. Thisis the view of Anttila (ibid.: 159). Further evidence for the archaic nature of the ablaut distribution in the inflection of these neuter nouns (nominative/accusative singular: full grade of root, 0-grade of stem-forming suffix; oblique singular: 0-grade of root, full grade of stem-forming suffix) is found in the parallel ablaut distribution found in the original inflection of the word for 'name', which like the nouns under discussion is a neuter noun, namely *nomn: *nmen-s etc.; cf. Anttila ibid.: 126-7 with references to earlier literature concerning the reconstruction of the inflection of *nomn. It would therefore seem that *gonu and *doru (as well as *sonu and *Aoyu) cannot be considered bona-fide members of the aberrant i-and ustems with "open" inflection. nails:32 G sg. n6es (96x) beside ne6s (19x) with prevocalic shortening of e; D sg. nei (99x); N pi. nees (52x) beside ndes (13x), both with secondary "root" accentuation comparable to that found in Skt. navas and rayas; A pl. neas (193x) beside neas (16x), also with secondary "root" accentuation; for attestations cf. Gehring 1891: s.v. For the derivation of this paradigm from a stem *neA-u-, cf. Szemerenyi 1956: 185-6. Notice, however, that as in Sanskrit, the nominative plural is phonologically ambiguous, derivable both from *neA-w-es and from *neA-ewes. huus:33 G sg. huros (14x) beside huieos (6x), D sg. huii (lOx) beside huidi (llx) and huiei (5x), N pl. huies (52x) beside huieeslhuieTs (18x), A pl. huzas (36x) beside huieas (8x). In addition, there are a few thematic forms of the type G sg. huioi.34 Compare Schwyzer 1939: 573-4. As Schwyzer (ibid.: 573, fn. 5) pointed out, the statistics On the secondary nature of the Homeric (Ionic) nominative singular nei1s, cf. Szemerenyi 1956: 158. 33 Attested three times in Old Attic inscriptions, in addition to a single occurrence of contracted his; cf. Schwyzer 1939: 574. Elsewhere, the nominative singular generally has been replaced by huids with o from *ui by dissimilation and with generalization of the -i- from the oblique; cf. ibid. 34 This does not include the starting points for this thematic inflection, namely N sg. hui6s (215x), A sg. huidn (170x), V sg. huie (29x), which are the regular Homeric forms in the paradigm of the word for 'son' and which owe their existence to the developments described in fn. 33.
32

triebene Geistreichelei. Viel Alliteration." It is therefore most likely that also this form is a hyperarchaism. forms are In (Homeric) Greek, the following found:30 6is: G sg. dios (3x) beside oids (5x) with accent on the case ending on the analogy of the root nouns (?),31 N pl. dies (5x), A pl. dis (7x) from *owins (cf. Argolic owins apud Schwyzer 1939: 573); for attestations cf. Gehring 1891: s.v. gonu: G sg. gounds (2x) beside obviously remade gottnatos (lx), N/A pl. goiina (12x) beside gotinata (56x), etc.; cf. Gehring 1891: s.v. For a discussion of the value of these forms, cf. the next entry in this list. ddru: G sg. dourds (19x) beside doziratos (Ix), 1) sg. douri (152x) beside dourati (lx), etc.; cf. Gehring ibid.: s.v. At first glance, the inflection of gdnu and ddru would seem to be excellent evidence for an original "open" reflection with genitive singular in *-w-os. It might be objected that, since compared to the Indo-Iranian type Skt. janu: ji6s*, daru: drds (Av. draos), sanu: sn6s, Av. dyii: yaos, the full grade of the root in the Greek genitive singular and other oblique forms must be secondary, generalized from the nominative/accusative singular, the suspicion arises that also the combination *-w-os of the stem-forming suffix plus case ending is innovated compared to the *-e/ow-s of Indo-Iranian. This objection might in turn be countered by the observation that the productivity of the type in *-elow-s and the moribund state of "open" inflection in Indo-Iranian make it more
30 I do not include in this list the dialectal type pdlis, G sg. pdlios, N pl. pdlies. The virtual limitation of this type to the i-stems, combined with the fact that the stemforming suffix always appears in vocalic shape (i), never as *y, makes it quite likely that Schwyzer (1939: 571) was correct in viewing this type of inflection to reflect a secondary transfer of the i-stems to the i-stems. Compare the similar developments in Sanskrit, where the transfer however remained largely optional and limited to the feminines. 31 Schwyzer (1939: 573) believed that the genitive singular form 6ios is "fiilschlich" for oids. The reason for this belief apparently lay in the oxytone accentuation of the genitive plural oidn. However, the constant root accent of Sanskrit duis, dvim, dvyas, combined with the Greek root accentuation in 6is, din, pl. dies, dis, as well as the actually attested root accentuation in G sg. 6ios, makes it more likely that also the Proto-Indo-European paradigm of the word for 'sheep' had root accentuation throughout, and that the end accentuation found in (some of) the Greek genitive singular and plural forms is secondary.

HOCK: Indo-Iranian
show that the forms of the type hufas are not secondary developments from the type huieas. That is, the type hulas can be considered more archaic.

Accusative Plural

91

the bona-fide members, forms with "open" inflection definitely outnumber forms with "closed"
inflection. This would seem to indicate that in

The bona-fide members of the i- and u-stems


with "open" inflection then are Skt. dri-, dvi-,

the early periods of Greek and Sanskrit, the bonafide members of the i- and u-stems with "open"
inflection by and large still resisted transfer to

rayi-, *kcftu-, krdtu-, ndu-, pasu-, pitu-, mddhu-, sisu-, Gk. o's, nalzs, huus. The Sanskrit and Greek sets overlap in the words for 'sheep' and 'ship'. Four of the words in question have clear attestations of predesinential O-grade in the nominative plural, namely dvi-, mddhu-, dis, and huus. Two further words, rayi- and nad-/nais, have attestations of nominative plural forms which are not incompatible with the assumption that in the "open" i- and u-stems the nominative plural originally had predesinential 0-grade, although they could also be derived from pre-forms with predesinential full grade. The two stems, however, are more valuable on a different count: They are the only
stems which, excepting the ambiguous dative singular and nominative plural (and the secondary "root" accent), have more than one form of the attested and show consistent paradigm "open" inflection in those forms, the genitive singular rayas being attested altogether 108 times, the accusative plural rayas 26 times, the genitive singular nauvs once, the genitive singular nods al115 times, and the accusative together plural neas/neas altogether 209 times. Other forms with are dvi- (with G sg. dvyas, only "open" attestations

the synchronically more productive "closed" inflection. As in the case of the aberrant n-stems, this resistance to analogical transfer to the productive
mode of inflection can be attributed to the mean-

ings of the words in question, at least in most cases. Again we find words with strong religious connotations (dri- and krdtu-) and words for cattle and other basic expressions for property (dvi-/ois, pdsiu-,36 rayi-). In addition, we find core-vocabulary items, namely terms for 'food and drink' (pitu-, mddhu-), and words for '(male) offspring, (9iSu-, huus). All of these words belong to spheres of the vocabulary which are notoriously resistant to analogical restructuring. As for the two remaining words, *kftu- and nau-/naus, it does admittedly not seem very likely that it is their meaning which preserved them from remaking.37 However, in the case of *kftu- the fact is significant that the only attestations are two occurrences of the fossilized accusative plural in adverbial function; for such fossilized case forms

with adverbial function notoriously preserve arof the "closed" endings -ins and -uns are modeled on the In adpattern of the productive "closed" inflection. dition, in the case of Indo-Iranian, the fact that the overriding tendency of this dialect group is to differentiate, not identify, nominative and accusative plural makes it unlikely that the endings -yas and -vas, which are identical with the corresponding nominative plural endings, are secondary. On the contrary, it strongly reinforces the view that the accusative plural forms -yas, -vas, precisely because of their phonological identity with the nominative

13x), *kftu- (with the original accusative plural kritvas, 2x), and pitu- (with G sg. pitvas, 9x). Elsewhere, encroachments of the synchronically productive "closed" inflection can be observed to a greater or lesser degree; cf. for instance D sg.
drate (lx), krdtave (lx), A pl. pasuan (lx), Gk.

6is (< *owins; 7x).35 However, in virtually all of

accusative plural forms of the type Skt. pasun, Gk. d6s more original than the forms of the type dryas, hulas, although Szemerenyi (ibid.) admitted the possibility that in some words, the accusative plural in Skt. (-yas,) -vas may be equally original and that the two types of formations may perhaps even have coexisted in Proto-IndoEuropean. Considering that the endings -ins, -iuns are the exclusive accusative plural endings in the productive i- and u-stems with "closed" inflection, it would a priori seem more likely that the endings *-yns, *-utns which are exclusively found in the "open" inflection are more original in that inflection and that the occasional occurrences

35 Some linguists, such as Schwyzer (1939: 573, fn. 5 [on p. 574]) and Szemer6nyi(1956: 174), have considered plural forms, are more archaic, and it suggests that the the
forms in -in, -in, as well as the nominative plural forms in (-ayas,) -auas, owe their existence not only to the pressure of the synchronically productive "closed" inflection, but also to the tendency to differentiate nominative and accusative plural. 36 For a view that the original meaning of PIE *pekuwas 'possession mobiliere personelle' rather than simply 'cattle', cf. Benveniste 1970. 37 With the sea-faring Greeks, nais may perhaps have been part of the core vocabulary. HIowever, in the world

of the speakers of Rig-Vedic Sanskrit the ship or boat hardly had the same significance.

92

Journal of the American Oriental Society 94.1 (1974) would have merged through regular phonological merger of the case endings; cf. (2.1,) 2.3, 2.4, 3.1, 3.8. (6) Such a morphological development leading to the differentiation of nominative and accusative (plural) is not without precedent; cf. 2.4. (7) It is not necessary to postulate two different models for the predesinential 0-grade, one for the r-stems and one for the other consonant stems. For the aberrant behavior of the r-stems can be explained as due to secondary influence from the productive i- and u-stems with "closed" inflection, that is, with regular accusative plural endings IIr. *-i-n?, *-u-ns; cf. 2.6 and fn. 28. 4.2 As the discussion in 3.1-9 has shown, only one linguist, namely Kurylowicz, has offered what can be considered a model worthy of discussion, motivating the replacement of predesinential full grade by 0-grade in the accusative plural of consonant stems. However, the discussion in 3.9 also has shown that Kurylowicz's model for the predesinential 0-grade of the majority of consonant stems is seriously defective. Rather than concluding, however, that consequently the hypothesis of the secondary nature of the Indo-Iranian predesinential 0-grade must be considered defective also, it would seem preferable to look for a less vulnerable model which avoids the difficulties involved with Kurylowicz's extraparadigmatic motivation of the 0-grade. Such a model can be readily found in the shape of the following INTRAparadigmatic model of predesinential ablaut grade relationships: N pl. -N sg. -v(cf. pita) (cf. pitaras)

chaic features of inflection. As for naui-/naus, its synchronically opaque inflection (like that of rayi-), may be considered responsible for the resistance to remaking, as well as the tendency to adopt root noun accentuation and inflection. 4.1 Before attempting to give a model for the secondary origination of the synchronically regular predesinential 0-grade in the accusative plural of consonant stems, it will be good to recapitulate the conclusions so far reached. (1) The end accentuation of the accusative plural is secondary, replacing a more original predesinential accentuation which actually is, as a rule, still preserved in the Rig-Vedic inflection of the nonablauting 0-grade root nouns; cf. 2.3, 2.4, 2.5, 2.10.1, 3.8 above. Considering the usual coincidence of full grade and place of accent in ablauting inflection, it is therefore more likely that the predesinential element had full grade than that it had 0-grade. It must, however, be admitted that the Indo-Iranian and Germanic evidence for a double 0-grade in the genitive/ablative singular ending *-f-s of the r-stems indicates that (accented) double 0-grades can occur in nominal inflection; cf. 2.10.1 and 3.6-7. The accent argument in favor of an original predesinential full grade therefore is not a very strong one. (2) The spread of end accentuation to the root nouns is tied up with whether or not the root ablauts; in ablauting roots it takes place earlier; cf. 3.8. (3) The Proto-Indo-European accusative plural ending -ns patently is to be derived from the accusative singular ending -m plus a plural marker -s. It is therefore to be expected that also the preceding stem would have originally been identical; that is, the accusative plural should originally have been a strong case like the accusative singular. Cf. 2.3, 2.11, 3.1, 3.6. (4) The accusative plural forms uksdnas, vfsanas, -hdnas, with predesinential guna, are archaisms, occurring in the paradigms of stems which are archaic also in other respects. This would seem to indicate that predesinential full grade is an archaism and that the synchronically regular predesinential 0-grade is an innovation. Cf. 2.10.3 and 3.8. (5) There is a plausible motivation for a replacement of an original predesinential full grade by the 0-grade, for this replacement leads to a differentiation of nominative and accusative plural which, if both originally were strong cases,

A sg.

-V(cf. pitdram)

A pl.

X = -0(cf. *pitOrds38)

I-ere the predesinential mora difference between the (bimoric) extended-grade vocalism of the nominative singular and the (monomoric) full grade of the accusative singular is considered to have entailed a similar mora difference between the (monomoric) full grade of the nominative plural and X = 0-vocalism in the accusative plural. The model accords well with Kurylowicz's law that in analogical remaking a "forme de fondation"
38 *pitrds in turn is later replaced by IIr. *pitrn? pre-Skt. *pitrnS > Skt. pitfn; cf. section 2.6 above. (But cf. also fn. 39 below.)

HOCK: Indo-Iranian founds a "forme fondee" (1945: 164), for in respect to the accusative plural, all of the three other formations in this model are "formes de fondation" The model also accounts for the fact (correctly observed by Kurylowicz) that the nonablauting root nouns were the latest to be affected by the new end accentuation of the accusative plural. For the end accentuation was only a consequence of the replacement of full grade by 0-grade, and this replacement was in turn motivated by the above intraparadigmatic ablaut model which the nonablauting roots by definition did not fit into. Conversely, the late spread of end accentuation to the nonablauting root nouns lends welcome support to the present hypothesis, which in this particular respect agrees with the (implicit) hypothesis of Kurylowicz, namely that it was considerations of ABLAUT (rather than of accent or some other feature) which formed the formal foundation for the differentiation of nominative and accusative plural. The possible objection that most consonant stems regularly have vrddhi in the accusative singular and nominative plural, not the guna of pitdram and pitdras, can be countered by assuming that the replacement of full grade by 0-grade took place before Brugmann's Law (whether this be considered a regular phonetic law or a morphological process) had brought about the vrddhi found in most accusative singular and nominative plural forms. That is, the replacement of full grade by 0-grade in the accusative plural must have taken place earlier than the generalization of vrddhi in the "aberrant" n-, original en-stems discussed in section 2.10.3 above.39

Accusative Plura,

93

The only potentially serious problem which I can see is that, as in Kurylowicz's hypothesis, the ultimate motivation for the replacement must have been the phonological merger of the nominative and accusative plural endings *-es and *-ns into -as. For those linguists who consider Brugmann's Law a regular phonetic development might hesitate to accept a merger of PIE *e and *n to a prior to the split of ablauting *o to a and a. However, even this objection is not insurmountable, since it might be argued that whatever phonetic characteristic (most likely 'half length'40)

lack of an intermediate stage *-rds would seem to be best explained by the assumption that the change of *-dras to *-rns preceded the general replacement of full grade by 0-grade in the (other) consonant stems. Since both morphological developments are limited to IndoIranian, there would seem to exist no compelling outside evidence against this relative chronology of events. As far as I can see, the only affect that this modified hypothesis would have on the hypothesis of the main body of the text is a very trivial one, namely that the (partial) paradigm of pita, used to illustrate the relationships in the intraparadigmatic ablaut model, would have to be replaced by a non-r-stem paradigm. 40 Cf. Cowgill 1968 for speculations that the original differencebetween Proto-Indo-European*e and ablauting *o was one of length (rather than timbre) and that the phenomena covered by Brugmann's Law are a consequence of the preservation of the original length of ablauting *o in open syllable, and the loss of that length in checked syllable. To make Cowgill's views fit the present hypothesis, it must be assumed that new long ("extended-grade")vowels arose in Proto-Indo-European, and that as a consequence the original length of the ab39 Considering that also the kinship r-stems are syn- lauting *o became half length. Although a system with chronically aberrant in showing predesinential guia in three degrees of length seems to be attested in Estonian, the accusative singular and nominative plural and that such systems are typologically sufficiently rare to be this aberrancy is no doubt also due to archaic retention considered quite unstable. It can therefore be assumed in basic (kinship) vocabulary, the fact that they do not that outside of Indo-Iranian the half-long ablauting *o show archaic retention of guna in the accusative plural soon merged with the nonablauting *o. In Indo-Iranian, may perhaps make it necessary to introduce the following however, its half length must be assumedto have survived modification of the present hypothesis: The formal con- the merger of non-high vowels into a-vowels. It can tacts with the productive i- and u-stems with "closed" then be assumed that subsequently also Indo-Iranian inflection led to the remaking of the inherited accusative eliminated the unstable system of three degrees of length plural ending (*-er-ns >) *-dr-as directly to -f-n?, rather by merging half-long [a.] with short [a] in checked sylthan through an intermediary stage *-r-ds. And unlike lable (following the universal tendency toward shortening the replacement of full grade by 0-grade in the other of vowels in that environment) and with long [a:] in open consonant stems which was motivated by the intrapara- syllable (following the universal tendency toward lengthdigmatic ablaut model, the replacement of *-dras by ening in that environment).- All of these developments, *-rng took place across-the-board, without any exceptions of course,have to be postulated only if the view is accepted even in basic vocabulary. This difference, as well as the that Brugmann's Law is a regular phonetic sound shift.

94

Journal

of the American

Oriental

Society

94.1 (1974)

*o Proto-Indo-European ablauting distinguished *o survived the Indo-Iranian from nonablauting merger of the non-high vowels into a-vowels, that the resulting (half-long ?) vowel continued to function as a short vowel as far as ablaut is concerned, and that it was only the SPLIT of this feature into [ + length] in open syllable and [- lenigth] in checked of prethe replacement syllable which postdated desinential full grade by 0-grade in the accusative plural of the consonant stems.41 REFERENCES Anttila, Raimo. 1969. Proto-Indo-European Schwebeablaut. Revised edition of 1966 Yale University Ph. D. dissertation. (University of California Publications in Linguistics, 58). Berkeley and Los Angeles: University of California Press. Arnold, E. Vernon. 1905. Vedic Metre in its Historical Development. Cambridge: University Press. Bailey, Charles-James N. 1969. Inflectional Pattern of Indo-European Nouns: Preliminary Edition. (Working Papers in Linguistics, 2: 1.) Department of Linguistics: University of Hawaii, 1970. Bartholomae, Christian. 1894. "Vorgeschichte der iranischen Sprachen." Wilhelm Geiger and Ernst Kuhn, eds.: Grundriss der iranischen Philologie 1: 1: 1-151. Strassburg: Trdbner, 1895-1901. Benveniste, Emlile. 1970. "Les valeurs 6conomiques dans le vocabulaire indo-europeen." George Cardona, et al. eds.: Indo-European and Indo-Europeans: Papers Presented at the Third Indo-European Conference at the University of Pennsylvania [19661. (Haney Foundation Series, 9). Philadelphia: University of Pennsylvania Press. Brugman[n], Karl. 1876. "Nasalis sonans in der indogermanischen Grundsprache." Curtius Studien 9: 285-338. Brugmann, Karl. 1892. Grundriss der veryleiclhendenGrammatik der indogermanischen Sprachen. 1st ed., vol. 2: 2. Strassburg: Trubner. -, 1904. Kurze vergleichende Grammatik der indogernmanischen Sprachen. Strassburg: Triibner. -, 1906, 1911: Grundriss der vergleichenden Grammatik For convincing arguments in favor of this latter view, cf. Cowgill 1957: 129-59. 41 An earlier version of this paper was read at the One Hundred and Eighty-Second Meeting of the American Oriental Society, April 19, 1972, at Chapel Hill, North Carolina. I am grateful to Profs. H. M. Hoenigswald, D. H. H. Ingalls, and L. Rocher for comments and suggestions made at that meeting. However, the responsibility for any errors is entirely my own.

der indogermanischen Sprachen. 2nd ed., vols. 2: 1 and 2: 2. Strassburg: Trabner. Burrow, Thomas. 1955. The Sanskrit Language. 1st edl. London: Faber & Faber. -, 1965. The Sanskrit Language. 2nd ed. London: Faber & Faber. Chantraine, Pierre. 1942. Grammoire homerique. Vol. 1. Paris: Klincksieck. Cowgill, Warren. 1957. The Indo-European Long- Vowel Preterits. Yale University Ph.D. dissertation. -, 1968. "A Speculative Reconstruction of the PreIndo-European Vowel System." (Abstract) of a Paper read at the Forty-Third Annual Meeting of the Linguistic Society of America at Newv York. (Meeting Handbook, p. 22). Debrunner, Albert. 1954. Allindische Grammatik. Vol. 2: 2, begun by Jacob Wackernagel. Gottingen: Vandenhoeck & Ruprecht. Debrunner, Albert, and Jacob Wackernagel. 1930. AllGrammatik. Vol. 3. Gottingen: Vandenhoeck & inmdische Ruprecht. Gehring, Augustus. 1891. Index Homericus. Leipzig: Teubner. Geldner, Karl Friedrich. 1923(-1929). Der Rig- Veda ... Postumously edited by Charles R. Lanmnan; parts completed by W. E. Clark. 3 vols. (Harvard Oriental Series, 33-35.) Cambridge, Mass.: Harvard University Press, 1951. zum R ig- Veda. Grassmann, Hermann. 1872. WVorterbuch Leipzig: Brockhaus. (Reprinted 1964 by Harrassowitz, Wiesbaden). Hillebrandt, Alfred. 1878. "Zur Lehre von den starken und schwachen Casus." BB 2: 305-35. I-irt, Hermann. 1927. Indogermanische Grammatik. Vol .3. Heidelberg: Winter. Kretschmer, Paul. 1892. "Indogermanische accent- und lautstudien." KZ 31: 325-472. Kurylowicz, Jerzy. 1945(-1949). "La nature des proces dits 'analogiques'." Acta Linguistica 5: 121-38. Reprinted in E. P. Hamp, et al., eds.: Readings in Linguistics. Vol. 2: 158-74. Chicago and London: University of Chicago Press, 1966. [My references are to the 1966 reprint.] -, 1956. L'apophonie en indo-europden. (Prace Jqzykoznawcze, 9). WVroclaw: Polska Akademia Nauk. , 1964. The Inflectional Categories of Indo-European. Heidelberg: Winter. --, 1968. Indogermanische Grammatik. Vol. 2. Heidelberg: Winter. Lanman, Charles R. 1877. "A Statistical Account of Noun-Inflection in the Veda." JAOS 10: 325-601. Mayrhofer, Manfred. 1956, 1964, 1970. Kurzgefasstes

HoCK:

Indo-Iranian

Accusative

Plural

95

etymologisches Worterbuch des IAltindischen. Vol. 1; Lieferung 18; Lieferung 21.1 Heidelberg: Winter. Osthoff, Hermann. 1876. "Zur frage des ursprungs der germanischen n-declination. (Nebst einer theorie uber die ursprtingliche unterscheidung starker und schwacher casus im indogermanischen.)" PBB 3: 1-89. -, 1901. "Zur armenischen Laut- und Wortforschung." Lukas v. Patrubany, ed.: Sprachwissenschaftliche Abhandlungen. Vol. 2: 49-136. Pott, August Friedrich. 1833, 1836. Etymologische Forschulgen auf dem Gebiele der indogermanischen Sprachen. 1st ed., vols. 1 and 2. Lemgo: Meyersche Hofbuchhandlung. de Saussure, Ferdinand. 1879. M1emoire sur le systdme primitif des voyelles dans les langues indo-europeennes. Leipzig. (Reprinted 1968 by Olms, Hildesheim). Schwyzer, Eduard. 1939. Griechische Grammatik. Vol. 1.

M[inchen: Beck. Senn, Alfred. 1966. Handbuch der litauischen Sprache. Vol. 1. Heidelberg: Winter. Szemer6nyi, Oswald. 1956. "Latin res and the IndoEuropean Long-Diphthong Stenm Nouns." KZ 73: 167-202. -, 1970. Einfiihrung in die vergleichende Sprachwissenschaft. Darmstadt: Wissenschaftliche Buchgesellschaft. Thieme, Paul. 1938. Der Fremdling ini Rgveda. (Abhandlungen fir die Kunde des Morgenlandes, 23: 2). Leipzig. Thumb, Albert. 1905. Handbuch des Sanskrit. Vol. 1. Heidelberg: Winter. Thumb-Hauschild. 1959. = Albert Thumb. Handbuch des Sanskrit. 3rd ed., rev. by Richard Hauschild. Vol. 1: 2. Heidelberg: Winter. Wackernagel, Jacob. 1896. Altindische Grammatik. Vol. 1 Gottingen: Vandenhoeck & Ruprecht.

Das könnte Ihnen auch gefallen