The Oblique Stem Marker in Plural Paradigms in Udi

The Oblique Stem Marker in Plural Paradigms in Udi:
Synchronic and Diachronic Evidence1
Alice C. Harris
SUNY Stony Brook
Table 1 shows a part of a noun paradigm in Udi that is typical of noun stems consisting
of a single syllable ending in a non-sonorant consonant.
Case
Singular
Plural
Nominative
mex
mexrux2
Ergative
mexen
mexrupon
Genitive
mexnaj
mexrupoj
Dative
mexnux
mexrupox
Table 1. Partial paradigm of mex ‘sickle’, after
Sixaruli¥e (1987: 43).
While there are many more than these four cases in Udi, limiting the discussion to these four
helps to focus on the issue of the oblique stem marker. The analysis of this paradigm provided in
Jeiranišvili (1971), Pančvi¥e (1974a), Sixaruli¥e (1987), Schulze (1982), Schulze-Fürhoff
(1994), Schulze (2001: 234) is summarized in Table 2.
1
The research reported here was supported in part by Grant Number BCS 0215523 from
the National Science Foundation, and I am grateful for this support. I thank Andrei Antonenko
for discussing the Arči data with me.
2
/x/ and /p/ in Udi are post-velar sounds. Except as noted, data presented in this paper
are from the Vartašen dialect; examples not otherwise marked are from the author’s fieldwork.
Harris, 2
Case
Singular
Plural
Nominative
mex
mex-rux
Ergative
mex-en
mex-rup-on
Genitive
mex-n-aj
mex-rup-oj
Dative
mex-n-ux mex-rup-ox
Table 2. Partial paradigm of mex ‘sickle’,
with traditional analysis.3, 4
The addition of genitive and dative suffixes to an “oblique stem” that consists of the root (mex)
plus the marker of the ergative case (-en) is a pattern found in most of the languages of the
Daghestan portion of the Nakh-Daghestanian family (Alekseev 1985, Kibrik 1991, etc.).
Traditional sources on Udi make the following observations about the forms (possibly mixing
synchrony and diachrony):
!
!
!
The formant of the ergative case, -en, is reduced to -n when the genitive and dative
suffixes are added; thus mex-n-aj < *mex-en-aj, mex-n-ux < *mex-en-ux, etc.
The final consonant of the exponent of the plural, here -rux, undergoes voicing when it
occurs between two vowels; thus mex-rup-on, etc.
In the plural, the vowels of the case endings, -en, -aj, -ux, etc. assimilate to the preceding
u of -rux/-rup; thus mex-rup-on < *mex-rux-en, mex-rup-oj < *mex-rux-aj, mex-rup-ox
< *mex-rup-ux, etc.
In this paper I argue that the analysis summarized in Table 2 is not the best analysis. I use the
approach pioneered by Alexandre Kibrik in his 1991 paper but distinguish singular patterns from
plural ones, as is necessary for examining the issue at hand. There are two distinct questions: Is
this the best analysis from a synchronic point of view? Does the analysis in Table 2 correctly
represent the diachronic origins of the forms of the plural? In §1 below I discuss the correct
synchronic analysis, and in §3 the diachrony of the forms. In this paper I focus on just the four
cases listed above because additional cases follow the model of the genitive and dative; thus
3
Sixaruli¥e (1987) presents most of the paradigms with the analysis I term “traditional”,
but on page 110 he proposes a different view, that the older form of the plural marker is -xo,
which is retained in the Nij dialect, and that up < ux < ox < xo.
4
Schulze (1994: 462) lists -o(y) as the genitive plural and -o as the dative1 plural, but on
page 463 lists a plural paradigm that seems to separate out a stem formant -o. The former
appears to be consistent with the analysis in Table 2, but the latter seems to agree with the
proposal summarized in Table 4 below.
Harris, 3
nothing is gained by including additional cases.
1. The Synchronic Analysis
Let us begin by considering in detail each of the statements above characterizing the
paradigm in Table 2. Udi e is reduced in a similar environments in the forms listed in (1), among
others.
(1)
Infinitive
ak’-es
aq’-es
q’i¨-b-es
ak’-s-un ‘seeing’
aq’-s-un ‘taking, receiving’
q’i¨ -b-s-un ‘being afraid of’
From the forms in (1) we can see that /e/ syncopates if (a) the (simple or complex) verbstem ends
in a single consonant and (b) -es is followed by -un, the formant of the masdar. On this basis, I
have written the rule as in (2) (Harris 2002: 82).
(2)
e-Syncope
e ÿ 0/ / VC + __sV
(2) is written here for a specific set of facts, rather than as part of a general phonology of Udi.
The partial paradigm in Table 3 shows that this can be generalized to apply to the /u/ of -ux,
before CV, as in (3).
Case
Singular
Plural
Nominative
gurat’
gurat’-ux
Ergative
gurat’-en
gurat’-p-on
Genitive
gurat’-un
gurat’-p-oj
Dative
gurat’-ax gurat’-p-ox
Table 3. Partial paradigm of gurat’ ‘shirt’, after
Jeiranišvili (1971:54), Sixaruli¥e (1987: 35).
This process is also found in words such as č:o-rx-on [face-PL-ERG],5 o-rx-on [grass-PL-ERG]
(Jeiranišvili 1971: 56).
5
The following abbreviations are used in this paper: ABL ablative, C consonant, DAT
dative, ERG ergative, FUT future, GEN genitive, INST instrumental, MAS masdar, NEG negative,
NOM nominative, OBL oblique, PL (in glosses) plural, PL (in the text) Proto-Lezgian, SG singular,
V vowel. = is a clitic boundary, while - or + represents a morpheme boundary.
Harris, 4
(3)
u-Syncope
u ÿ 0/ / VC __CV
(The rule in (3) does not, of course, preclude the application where the sequence is split by
morpheme boundaries.) We cannot, however, generalize the syncope rule to all instances of
these environments or to all vowels, since we find, for example ga-n-ux-o [place-OBL-DAT-ABL],
me-n-ax-o [knife-OBL-DAT-ABL] (Jeiranišvili 1971: 57-58). These satisfy the conditions in (3)
but do not undergo the rule.
(4)
(a)
g a- n- ux - o
V C uC V
(b)
me- n- ax - o
V C aC V
(4b) shows that the rule cannot be generalized to all vowels, in this case to a. In spite of the
exception in (4a), given the syncope shown in (1) and in Table 3, it seems clear that this process
accounts for the forms of the oblique stem formant in Table 2, -n.
The second process assumed in the traditional account is voicing of /x/ between voiced
segments, often both vowels. Table 3 shows the same process in the simple plural marker -ux,
parallel to the diachronically complex -rux. Although intervocalic voicing does not seem to
apply elsewhere in Udi, we know that it is a natural process found in many languages.
The third process posited is vowel assimilation; all vowels are said to assimilate to the u
of -ux, becoming o. This assumption is especially troubling if the vowel in the dative case
marker is u, as it may be in Table 2, or if it is i. The vowel in the dative case singular can be /a/,
as in Table 3, /u/, as in Table 2, /i/, as in äyz-ix, from äyz ‘village’, or /e/, as in k’a¨va¨n-ex from
k’a¨va¨n ‘meadow’ (Jeiranišvili 1971: 55); -ox is not found in the singular. The fact that u and i
are not phonetically likely to become o by assimilation to a preceding u makes this analysis
suspect. Further, the proposed assimilation is not a productive process in other similar
environments in Udi. For example, consider the paradigm in (5).
(5)
b-es-un
b-es-un-en
b-es-un-un
b-es-un-ax
[do-INF-MAS]
[do-INF-MAS-ERG]
[do-INF-MAS-GEN]
[do-INF-MAS-DAT]
‘doing (NOM)’
‘doing (ERG/INST)
‘doing (GEN)’
‘doing (DAT)’, etc. (Jeiranišvili 1971:107)
Like Tables 2 and 3, this is a declension, including a morpheme of the form -uC immediately
before the case suffixes, and the vowels of these suffixes do not change to o or otherwise
assimilate. This is a second reason to doubt the analysis summarized in Tables 2 and 3.
A third reason to doubt the assimilation proposed as part of the analysis in Tables 2 and 3
is that in general, assimilation in Udi seems to be regressive, applying from right to left, while
the proposed process would be progressive, applying from left to right. Probable examples of
regressive vowel assimilation (from Jeiranišvili 1971: 15) are šeno ÷ šono ‘he, she, it yon’ and
ipel ÷ epel ‘sheep’; in both words both forms are used. Consonant assimilation is more certain
Harris, 5
and is always regressive, as in efal=le ‘he will keep’ ² *ef-al=ne [keep-FUT=3SG] (note
occurrence of =ne in te=ne ‘he (is) not’ [NEG=3SG]), hametar=re ‘in this way he’ ²
*hametar=ne [in.this.way=3SG], Nij č’ap-p-e ‘he hid’ ² *č’ap-b-e. Note, too, that Dirr provides
examples of the assimilation of the plural marker -up, to the following -o, rather than the
direction that is posited in Table 2: ayl-op-ox ‘to the children’ (Dirr 1904: 91:5), ail-op-ox-olan
‘with children’ (Dirr 1904: 91:6). Because languages tend to have assimilation in one direction
only, the existence of regressive assimilation is troubling for the analysis in Tables 2 and 3,
which assumes progressive assimilation. Dirr’s examples make the analysis proposed in the
tables even more suspicious.
In this context, it seems wise to consider a second hypothesis. A different analysis of the
data in Table 1 yields Table 4.
Case
Singular
Plural
Nominative
mex
mex-rux
Ergative
mex-en
mex-rup-o-n
Genitive
mex-n-aj
mex-rup-o-j
Dative
mex-n-ux mex-rup-o-x
Table 4. Partial paradigm of mex ‘sickle’,
proposed analysis.
Table 4 posits a dual base analysis for both the singular and the plural. This is a basic pattern
found in the singular or the plural or both in a large number of Daghestan languages, including
the plural in Arči (Archi) on one possible analysis. The analysis in Table 4 is like that in Table 2
in assuming e-syncope in the genitive, dative, and other singular cases, and in assuming
intervocalic voicing of x. It differs from the analysis in Table 3 in not positing assimilation of all
vowels to u, being realized as o. Instead, o is posited as the underlying representation of all
oblique stem formants in the plural.
Case
Singular
Plural
Nominative
qIin
qIonn-or
Ergative
qIinn-i
qIonn-or-ča-j
Genitive
qIinn-in
qIonn-or-če-n
Dative
qIinn-i-s qIonn-or-če-s
Table 5. Partial paradigm of qIin ‘bridge’,
Arči, after Kibrik (1991: 256).
Udi has lost most of the locative markers that Kibrik includes in his discussion, and I therefore
Harris, 6
omit these from my discussion. Regarding cases cited here, Kibrik (1991) states that this
paradigm in Arči is characterized by the following facts.
!
!
!
!
“Case markers are usually special morphemes, which occur in final position and
do not depend on number: -n genitive, -s dative,....
“The singular has no special means of expression; the plural is the marked
number and is indicated by -or...; this marker -or is present in all forms of the
plural.
“The nominative has no special marker and its form coincides with the base form
of the appropriate number, singular or plural.
“In all forms of the singular paradigm, except the nominative, the [oblique stem]
marker -i appears; similarly in the plural we find the [oblique stem] marker
-čaj/če.”
(Kibrik 1991: 256-257)
The Udi paradigm in Table 4 has all of the characteristics quoted above from Kibrik, if we
abstract away from specific sound-function correspondences of morphemes:
C
!
!
!
Case markers are usually special morphemes, which occur in final position and do
not depend on number: -(e)n ergative, -(V)n or -(V)j genitive, -(V)(x) dative, etc.
The singular has no special means of expression; the plural is the marked number
and is indicated by -ux/up, -ur, -mux/mup, or a combination of these; at least one
of these markers is present in all forms of the plural.
The nominative has no special marker and its form coincides with the base form
of the appropriate number, singular or plural.
For many nouns, in all forms of the singular paradigm, except the nominative, the
[oblique stem] marker -n appears; similarly in the plural we find the [oblique
stem] marker -o.
(adapted from Kibrik 1991: 256-257 for Udi)
There is one difference, however, between Tables 4 and 5, which Kibrik did not particularly
draw attention to. In this paradigm in Arči, the ergative plural consists of the root, the plural
marker
-or, and the marker of the oblique stem, -če/-čaj. It is not altogether clear whether it contains an
explicit ergative case marker, -(a)j- or whether, on the other hand, -če and -čaj are allomorphs of
a single morpheme. The Udi ergative plural in Table 1, on the other hand, also consists of the
root, the plural marker -rux/-rup, the oblique marker -o; but in addition it contains an explicit
ergative marker, -n. Patterns with and without an explicit ergative marker are found elsewhere
in Daghestan languages.
If we look closely at the singular in Udi, we see that there are several different markers
for the genitive and dative cases, but I would like to set that aside for the moment to concentrate
on the patterns. We find the patterns shown in Table 6, which repeats those from the earlier Udi
tables.
Harris, 7
Case
Schema A
(Table 3)
Schema B
Schema C
(Table 4)
Nominative
gurat’
‘shirt’
me
‘knife’
mex
‘sickle’
Ergative
gurat’-en
me-n-en
mex-en
Genitive
gurat’-un
me-n-ej
mex-n-aj
Dative
gurat’-ax
me-n-ax
mex-n-ux
Alat
gurat’-ač’
me-n-ač’
mex-n-uč’
Super
gurat’-al
me-n-al
mex-n-ul
Table 6. Three schemas of singular declension in Udi
Both Schema B and Schema C may be termed “dual base” declension, since each has two bases,
mex, mex-(e)n-, and me, me-n-. Schema A has a single base, gurat’. Schemas B and C differ in
that the former has an explicit ergative case marker, -en, while in the latter the oblique stem
serves also as the ergative case. It is lexically determined which nouns follow which of these
Schemas; thus they may be considered declension types. The proposal embodied in Table 4, the
recognition of dual base declension in the plural in Udi, is made the more likely by the existence
of this pattern, Schema B, in the singular of nouns such as me ‘knife’.
Schema B is also found in other Lezgian languages, though it is less common than either
Schema A or C. Appendix A provides examples of declensions using Schema B in the singular
or the plural in Rutul, C’axur, Budux. Though not illustrated here, Schema B is also found in
other Nakh-Daghestanian languages, including Andi, Botlix, Γodiberi, Axvax, Hinux, Hunzib,
and Lak.
In the singular, realizations of the marker of the ergative singular in Udi include -n, -en, in, -jn. Realizations of the markers of the genitive singular include -j, -aj, -ej,-uj, -i, -un, -in; in
addition, in the forms that end in j, this consonant is optionally omitted, even when it is the only
marker of the genitive case (examples below). Realizations of the dative singular include -x, ax, -ux, -ix, -ex; in addition, final x can be omitted. Thus, if -o- is analyzed as the oblique stem
formant, as proposed in Table 4, we may assume underlying forms -n, -j, and -x. An alternative
analysis would posit underlying -Vn, -Vj, and -Vx and assume that any vowel (other than o)
elides immediately following o, as in (6).
(6)
V ! 0/ / o + ___ C
To test the reasonableness of both analyses proposed here, we can examine the
declension of Udi vowel-stem nouns in the singular. Vowel-stem nouns have either dual-base or
single base declension (lexically determined). A selection of declensions is given in Table 7.
Harris, 8
Case
‘mother’
‘husband’
s
brother’
‘face’
‘brother’
‘dog’
Nominative
nana
sevče
č:o
viči
xa¨
Ergative
nana-n
sevče-n
č:o-en
vič-e-n
xa¨-jn
Genitive
nana-j
sevče-j
č:o-e-j
vič-e-j
xa¨-j
č:o-e- vič-e-x
xa¨
x
Table 7. Singular declensions of vowel stem nouns. (Data from
Jeiranišvili 1971: 55, 56; Sixaruli¥e 1987: 67, 90, 97.)
Dative
nana-x
sevče-x
(In Table 7, the optionality of j and x is not indicated.) The frequency of these paradigms is
relevant. According to Jeiranišvili (1971) and Sixaruli¥e (1987), the types of nana ‘mother’ and
č:o ‘face’ are relatively common, while those of sevče ‘husband’s brother’, and viči ‘brother’ are
uncommon.6 The declension of xa¨ ‘dog’ appears to be unique to this noun. Although only
human nouns follow the nana ‘mother’ declension type today, this may not always have been
so.7 Of those presented in Table 7, the č:o ‘face’ declension seems to provide the best parallel to
the plural, yet here the vowel that follows o uniformly fails to delete. Still, this noun root is
monosyllabic, while the plural stems are all multisyllabic. The declension of viči ‘brother’ may
be analyzed as an example of vowel deletion, but it is not very relevant, since we cannot
reasonably make inferences about the elision of vowels in hiatus with o on the basis of the
treatment of vowels in hiatus with i.
A selection of vowel-stem nouns with dual-base declension is illustrated in Table 8.
6
Jeiranišvili (1971: 55) and Sixaruli¥e (1987:91) list the following nouns that decline
like nana ‘mother’: baba ‘father’, xala ‘mother’s sister’, dädä ‘father’s sister’ nävä ‘grandchild”
(one of two declensions in this dialect), personal names such as Olisa, Tamara, Kolja, Asya,
Nadja, Nina, and other nouns referring to people such as apa ‘aga, high official’, ust’a
‘craftsman’, tamada ‘toastmaster’. Like sevče ‘husband’s brother’ are sejne ‘husband’s mother’
and sejde ‘husband’s father’. Jeiranišvili (1971: 56) lists only xunči ‘sister’ as declining like viči
‘brother’. Jeiranišvili (1971: 56) lists ten nouns that decline like č:o ‘face’, but only two others
are o-stems (and one additional stem ends in o¨).
7
In C’axur, one ergative case marker is used for humans and another for non-humans; but
this appears to be an innovation in C’axur, not the reflex of an inherited property. It is likely that
the division we see in Udi is likewise an innovation peculiar to this language.
Harris, 9
Case
‘cloud’
‘millet’
‘knife’
‘place’
‘day’
‘name’
Nominative
haso
t’ajna
me
ga8
pi
c’i9
Ergative
haso-n-en
t’ajn-in-en
me-n-en
ga-n-en
pe-n-en
c’i-jen
Genitive
haso-nun
t’ajn-in
me-n-ej
ga-n-ej
pe-n-ej
c’i-j-ej
ga-npe-n-ax c’i-jux
ex
Table 8. Dual-base singular declensions of vowel stem nouns. (Data from Jeiranišvili
1971: 57-58; Sixaruli¥e 1987)
Dative
haso-n-ax
t’ajn-in-ax
me-n-ax
Among those listed in Table 8, only the declension of haso ‘cloud’ is common,10 and this
declension provides the best parallel to the declension of plurals. Because the vowel of the
oblique stem formant -en uniformly deletes following o in haso, we can reasonably continue to
consider the analysis that involves suffixes containing vowels, together with rule (6).
In summary, there are two kinds of problems with the analysis summarized in Table 2. (i)
The proposed assimilation is phonetically unlikely for two of the vowels that appear to constitute
underlying vowels in case markers in Table 2. (ii) In singular declensions containing some other
suffix with the vowel u, that vowel does not cause assimilation of the vowel of the following
case suffix in the way claimed to happen in the plural. The alternative proposed in Table 4 has
neither of these problems; further, it is supported by the synchronic remedy of o+V hiatus in
multisyllabic words shown in (6). Specific support for this resolution of hiatus is found in the
declension of haso ‘cloud’ in Table 8. I conclude that Table 4 provides a better analysis, from a
synchronic point of view, than that traditionally assumed. Whether the case endings contain a
vowel or not need not be resolved here; either possibility provides a reasonable analysis from the
point of view of Udi.
2. The Diachronic Analysis
Here I discuss only issues relevant to determining whether the analysis summarized in
8
There is a second way of declining ga ‘place’: ga, ga-l-en, ga-l-aj, ga-l-ax.
9
Also ergative c’i-en, dative c’i-ax.
10
The declension of ga ‘place’ seems to be unique to this noun. For the declensions of
‘name’ and ‘day’, Jeiranišvili (1971: 58) lists only a total of two parallels (and one cannot
determine there which of these two they follow). For the declension of me ‘knife’, four parallels
are listed, all e-stem nouns. In contrast, Jeiranišvili (1971: 57) lists eleven nouns following the
declension type of haso ‘cloud’; among these are o-stems (nine), one u-stem, and one e-stem.
Harris,10
Tables 2-3 is better than that proposed in Table 4. Is there diachronic evidence that supports one
analysis in favor of the other?
In addition to a complex set of spatial cases, Alekseev 1985 reconstructs for ProtoLezgian, the subgroup to which Udi belongs, the general declension schema in (7).
(7)
Reconstruction to Proto-Lezgian
Schema: Direct stem = Nominative case
+ -V-/-CV = Oblique stem = Ergative case
+ -n = Genitive case
or + -s: = Dative case
There are two distinct diachronic issues that must be addressed: (i) is the general schema of the
declension paradigm in Udi a direct reflex of one from Proto-Lezgian (PL), or has it been
restructured? (ii) Are the specific formants inherited or innovations? I take these to be entirely
independent questions, since I assume that it is perfectly possible for a language to retain the
general declension schema but replace some or all of the specific formants; likewise it is possible
for a language to retain reflexes of the formants but restructure the schema of the declension.
Though the questions must be kept distinct, they are best discussed together. With respect to
formants, the issues are the reconstruction of (a) the plural marker, (b) the markers of cases, (c)
the marker of the plural oblique stem. I discuss them in this order.
Citing Jeiranišvili’s unpublished dissertation on Rutul and Tsakhur (1966), Topuria (1973:
259) agreed that Udi -ux/up is a reflex of the oldest plural marker in the Lezgian languages and
proposed that other formants of plurality in Lezgian languages are reflexes of class markers.
Among the case markers, I discuss here only the ergative, genitive, and dative. The
ergative was probably reanalyzed from the singular oblique stem, as others have also suggested
(Bokarev 1960, Alekseev 1985, Schulze 2005, and others), not only for Udi but for a number of
other Nakh-Daghestanian languages. Among the specific markers of the (singular) oblique stem
(ergative case) reconstructed by Alekseev are *-n:e and *-ni. Alekseev does not include Udi -en
as a reflex of either of these two oblique formants. Like Udi, some other Lezgian languages
have some declensions where a reflex of one of these stem formants is an ergative case marker in
single-stem declensions (as in the declension of gurat’ ‘shirt’ shown in Table 3). For example,
C’axur has the declension shown in Table 9.
Case
‘sheep’
‘dog’
Nominative
va¨q’a¨
xjöä
Ergative
va¨q’a¨-n
xjöä-n
Genitive
va¨q’a¨-n//-ni
xjöä-n//ni
Dative
va¨q’a¨-s
xjöä-s
Table 9. Partial paradigms, after
Jeiranišvili (1983: 241).
Harris,11
Thus, the occurrence of Udi -en/-n is parallel to that of other reflexes of singular oblique stem
formants in Lezgian languages, in that it occurs both as a stem formant and as an ergative case
marker. As discussed above, Udi has only the formants -(e)n (homophonous with a marker of
the ergative case), -in (homophonous with a marker of the genitive case) and possibly -jn as
formants of the singular oblique stem; most of the other Lezgian languages show much greater
variety than this. On this basis, I suggest that the Udi singular oblique stem formants and
ergative case marker are both reflexes of a Proto-Lezgian morpheme along the lines of that
reconstructed by Alekseev (1985).11 What is relevant in the present context is that the o found in
the plural declension does not enter the paradigm in Pre-Udi through the ergative case marker.
That is, the origins of the ergative case marker show no evidence of containing o, and there is no
evidence that the o found in the plural declension was generalized from an earlier form of the
ergative case marker.
Udi retains reflexes of the reconstructed genitive *-n, which can be seen in the
declensions of haso ‘cloud’ and t’ajna ‘millet’ in Table 8 above, as well as in gurat’ ‘shirt’ in
Table 3. It also has an innovative genitive -Vj. Alekseev (1985: 45) observes that the origin of
this marker is not known, and he suggests PL *-§, a spatial case meaning ‘within’, while Schulze
(2005) discusses relationships to other spatial cases, including the ablative. The meaning of PL
ablative *-aj (Alekseev’s reconstruction, 1985: 48, also Schulze 2005) is, however, more in
keeping with its adaptation to the meaning of the genitive. The assumption that the PL ablative
was historically not adjacent to the oblique stem, where the modern genitive occurs, may be the
basis of an objection to this proposal; but the Udi declension is believed to have been completely
restructured with loss of the distinctive Nakh-Daghestanian spatial case system, and the meaning
of the ablative might have been retained and redirected under these circumstances. The question
relevant here is whether either of these markers of the genitive case could have introduced the
vowel -o, which we have seen functions as the plural oblique formant in the modern Udi
declension. It is unlikely that the *-n marker did so, as it is generally believed to have lacked a
vowel. The *-aj, however, could have introduced the etymon of the present-day o; see more on
this below.
Many accept the suggestion of Dirr (1928) and Bokarev (1960) that the dative case
marker of Udi has been lost and replaced by one of the spatial cases. While some reconstruct
only the consonant (e.g. Bokarev 1960: 47), Pančvi¥e (1940: 139) suggests that the underlying
form is -ax; I assume that he meant that it is the reconstructed form, *-ax. (See also Schulze
2005 on the origin of a in the dative.) If this suggestion is correct, *a of the one-time spatial
cases, now dative, provides a possible origin of modern o; see further on this below.
Thus, one hypothesis is that the o of the plural declension originated as the vowel portion
of the spatial case (reanalyzed as dative), *-ax, or of the innovative (Pre-Udi) genitive, *-aj.
Either of these assumes (i) a change in the value of the vowel, as well as (ii) reanalysis of the
vowel as a separate morpheme, an oblique stem formant, and (iii) restriction of the new
morpheme to the plural. I return to this hypothesis, Hypothesis 1, below.
11
See also Schulze (2005) on the variety of oblique stem formants in PL and on the origin
of the ergative case marker.
Harris,12
Other singular oblique stem formants reconstructed for PL include formants of the form
*CV or *V. Reconstruction of the latter (PL *-e, *-y, *-a, *-a¨) is attributed by Alekseev (1985:
27-28) to unpublished work by Starostin (see also Schulze 2005). A second hypothesis,
Hypothesis 2, is that PL *a,12 one of the singular oblique stem formants, was extended to the
plural and became the plural oblique stem formant, -o, in Udi immediately after (and possibly
also immediately before) a velar or postvelar consonant. Much work remains to reconstruct the
PL sound system, and it is not my goal here to contribute to this. However, the forms cited in
Appendix B show that there is some evidence to support the existence of a development of the
kind suggested here, from PL *a to Udi o. This hypothesis assumes (i) a change in the value of
the vowel, as well as (ii) extension of the morpheme from the singular to the plural, and (iii)
restriction of this morpheme to the plural.
It remains to reconstruct the pattern(s) of declension of nouns in the plural in PL. To do
this, we need to take into consideration differences in the Nij dialect, shown in Table 10.
Case
‘horse’
Nominative
e¨k-urux // e¨k-urux-o
Ergative
e¨k-urux-o-n
Genitive
e¨k-urux-o-j
Dative
e¨k-urux-o-x
Table 10. Partial plural paradigm, Nij,
Pančvi¥e (1974a: 67).
In the Nij paradigm with -o in the nominative, the vowel in fact becomes part of the plural
marker, since it is used in all forms; this then is an example of Schema A. It is likely that in Nij
the oblique stem formant was reanalyzed as part of the plural marker. If we assume that this is
correct, this declension is not relevant to reconstruction.
Either of the hypotheses proposed above can account for the origin of the plural oblique
stem formant -o in modern Udi, and at this time we have no evidence to prefer one over the
other. Perhaps the publication of Old Udi texts will provide additional evidence in favor of one
of these, or perhaps it will reveal the need for a further hypothesis. Neither hypothesis provides
any evidence that Table 4 does not provide the correct synchronic analysis.
3. Conclusions
Table 4 provides a synchronic analysis more in keeping with the phonology and
morphology of Udi than the traditional analysis summarized in Tables 2 and 3. The new analysis
is favored by the improbability of u and i assimilating to o after u and by the fact that vowels do
not assimilate to a preceding u in phonological environments similar to that hypothesized as the
12
For Schulze (1988) this has the value *c.
Harris,13
earlier form. The analysis in Table 4 is also supported by the fact that in general, assimilation in
Udi is regressive, not progressive as posited in the analysis summarized in Tables 2 and 3.
I have shown that o in the plural in Table 1 did not originate as part of the PL plural
marker, but that it could have been reanalyzed from the genitive or dative case. I suggested as an
alternative that it originated as a PL singular oblique stem formant, possibly *a, and was
extended to the plural in Pre-Udi. Whether or not one of these diachronic hypotheses turns out to
be correct, the analysis summarized in Tables 2-3 is inappropriate synchronically.
Appendix A
This appendix provides samples of declensions that follow Schema B in the singular or plural or
both, but not illustrating the full variety of formants for oblique stems, cases, or plurals.
Rutul
Nominative
Ergative
Genitive
Dative
Plurals
Nominative
Ergative
Genitive
Dative
C’axur
Nominative
Ergative
Genitive
Dative
Nominative
Ergative
Genitive
Dative
nin ‘mother’
nin-ä
nin-dc
nin-cs
[Schema A]
dux ‘son’
dux-ar-a
dux-ar-dc
dux-ar-cs
[Schema B]
jac ‘pier’
jac-ar-a
jac-a-d
jac-a-s
[Schema B]
nin-a-ba-r
nin-a-ba-š-ä
nin-a-ba-š-t’c
nin-a-ba-š-cs
[Schema B]
dux-jä
dux-räš-ä
dux-räš-dc
dux-räš-is
[other]
jac-bc-r
jac-mc-ra
jac-mc-d
jac-mc-s
[other]
iš ‘work’
iš-i-n
iš-i-n/na (56)
iš-i-s
[Schema B]
Singular
jaIq ‘road’
jaIq-:v-n
jaIq-:v-na/n
(also jaIq-v-na)
jaIq-:v-s
Plural
jaIq-bv
jaIq-bi-š-e
jaIq-bi-ši-n/na
jaIq-bi-ši-s
(Jeiranišvili 1967: 583)
Harris,14
[Schema B]
[Schema B]
(Kibrik et al. 1999: 56-57)
Budux
Nominative
Ergative
Genitive
Dative I
Singular
k’ul ‘house’
k’ul-:̌-er
k’ul-:̌-a/u
k’ul-:̌-a/u
[Schema B]
Plural
k’ul-ib-er
k’ul-(i)b-era
k’ul-(i)b-a
k’ul-(i)b-a
[Schema A]
(Pančvi¥e 1974b: 178-179)
Appendix B
All data below are from Nikolayev and Starostin (1994), and “page” refers to that work. The last
column represents Nikolayev and Starostin’s reconstruction. Other columns represent the vowel
in the cognate, by language, that corresponds to o in the Udi word in the first column. (L =
Lezgi, T = Tabasaran, A = Apul, R = Rutul, Ts = C’axur, B = Budux, Ar = Arči.) In sum, while
these data do not prove that PL *a > Udi o when following (or preceding) a velar or post-velar
consonant, they do show that such a development is worth investigating.
Udi
bo“Ia-esun ‘find’
boq:I ‘pig’
bost:un ‘cut, throw’
dömbä ‘corner,edge’
k:oda ‘wood shovel’
k:ori ‘hooked’
la¡o ‘up, on top’
q:oq: ‘throat’
qIol ‘bark, shell’
qo-š ‘behind’
¡od ‘tree’
¡oj ‘descendants’
¡o¡o-pesun ‘drizzle’
page
168
1047
272
401
688
631
551
909
789
1027
1079
577
1063
L
a
a
a
0/
T
a[?]
a
A
a
a
a
R
a
a
a
a
Ts
o
a
a
K
B
a/i[?]
a
c
a
o
e
Ar
o
a
a
a
a
a
w
i
a
0/
w
a
w
a
a
i
a
a
a
w
a
0/
a
e
u
a
a
w
a
0/
0/
a
u
v
w
a
v
a
a
a
a
0/
0/
uj
ä
o
i
w
i
a
a
ablaut
*a
*a
*a
*a
*wa
*ä
*a
*a
*V, *a
*wa
*wa
*a
0/ means that there is a reflex, but it appears to have no vowel in the corresponding position; a
blank in the chart indicates that there is no reflex, according to Nikolayev & Starostin (1994). In
some instances it is not clear to me whether the vowels in the reflex are thought to reflect the
reconstructed vowel, e.g. in Kryz ja=t. r ‘cut’, said to come from PL *§a t. wv.
Regarding the fourth item, note that dömbä is proposed only tentatively as a reflex of PL
*t:am[a]. Regarding the nex-to-last item, note that the proposed cognates of ¡oj represent verbs;
the Udi form is said to represent the same root; the proposed verbal reflex in Udi has no vowel in
this position.
Harris,15
References
Alekseev, M.E. 1985. Voprosy sravnitel’no-istoričeskoj grammatiki lezginskix jazykov:
Morpologija, sintaksis. Moskva: Nauka.
Bokarev, E.A. 1960. K rekonstrukcii padežnoj sistemy pralezginskogo jazyka. Voprosy
grammatiki: Sbornik statej k 75-letiju Akademika I.I. Meščaninova, 43-50. Moskva:
Akademija.
Dirr, Adolph. 1904. Grammatika udinskogo jazyka. (Sbornik materialov dlja opisanija
mestnostej i plemen kavkaza, vol. 33.) Tbilisi.
Dirr, Adolph. 1928. Einführung in das Studium der kaukasischen Sprachen, mit einen Sprachen
karte. Leipzig: Verlag der Asia Major.
Harris, Alice C. 2002. Endoclitics and the origins of Udi morphosyntax. Oxford: Oxford
University Press.
Jeiranišvili, Evg. 1967. Rutul’skij jazyk. Jazyki narodov SSSR, IV: Iberijsko-kavkazskie jazyki,
ed. by K.V. Lomtatidze (chief), E.A. Bokarev, Ju.D. Dešerijev, G.B. Murkelinskij, M.A.
Kumaxov, S.M. Xajdakov, A.K. Šagirov, 580-590. Moskva: Nauka.
Jeiranišvili, Evgeni. 1971. Udiuri ena. [The Udi language.] Tbilisi: University.
Jeiranišvili, Evg. 1983/1984. C’axuri da muxaduri (rutuluri) enebi, II: morpologia. [The
C’axur and Rutul languages, II: Morphology.] Tbilisi: Universit’et’i.
Kibrik, A. E. 1977. Opyt strukturnogo opisanija arčinskogo jazyka. Tom III, Dinamičeskaja
grammatika. Izdatel’stvo Moskovskogo Universiteta.
Kibrik, Aleksandr E. 1991. Organising principles for nominal paradigms in Daghestanian
languages: Comprative and typological observations. Paradigms: The Economy of
Inflection, ed. by Frans Plank, 255-274. Berlin: Mouton de Gruyter.
Kibrik, Aleksandr E., editor and organizer, Ja.G. Testelec, co-editor. 1999. Èlementy
caxurskogo jazyka v tipologičeskom osvešenii. Moskva: Nasledie.
Nikolayev, S.L., and S.A. Starostin. 1994. North Caucasian etymological dictionary. Moscow:
Asterisk Publishers.
Pančvi¥e, Vl. 1940. Ak’uzat’ivis genezisisatvis udur enaši [On the origin of the accusative in
the Udi language]. Enimk’is Moambe 5-6: 137-152.
Pančvi¥e, Vl. 1974a. Uduri enis gramat’ik’uli analizi. [A grammatical analysis of the Udi
language.] Tbilisi: Mecniereba.
Pančvi¥e, Vl. 1974b. 3iritad brunvata c’armoeba da punkciebi buduxur enaši. IK’E 19: 168179.
Schulze, Wolfgang. 1982. Die Sprache der Uden in Nord-Azerbajdžan. Wiesbaden: Otto
Harrassowitz.
Schulze, Wolfgang. Ms. 1988. Studien zur Rekonstruktion des Lautstandes der Südostkaukasischen (Lezgischen) Grundsprache. Bonn.
Schulze-Fürhoff, Wolfgang. 1994. Udi. The indigenous languages of the Caucasus, Volume 4
(Part 2 of North East Caucasian), ed. by Rieks Smeets, 447-514. New York: Caravan
Press.
Schulze, Wolfgang. 2001. The Udi Gospels: Annotated text, etymological index, lemmatized
concordance. Munich: Lincom Europa.
Harris,16
Schulze, Wolfgang. 2005. From case to case in Udi. Haptačahaptāitiš: Festschrift for Fridrik
Thordarson, ed. by Dag Haug and Eirik Welo, 251-256. Oslo: Novus Forlag.
Sixaruli¥e, T’ariel. 1987. Arsebiti saxelis bruneba udiurši [The declension of nouns in Udi].
Sak’andidat’o disert’acia, Sakartvelos Mecnierebata Ak’ademiis Enatmecnierebis
Inst’it’ut’i, Tbilisi.
Topuria, Guram. 1973. Saxelta mravlobiti ricxvis c’armoebisatvis lezgiur enebši. IK’E 18:
254-66.