The Morphophonology of Reduplicated Presents in Vedic
and Indo-European*
University of California, Los Angeles
1. Introduction and aims
The verbal system of Vedic Sanskrit employs a diverse array of reduplicated
formations that are indispensable for the reconstruction of reduplicated verbal
categories in Proto-Indo-European; see Kulikov 2005:43 1-5 and citations therein
for a linguistically current resume of all Vedic verbal formations with reduplication. Among these formations, the reduplicated present (Class I11 in the traditional nomenclature of Sanskrit present classes; cf. Whitney 1889:227-78) shows
the greatest diversity of vocalism in its reduplicating syllable.' Specifically, reduplicated presents exhibit a variation between i- and a-vocalism in the reduplicant
that lacks an immediate phonological or morphological explanation. For most
roots that form reduplicated presents in the Vedic period, the reduplicant vocalism is stable by lexeme; for instance, ddii 'give' and dbhas 'chew' build
3.sg.pres.act.ind. forms dddiiti and bdbhasti, whereas dbhr 'bear' and dvac
'speak' build 3.sg.pres.act.ind. forms bibharti and vivakti, respectively. Nevertheless, a few roots show an alternation between i- and a-vocalism in the reduplicant, e.g., 3.sg.pres.act.ind. si&i but 3.~1.sascati to dsac 'follow', or 3.sg.
pres.act.ind. jigiiti but (fossilized) pres.part.act. jdgat- to dgd 'go'.
Although reduplicated presents are non-productive already from the earliest
phase of the Vedic language, for the reduplicated present stem to be a synchronically non-derivable, lexicalized fossil is unhappy, because the burden of accounting for all the peculiarities of its inflection then falls on the prehistory of the
formation. Precisely in this vein, some scholarship (e.g., LIV:17, Tichy
2006:1134, and Jasanoff 2003:128-32) projects the distinction between Vedic
*
1
Thanks to Andrew Byrd, Bruce Hayes, Alexander Lubotsky, and Kie Zuraw for profitable
discussions on various aspects of this paper. I owe further thanks to the editors for their diligent work, especially Stephanie Jamison. Responsibility for any remaining infelicities or misjudgments rests solely upon me.
Henceforth, I will use the terms "reduplicant" and "base" to refer to what one might otherwise
designate as the "reduplicating syllable" and the "root." I ignore any theoretical implications
that the former set of terms may have.
Stephanie W. Jamison, H. Craig Melchert, and Brent Vine (eds.). 201 1.
Proceedings of the 22nd Annual UCLA Indo-European Conference. Bremen: Hempen. 223-54.
Ryan Sandell
a-reduplicating and i-reduplicating presents back to distinct PIE types with *eand *?"-reduplication.Such a reconstruction simplifies the morphological and
phonological problems of inflection for the Vedic reduplicated present, but introduces a problem of semantic distinction between the two types; furthermore, the
reconstruction of two PIE types does not account for the u-reduplication found in
Indo-Iranian in a direct way.
The approach and objective of this paper are to attempt a systematic morphophonological account of the reduplicated present in Vedic that can capture the
variations in reduplicant vocalism at the synchronic level. I will further discuss
the implications of the reduplicative process at work in Vedic for
theories of reduplication situated within the larger framework of generative phonology
and
the reconstruction of the reduplicated present for Proto-Indo-European.
count of every detail in the inflection
nts. Rather, this work primarily seeks
ion that can generate the Vedic data
well a given stem is attested in the
this paper suggests how one might
ccount for certain forms that are not
eptible to the synchronic analysis.
flectional and phonological classification of the Vedic reduplicated
section introduces the sources of data on reduplicated presents in Vedic employed for this study, the criteria for classification of verbal forms as genuine reduplicated presents, and a classificatory scheme of the reduplicated presents.
drawn data concerning the inflection of the Vedic reduplicated present
the Rgveda (RV), via the concordance of Lubotsky (1998);
aveda (AV), via the index verborurn
The Morphophonology of Reduplicated Presents
225
lemmata of Whitney 1885, for forms from Vedic texts other than the
above-mentioned.
Relatively few relevant forms are attested outside of the RV or AV, and indeed
many traces of reduplicated present inflection for a given root are entirely limited
to the RV. Whitney (1885:212-3) lists forty-nine roots as attesting to the stem of
an athematic reduplicated present. However, Whitney's classifications include
some forms that are properly to be understood as pluperfects or modal forms to a
perfect stem. I apply the following criteria to determine whether a given root has
a reduplicated present:
1. If an unmistakable perfect stem is attested:
(a) forms with primary endings, active or middle, in indicative mood,
clearly indicate a reduplicated present stem, even if ultimately derivative of a perfect;
(b) if no forms with primary endings, or only modal forms are attested,
those forms probably belong to the perfect stem;
(c) if the perfect stem to the root shows a-reduplication, a stem with ireduplication may reflect a reduplicated present stem.
2. If no unmistakable perfect stem is attested, then an imperfect or modal
form may constitute evidence of a reduplicated present.
Based on these criteria, one should exclude ddhi 'think', dpi 'swell', dvyac 'surround', and probably ^man2 'waif2 on the basis of criterion l(b). At the same
time, however, criterion l(a) then demands that one consider bibhiti (to dbhi
'fear') and mamatsi (to dmad 'enjoy7) as reduplicated presents, despite the fact
that such forms do ultimately derive from perfects; cf. Cardona 1992 and Insler
1972.~Table 1 presents those roots that attest to the synchronic formation of
athematic reduplicated presents.
2
3
Provided that the pluperfects and modal forms with the sense 'wait' ultimately derive from
^man, 'think'; see the discussion in Kiimmel 2000:364-6. If one insists that these forms belong to a separate root, then one may consider these forms as attesting to a reduplicated present
based on criterion 2.
Indeed, that these derived reduplicated presents seem to conform to the morphophonological
specifications for reduplicated presents given below suggests that the reduplicated present was
not an absolutely closed and lexicalized class in the early Vedic period.
Ryan Sandell
^^Iga
x
x
* hser-
dghr
dci
'move'
'go'
'drip '
'observe?
*g''eh2*gH*er*kei-
gh&citb-
djus
'enjoy'
*gep
jkta-, jus&i-
dtq
'be sharp'
* (s)tejg-
tiktd-^
Â¥to
'give'
*dehr
ddii
'point'
*de&
dattd-, -data-,
-ttadisci-
ddha
ddhr
'put'
'hold'
*dhehl*dher-
hitd-, -dhitadhrtd-
dnij
'wash'
*nelgw-
nikta-
dni
'lead'
*neiff-
nitd-
dpr
^jprc
'carry, keep'
'mix'
*per*perk-
prkta-
dbhas
dbhi
dbhr
dmad
'chew'
'fear'
'bear'
'enjoy'
*bhes*bhe&*bifer*med-
x
bhasita-c
bhitdbhrtdmattd-AV
iyarti
jigati
jigharti
2.sg.pres.act.ind.
cik&iAV
3 .pl.pres.act.subj.
jujusan R
2.sg.pres.act.impv.
titigdhiMsH
dddati
3.sg.impf.mid.inj.
didiga
dddhati
2 .pl.pres.act.impv.
didhrtd
2.pl.pres.act.impv.
ninikta H
2.du.pres.act.ind.
ninithds
piparti
2.sg.pres.act.impv.
piprgdhi R
bhbhasti
bibhktiv
bibharti
2.sg.pres.act.ind.
Continued on nextpage
Notes: X = unattested; R = rare (i.e., the stem is known from three or fewer forms), H = hapax legomenon. All forms are RV and 3.sg.pres.act.ind. unless otherwise marked: AV = Atharvaveda, MS = Maitrayam Samhitii, V = Vedas, C = Classical Sanskrit. Accents are marked only
when attested.
Glosses generally follow Whitney (1885) and Kiimmel(2000).
a
b Forms of PIE roots follow LIV.
The Morphophonology of Reduplicated Presents
Table 1-Continued
-
-
--
-
Skt. root
Gloss
'measure'
^ma\
dm&
dmi
'low7
'(ex)change'
dyas
^Ira
'(be) heat(ed),
boil'
'separate'
'give'
dvac
dvai
'speak'
'wish, want'
&is
4 5
dsac
'handle'
'sharpen'
'follow'
&as
dsu
'sleep'
'press'
dsr
dhas
dh'i
^Ihu
'let loose'
'laugh'
'leave behind,
depart'
'pour, offer'
'call'
dhri
'be ashamed'
^YU
dhu
PIE root
Past pass. part.
yutdrat&
vistaSitd-
x
s&-~
hasita-^
hanu-^, hind-,
jahitdhutdhad*gi'rejH-
hrit~-~
Class I11 example
3.sg.pres.act.impv.
mimiitu
mimati
3.du.pres.act.ind.
mimitas R
3 .sg.pres.act.impv.
yayastu H
yuyoti
2.sg.pres.act.impv.
ririhi
vivakti
vivagi H, 2.sg.pres.
act.ind. vav&i H
vivesti
SiSZti
si&i, 3.pl.pres.act.
ind. sdicati
sasdstiTsR
3.pl.pres.act.ind.
susvati H
sisarti
pres.part.act.jdkyat- H
jdhzti, 3.sg.pres.mid.
ind. jihite
juhbti
1.pl.pres.act.ind.
juhumdsi R
jihretiB
Notes: TS = Taittiriya Samhita, AB = Aitareya Brahmans, B = Brghmanas, E = Epic Sanskrit.
Some forms given here require further comment:
jujusan (RV 7.61.6d) to djus could easily be a perfect subjunctive,
though all other RVic perfect subjunctives to this root show full grade of
the base. Also notable is 2.pl.preslperf.act.impv. jujustana. The forms'
interpretation as reduplicated presents thus mostly relies on the etymological connection with Hittite ku-ku-us-zi 'tastes' and Young Avestan
Ryan Sandell
3.sg.pres.mid.hd. (ii-)ziizuite; cf. Watkins 2003:391. LIEl66-7, on the
other hand, does not believe that any Indo-Iranian forms continue an old
reduplicated present.
titigdhi to dtij is a nonce form, though with LIE593 "im Kontext fimgiert
titigdhi eindeutig als zu tijas- n. 'ScKarfeY Glut, Energie' gehoriger
Prasensstamm." titigdhi is a synchronically valid Class 111present.
The imperatives didhrtam and didhrtd most probably belong to the better
attested reduplicated aorist stem didhar-. didhrtam (RV 5.86.6eQ should
scan with long i in order to fit the lines' cadences; didhrta (RV 1.139.8g)
may scan with short i.
Note that, pace Whitney (1885:lOO), dP? Yilly (< *plehl-) probably does
not have a reduplicated present. Having read all the occurrences of the
stem p@ar-/p&r- in the RV and AV ~aunaka,I find that only AV
~aunaka1.35.4b (piparmi) lends itself well to an interpretation 'fill'. Furthermore, as Whitney (1905:520-l, 636) already mentions, the AV
Paippalada sometimes reads bibharti where AVS has piparti.
Both vivasti (RV 7.16.1 lb) and vavdk.yi (RV 8.45.6b) are problematic.
EWA:527 identifies both forms as ''Augenblicksbildungen"; Joachim
(1978:151-2) likewise suggests that both forms are nonces; LIE672-3,
misreading Joachim 1978, identifies only vivasti as an Augenblicksbildung? and tentatively supports a link between vavbbi and Hittite wewakki; Lubotsky (1998:1246) marks vavhksi as a nonce, but not vivasti.
A 3.sg.pres.act.impv. vivastu is found in the SSmavedaYbut in a mantra
which is a borrowing of RV 7.16.11. According to the synchronic principles of formation developed in $4 below, vavhbi has aberrant reduplicant vocalism, but could preserve an archaic accentuation; on the other
hand, vavabi could be built directly to a perfect stem, with perfect accentuation, like mambttu to dmad, but the perfect of dvai is rare (3x
RV), and in all cases shows a long reduplication vowely thus vGvai-;
see further Kummel 2000:477-80. In my assessment, the stem vivai- is
more likely the synchronically "grammatical" Class 111 formation, but
that Vedic inherited any reduplicated present to dvai is very doubtful.
Table 2 displays those reduplicated presents that also show thematization, but
exhibit phonological patterns that indicate the synchronic productivity of a reduplicated present stem:
The Morphophonology of Reduplicated Presents
Table 2. Thematized reduplicated presents
Skt. root
dghrii
Gloss
'sniff, smell'
PIE root
*gwhrehI-
Past pass. part.
ghr~th-~
dnas
'return'
*nes-
X
dpad
'go, step'
*ped-
pannh- AV
dmyab
dsthii
dhan
'affixy
'stand'
'strike'
*me&-(8)*steh2*gwhen-
x
sthit&hat&
Class I11 example
3.pl.pres.act.ind.
jighrantiMs R
3.sg.pres.mid.ind.
nihsate
pres.mid.part.
pibdamcna- H
mimibati R
t&hati
3.sg.pres.mid.ind.
jighnate
Notes: see notes to table 1.
Whether the reduplicated present stem to dghrii is thematic or athematic is not
entirely clear. Whitney 1885:43 reads: "jigrati AV.MS (3p.)." However, Whitney
(1905:694) acknowledges that the form in the AV (12.4.5) "can be ... either singular or plural." Furthermore, Amano (2009:556n2426) follows the emendation
of von Schroeder (1881) in reading the MS manuscript form @ratgjighrati as
@ratgjighranti. Kiithaka Samhitii forms (e.g., 3.sg.pres.act.opt. (ava)jighret, 2.sg.
pres.act.impv. (ava)jighra) also suggest thematic inflection.
Table 3 shows further forms that must historically continue reduplicated presents, but in which sound changes either eliminated a clear morphological division separating the reduplicant and the base or obscured an obvious phonological
relation between reduplicant and base.
Table 3. Historical reduplicated presents
Skt. root
daj > d q
Gloss
'drive7
PIE root
*h2e&
dr > dir
'move'
*h3er-
Past pass. part.
- ~ j i t a - 'H
~
Class I11 example
3 .sg.pres.mid.ind.
ijate R
3.sg.pres.mid.ind. hte (cf.
karti in table 1 above,
to which h e remains
the corresponding middle in early Vedic)
Continued on next page
Notes: $B = ~ata~atha
Briihtna~a.See notes to table I.
Ryan Sandell
dih
d p ~
&ad > &id
'praise'
'be eager'
*h2ejsd*He@"-
idithihit~-~
'drink'
'sit'
*peb
*sed-
pithAV
sannh-
2.sg.pres.mid.ind.
ihaseVsH
pibati
s fdati
Notes: VS = VFijasaneyi S ~ h i t i i See
. notes to table 1
Vedic speakers had probably reanalyzed these forms as either root (Class 11) or
thematic (Class I) presents to entirely different roots. One need not, therefore,
bring to account this last group from the synchronic point of view, though these
forms may contribute data to the Indo-European picture.
2.2. Phonologicalpatterns between root and redzplicunt
When we examine the data for relationships between the vocalism of the reduplicant and vocalism or other structure of the root, the following patterns emerge:
Roots that take a zero grade with u-vocalism invariably reduplicate with
u-vocalism (e.g., juhbti).
Roots that take a zero grade with i-vocalism invariably reduplicate with ivocalism (e.g., cikisi).
Roots that take a zero grade with y-vocalism invariably reduplicate with
i-vocalism (e.g., bibharti).
Thematic reduplicated presents invariably take zero grade of the root in
the base but reduplicate with i-vocalism (e.g.? tisthati).
Roots of other structures (i.e.? CaC or Cii) show either i- or a-vocalism in
the reduplicant, but some fbrther subregularities are observable:
1. CaC-roots in final coronal (i.e., dbhus, dnzad, dyas, and &as) always
show a-vocalism in the reduplicant.
2. dvac, dvui, and dsac typically show i-vocalism in the r e d ~ ~ l i c a n t . ~
Significant exceptions are the 3.pl.pres.act.ind. sdicati and corre-
4
See the discussion on viva+?i
and vavabi following table 1 above.
The Morphophonology of Reduplicated Presents
23 1
sponding pres.part.act. sbicat- to d ~ a cI; will
~ take up these forms in
$5.
3. CZ-roots (i.e., dg5, dd5, ddh5, dm&, dm&, dr5, d i ~ and
, dh5) exhibit the greatest diversity of reduplicant vocalism: dd5 and ddhZ invariably reduplicate with a-vocalism; drZ shows a-vocalism of the
reduplicant except in the 2.sg.pres.impv.act. ririhi; dh~?divides its
reduplicant vocalism into a for the active paradigm but i for the middle paradigm; the remaining four roots always show i-vocalism of the
reduplicant.
In the main, the vocalism of the reduplicant is predictable based on either a morphological criterion (if thematic, the reduplicant takes i-vocalism) or a
phonological criterion (root structure). To that extent, these patterns constitute
"Islands of Reliability" in the sense of Albright (2002). Namely, where a genera1
morphological pattern does not predominate, phonological subregularities may
come into play in order to determine the mo~hologicalexpression of a category.
Indeed, the vocalism of the reduplicant very closely corresponds to the vocalism
of the zero-grade allomorph of a given root as it appears in the past passive participle (see tables 1-3) or other morphological category that regularly takes zero
grade of the root. For instance, even in the diverse CZ-roots, diZ builds a PPP
&a- and a reduplicated present iiiiiti, while dd5 attests a PPP -d&a- or -tta- in
~) a
compounds (instead of perhaps expected %ditb-< *db3-tb-; cf. Greek 6 0 ~ 6 and
reduplicated present dhdZti. Regardless of the precise historical origin of the reduplicated present, then, such "Islands of Reliability" suggest that, at some point
in the prehisiory of Vedic (jerhaps Indo-Iranian), subcategories of reduplication
relying on common phonological patterns established themselves. These patterns
alone, however, do not constitute an account of how a Vedic reduplicated present
is built in the process of morphological and phonological derivation: unless one
can provide both historical and theoretical bases for these patterns, their collection here is little better than the listing of aberrant reduplicated forms found in
Whitney 1889:665-6 or Macdonnell 1916:144-5.
3. Reduplication cross-linguistically
The data presented in $2 pose a difficult problem in that many approaches to
reduplication, whether in the context of generative phonology or historical
5
Note that the participle is homonymous except in accent with the more common participle to
&ac 'dry up' (a-)sadist- (14x RV).
232
Ryan Sandell
linguistics, presuppose that the defining relationship between base and reduplicant is of a phonological nature. See, for instance, Wilbur 1973, Marantz 1982, or
even the more recent models of McCarthy and Prince 1995 and Raimy 2000. Although reduplication could be defined purely in terms of a phonological copying
mechanism, such a definition would immediately exclude the Vedic reduplicated
present, which does not build reduplicants of absolutely reliable phonological
form. Furthermore, most reduplicative processes known cross-linguistically are
morphophonological processes: the morphological marking found on a category
gives phonological expression to some syntactic or semantic value, but the,
phonological expression of the morpheme itself depends upon independent
phonological operations. A purely phonological definition of reduplication, then,
would not capture an essential aspect of many reduplicated categories across languages generally. Hence, Inkelas and Zoll (2005:2) carefully distinguish between
two distinct linguistic processes, which nevertheless bear some surface similarity
to one another:
1. Phonological Copying
2. Morpho-Semantic Feature Reduplication [Morphological Reduplication]
In the succeeding subsections, I will define and expand upon these two categories.
3.1. Phonological copying
Phonological Copying entails that the phonological features of some segment are
reproduced in another position in a word in order to satisfy structural demands of
a purely phonological nature, e.g., that no syllable lack an onset consonant or that
no syllable lack a coda consonant. Inkelas and Zoll(2005:2-3,20-I) cite cases of
this phenomenon in Hausa, Yoruba, and Spokane (Interior Salish); in examples
(1)-(2) I reproduce the Spokane case, in which an affix -e- meaning REPETITION
(REP)is added to the verbal root (data are taken from Black 1996:210ff.; cf. also
Bates and Carlson 1997:105-7). If the verbal root is of the form CC at the stage
in the derivation when -e- enters, the -e- is simply infixed; see (la)-(b), where
$1'- and 12'- become $-e-1'- and 1-e-6'-. If instead the verbal root is of the form
CVC at the stage in the derivation when -e- enters, the -e- is prefixed, and the
first consonant of the root is copied to provide an onset to the syllable, thus resulting in a form ClVC,V c see (2a)-(b), where id'- and nit'- become $-e-iilJand n-e-n 'it-.
The Morphophonology of Reduplicated Presents
Repetitive -e- infixes into an initial consonant cluster
(1)
a. I-e-, 3'-n'-t-an'/ + i-e-l'n 'ten '
REP, chop-CONTROL-TRANSITIVE-1
SGTRANSITIVESUBJECT
+ 'I cut it Up
repeatedly'
b.
I-e-, 1C'-n7-t-an'/+ 1'-e-c'n 'ten '
REP, tie-CONTROL-TRANSITIVE-~SGTRANSITIVESUBJECT
-+ 'I tied it Over
and over'
Phonological copying provides onset for -e-
(2)
a. l-e-, gal'/ +i-e-iil '
REP, chop -+ 'I cut up repeatedly'
b.
I-e-, niC7-n'-t-axw/+ n '-e-n 'En 'txM'
REP, cut-CONTROL-TRANSITIVE-~SGTRANSITIVESUBJECT
+ 'you kept On
cutting it'
The seeming reduplication in the examples under (2) is merely a phonological
repair to a wholly phonological problem: Spokane does not allow syllables without onsets, but rather than inserting some default consonant, the features of the
nearest consonant are copied into that onset position.
3.2. Morpho-semanticfeature reduplication
Work on reduplicative constructions that encode a syntactic or semantic value has
typically identified two major variations in phonological expression:
1. Fixed Segmentism: some phonological segment of the reduplicant remains constant throughout the paradigm of that reduplicated category,
regardless of the base's phonological constituency.
2. Phonological Correspondence: every segment of the reduplicant has a
correspondent in the base, though the segments of the reduplicant may
undergo changes through identifiable phonological processes.
Inkelas and Zoll (2005) further consider constructions with semantic identity, but
no necessary phonological relationship whatsoever, i.e., pure Semantic Doubling.
3.2.1. Fixed segmentism. Examples of fixed segment reduplication are readily
available in the form of the reduplicated presents and the perfect of Ancient
Greek, which are cognate with the Vedic reduplicated present and perfect. In
the reduplicated presents, one finds i-vocalism in the reduplicant throughout the
Ryan Sandell
entire paradigm, regardless of the vocalism found in the root; likewise, the Greek
perfect always displays e-vocalism in the reduplicant:
(3)
Greek reduplicated present
Root
-
doq 'puty
d8m 'give'
dpev remain'
(4)
1.sg.act.
Â¥ri0q
5i5mpi
pipvm
1.ul.act.
Â¥riQepe
Wiopev
pipvopev
1.%act.
Xskoma
7c67coda
116qevya
1.vl.act.
XeXoizapev
zezoi0apw
m<peVyapev
Greek perfect
&em 'leave'
ditei0 'convince'
"^/my 'flee'
Reduplication with fixed segmentism, however, is not restricted to the vocalism
of the reduplicant; Tubatulabal (Uto-Aztecan, California), for instance, exhibits
fixed segmentism in its reduplicant consonantism:
(5)
Tubatulabal (after Alderete et al. 1997:24; data from Vogelin 1958)
Base form
pi:*
le:win
piJIka
?a:ba?iw
Reduulicated form
?i:-bi:fin
?e:-1e:win
?i-pifika
?a:-?a:ba?iw
Q~ES
'he is snoring'
'to pack it'
'to slip'
'it is showing'
Here, the reduplicant faithfully copies the vocalism (including length) of the base,
but the onset consonant of the reduplicant universally appears as a glottal stop,
regardless of what the initial consonant of the base is.
3.2.2. Phonological correspondence. Instantiations of phonologically correspondent reduplication abound cross-linguistically, at least in comparison to fixed
segment r e d ~ ~ l i c a t i o Under
n . ~ the definition given above, phonologically correspondent reduplication may encompass both "full" (i.e., every segment of the
base also has surface expression in the reduplication) and "partial" (i.e., the redu6
Alderete et al. 1997 remains the only treatment in the linguistic literature, to my knowledge,
specifically to develop an account of fixed segmentism. I report anecdotally that cases of fixed
segment reduplication, despite the familiarity of the Greek examples, appear rather infrequently in linguistic literature. Moreover, queries to the Graz Reduplication Database (= Hurch
2005) yielded approximately three times as many constructions without fixed segmentism as
with fixed segmentism.
The Morphophonology of Reduplicated Presents
235
plicant is somehow phonologically reduced) reduplication. Example (6) shows
true "full" reduplication; example (7) demonstrates that the Vedic intensive follows a pattern of "partial" reduplication:
(6)
Indonesian noun pluralization (data from Cohn 1989:185)
hhk-hhk
bUku-bUku
karh-karh
kacil-kacil
wanita-wanita
miniman-minuman
'rights'
'books'
'monkeys'
'S~~~~-DISTRIBUTIVE'
'women'
'beverages'
Here, two phonologically identical forms (i.e., each segment in the reduplicant
has an obvious co-indexation to a segment in the base, h1&k3-h\dyk3 'rights') are
co-affixed to produce plural semantics.
(7)
Vedic intensive reduplication (data from Schaefer 1994, S.V. given roots)
w t
ddii 'point'
dnii 'bellow'
dvrt 'turn'
&J.3
mid. didi$e
mid. dkdiiate
act. nbnaviti
1.pl.act. nonumas
act. vdrvarti
act. vkrvrtati
In this case, the reduplicant (leftmost syllable) always shows the vocalism that
corresponds to the full grade of that root, while the root itself undergoes ablaut,
depending upon personlnumber and diathesis. Strictly speaking, where the root
stands in the zero grade (e.g., dgdiiate or nonumas), not every segment of the
reduplicant has a correspondent in the base: in Idlab-dli2S3-I,the a in the reduplicant lacks a corresponding segment in the base, while the i in the base lacks a
corresponding segment in the reduplicant. The morphophonological derivation
proceeds as follows:
(8)
Derivation of 3.sg.pres.mid.intens. didis-'
1. Root: /dais-/
2. Semantics: intensive
3. Morphology: reduplicate for intensive + IRED-do&/
4. Phonology: fill the reduplicant following restrictions (not more than one
coda consonant) + Idai-da&/
5. Semantics: middle diathesis
7
This derivation intends no theoretical claims whatsoever; it is merely illustrative.
236
Ryan Sandell
6. Morphology: ablaut to zero grade of the base, add desinence +
Id5i-dais-tag
3.2.3. Semantic doubling. Inkelas and Zoll (2005:6-8) farther classify certain
constructions in which no phonological relationship whatsoever obtains between
the constituents as reduplication:
(9)
Khmer synonym compunds (after Inkelas and Zoll 200.523; data from Orun
and Haiman:485-501)
Compound
cah-tum
kee-mordok
camaj-?ahaa(r)
Paarkambag
cbah-prakat
peel-weelia
Constituent semantics
'old+mature'
'heritage+heritage'
'food+food'
'secret+secret'
'exact+exact'
'time+time'
Compound semantics
'village elder'
'legacy'
'food'
'secret'
'exact'
'time'
Similar is a case from Vanatuan Sye. In Vanatuan Sye, every verbal root has two
phonologically distinct, but semantically identical, stem formations that cannot be
phonologically derived from one another. Historically, these stems derive from a
single verbal root:
(10)
Vanatuan Sye verbal allomorphs (data from Crowley 1998:83-5; see further
Inkelas and Zoll2005:524)
Stem 1
evsor
evtit
ocep
ochi
om01
oruc
ovoli
vag
Stem 2
amsor
avtit
agkep
aghi
am01
anduc
ampoli
amp%
'wake up'
'meet'
'fly'
'see it'
'fall'
'bathe'
'turn it'
'eat'
Which stem is used depends on the morphological category being expressed, e.g.,
the simple future takes Stem 2, whereas imperatives take Stem 1. Most interesting is a reduplicative construction that consists in the co-affixation of the two
stems:
(1 1)
cw-amolz-omoll
3p~Fu~-falL-fall1
'we will fall all over'
The Morphophonology of Reduplicated Presents
4. Modeling the Vedic reduplicated present
This section will attempt to formulate a synchronic analysis of the Vedic reduplicated present within two distinct approaches to reduplication: Base-Reduplicant
Correspondence Theory (McCarthy and Prince 1995) and Morphological Doubling Theory (Inkelas and Zoll 2005). Given the divergent theoretical claims
about the nature of reduplication that these two models make, the compatibility or
incompatibility of the Vedic evidence may serve to provide evidence in support
of or against one model or the other.
4.1. Base-Reduplieant Correspondence Theory and Morphological Doubling
Theory: Definitions
4 1 . 1 . Base-Reduplicant Correspondence Theoy. Under Base Reduplicant Correspondance Theory (BRCT), reduplication is a prosodic phonological process in
which an abstract RED(up1icant) morpheme receives its phonological specification and ultimate surface form through the interaction and ranking of four constraint relationships:
1. General I(nput)-O(utput) markedness constraints on acceptable surface
form.
2. I(nput)-R(edup1icant) faithfulness: the output reduplicant should be phonologically like the input base.
3. I(nput)-B(ase) faithfulness: the output base should be phonologically like
the input base.
4. B(ase)-R(edup1icant) identity: the output reduplicant should phonologically resemble the output base.
Figure 1 represents the interactions of these various relations:
OUTPUT
R [filled]-
B-R
B
Fig. 1. The Full Model of Reduplicative Identity (after McCarthy and Prince 1995:25)
238
Ryan Sandell
This model of reduplication depends on the theoretical claims and machinery of
parallel Optirnality Theory (oT).*
4.1.2. Morphological Doubling Theory. Morphological Doubling Theory (MDT)
defines reduplication in terms of morphology and semantics: the constituents of a
reduplicative construction are semantically identical, and the targets of such a
construction are morphological elements (such as a word, root, stem, or affix)
rather than phonological elements (cf. Inkelas and Zoll 2005:66). Thus, true reduplication, as opposed to phonological copying (cf. 93.1 above), consists in the
co-affixation of two semantically identical morphemes. The fact that semantically
identical morphemes are often phonologically similar, if not identical, results in
the phonological similarities seen between base and reduplicant in many reduplicative constructions cross-linguistically. The exact phonological shape that a reduplicative construction adopts is the result of interactions between three
cophonologies, is., distinct sets of phonological rules or constraints that apply to
the morphological constituents independently and then at the level of affixation
(cf. Inkelas and Zoll2005: 19):
Fig. 2. Interaction of cophonologies in reduplication
I will employ Optimality Theory to represent the operation of the various cophonologies in MDT, though MDT in itself is compatible with whatever
phonological framework one chooses to adopt.
4.2. Modeling under BRCT
Given that the data presented under $2 contain some patterns of reduplication in
which the reduplicant and base lack an obvious relationship of identity, BRCT
faces a significant challenge. The following examples will illustrate that one may
formulate constraint rankings that can capture either the reduplication pattern of
Ca- and CaC-roots (such as &a or dvac) that regularly show i-reduplication, or
the reduplication of roots that show a-reduplication (e.g., dbhas or dda), but not
8
See Prince and Smolensky 1993 for an introduction to Optimality Theory.
The Morphophonology of Reduplicated Presents
239
both. Either set of constraints can correctly predict the outcome for roots with
zero grades in i, u, or r.9
Since the primary concern of the following exposition lies solely in the quality of vocalism found in the reduplicant, whatever constraints necessary to produce the correct consonantism or prosodic shape will be omitted in order to
simplify the exposition. I therefore take for granted the following conditions on
the reduplicant for Sanskrit reduplicated presents:
The reduplicant admits only of palatal stops (e.g., cikisi).
All underlying aspirates surface as unaspirated [Grassmann's Law] (e.g.,
bdbhasti).
Where a root has more than one root-initial consonant, only the less sonorous of those consonants will appear in the reduplicant (e.g., tisthati).
Limitations of space here do not permit independent justification of the constraints in use.
4.2.1. Constraint ranking I: MAX-IB,IDENT-IB> *V[,hieh\3> IDENT-BR.In example (12), the reduplication of ^Ima 'bellow' from the full grade morphologically
required for the 3 .sg.pres.act.ind. surfaces as mimati, in winning candidate a:
for the following reasons:
The IDENT-BRviolations in (12a) result from a change in quality and
length of vocalism in the reduplicant; however, were the reduplicant
vowel to maintain the same quality and/or length, violations of higher> 1 or *V[_iiighl
would ensue; cf. candidates b and c.
ranking *q,
9
A stipulative constraint *R seems to be necessary to forbid an r in a reduplicant throughout,
though the same restriction was probably in operation for PIE itself; see further 55.
240
Ryan Sandell
*up> 1 and *Vr_~,igM
do not force any changes in the surface expression of
the base, because MAX-IB and IDENT-IB,demanding faithfulness to the
input, outrank those markedness constaints.
In example (13), the (e) (= [ail) vocalism of the base is impermissible in the reduplicant according to the given constraint ranking, because it would violate both
*ou> 1 and *V[-higM;cf. candidate b.
However, to resolve *up > 1 through the elimination of [i] still leaves a non-high
vowel in the reduplicant, as in candidate (13c). Winning candidate (13a), on the
other hand, violates low-ranking MAX-BR in deleting the [a] and violates IDENTBR in changing the underlying non-syllabic Ill to syllabic [i], but thus avoids any
violation of either *oF> 1 or *Vr-highl.
In order to satisfy both *q, > 1 or *V[-higill,as well as high-ranked *R, winning candidate (14a) shows deletion of all acceptable vocalic material originally
present in the base, into which an anaptyctic vowel [i] is then inserted.''
However, as examples (15) and (16) illustrate, this set of constraints predicts
that roots that show a-vocalism in the reduplicant would instead have i-vocalism.
Candidates (15a) and (16a) both pick out forms as winning candidates that do not
exist, rather than the expected candidates of (15b) and (16b); in particular, contrast example (16) with example (12).11 Therefore, this constraint ranking cannot
be viable.
10 The general spread of [i] as an intermorphemic linking vowel, on analogy to original setroots, could constitute a reasonable basis for assuming [i] as the standard anaptyctic vowel of
Sanskrit.
11 Kulikov (2005:438nl6) points out that a stem bibhas- is found in some manuscripts, but universally is emended to babhas-.
The Morphophonology of Reduplicated Presents
24 1
4.2.2. Constraint ranking 2: *oI,
> 1, MAX-BRL+hlgii1
>> IDENT-BR,MAX-BR. This
constraint ranking similarly employs *q,
> 1 to drive the moraic value of the reduplicant, but the removal of *V[-hiehiensures that forms such as bdbhasti and
ddddti can surface properly. MAX-BR[+~~~,,]
gives preference to high vowels in the
reduplicant, thereby securing forms such as cikksi over "cakesi.
Examples (17) and (1 8) are parallel to examples (15) and (16), demonstrating
that expected bdbhasti and ddddti can win using the new set of constraints:
Ryan Sandell
In (19), the faithfulness constraint MAx-BR[+~,~~~,~,
which establishes a bias for the
preservation of [+high] segments existing in the base over [-high] segments, allows cikki to surface:
(19)
However, the ultimate consequence of this set of constraints is that other roots
that show a type of a-vocalism in the base are predicted to show a-vocalism in
the reduplicant as well, contrary to the actual data; compare (20) with (12) above:
The Morphophonology of Reduplicated Presents
243
Again, the given constraint ranking is unable to provide a successful account of
all variation found in the Vedic data. Therefore, one must conclude that the theoretical machinery that BRCT offers is inadequate to explain the morphophonological patterns found in the Vedic reduplicated present.
4.3. Modeling under MDT
At least for the athematic reduplicated presents represented in table 1, MDT can
provide a coherent account: the formation of a reduplicated present consists in the
co-affixation of the zero-grade allomorph to a given root, as the reduplicant, to
another form of the same root, taking whichever ablaut grade the inflectional
paradigm specifies. The Vedic reduplicated present is akin to the reduplicative
constructions of Vanatuan Sye seen in example (1 1) in 53.2.3-two distinct allomorphs (in some cases, synchronically non-derivable allomorphs, as in zerograde forms with i-vocalism to Cii-roots) are bound to one another in order to
fulfill the need for two semantically identical elements, similar to the Khmer
synonym constructions in example (9). The derivations in examples (21)-(24)
illustrate the composition of a Vedic reduplicated present: cophonologies at the
appropriate daughter nodes shape the respective allomorphs for which the morphosyntax calls, and then further cycle through the mother cophonology to produce a final output.
As seen in the preceding section, the constraint *ou> 1, here present in the
Daughter A cophonology, drives the prosodic shape of the reduplicant. Likewise,
the constraints ranked above DEP-V result in a reduplicant of the form Ci- to
roots with zero grades in r. Throughout, I take the forms of the past passive participle shown in table 1 as indicative of the morphological form of the zero grade
to a given root.
244
Ryan Sandell
(21)
Morphosyntactic operation: construct the 3.sg.pres.act.impv. for 'measure'
1:;
mi-milmi-m-
l&l
Output: mi-ma-
Suffix: -tu
+
mimiitu
The constraint set needed to generate the correct output is very simple: Daughter
B (the base) and the Mother (the co-affixed stem) need only be strictly faithful to
the inputs; the reduplicant is largely faithful to the input as well, but its restricted
prosodic shape induces some change. The burden for correct outputs instead falls
almost entirely on the morphology, which selects the proper input form from
commands received from syntax or semantics.
Given that the zero grade allomorph to dyas is probably yas- as the past passive participle yastb- would suggest, then the a-vocalism found in the reduplicant
of yayastu (22) is simply a preservation of the underlying vocalism. Given that
the only free-standing form of the past passive participle to ^da 'give' attested
from the Rgveda forward is dattd-, or -data- or -tta- as the second member of
compounds, I submit that the zero grade allomorph that served as an input for
dd6 was not 'di-, as mi- is to drnii (21)' but rather dad- or da-, thus making the
vocalism of the reduplicant straightforward to obtain.12
12 Some special pleading, though, is necessary to explain dadhati to ddha, which attests only a
PPP hita-, or -dhita- as the second member of compounds and in sandhi. I see two possible
factors driving the formation of dadhati : (i) thoroughgoing identity of inflection to dda, and
(ii) hit& as a suppletive form for the PPP taken from a lower register of speech. If one or both
T h e Morphophonology of Reduplicated Presents
(22)
Morphosyntactic operation: construct the 3.sg.pres.act.impv. for 'boil'
(Â¥^yas
n
Output: ya-yasSuffix: -tu
+yayastu
The interaction of constraints in the Daughter A cophonology demonstrates the
outcome for roots with zero grades in r (23). As in $4.2, I assume that [i] is the
unmarked anaptyctic vowel. The validity of this assumption, however, gains further credence in that it helps to account for the formation of reduplicated presents
to roots that exhibit sarnprasGra~ain their zero grade, i.e., dvac, which has a PPP
uktd-. If one introduces a further constraint, ONSET,which penalizes the formation of syllables without an onset consonant, ranked above DEP-V, into the constraint set of the Daughter A cophonology, then the desired reduplicated present
form vivakti will result (24).13
of these conditions hold good, then perhaps a zero-grade allomorph "dadh- or "dha- to ^ d h ~
existed at some level of representation. One should also take note of the fact that data- is the
PPP to Avestan 4dG- (with conflation of Proto-Indo-Iranian 'dd2 'give' and 'ddhG 'set').
13 vivasi to dvai is likewise an expected outcome under the same account, if the weak stem of
the Class I1 present, e.g., 1.pl.pres.act.ind. u M s i , is taken as evidence of the proper zero
grade allomorph to d v a i
246
Ryan Sandell
(23)
Morphosyntactic operation: construct the 3.sg.pres.act.ind. for 'bear' (dbhr)
Output: bi-bharSuffix: -ti
+ bibharti
(24)
Morphosyntactic operation: construct the 3.sg.pres.act.ind. for 'speak'
/full grade allomorph
to 'soeaky/
to 'sneak' (= ucV
c.
ac-
1 *!
The Morphophonology of Reduplicated Presents
Output: vi-vac-
Suffix: -ti
+ vivakti
The presence of a constraint ONSETin the morphologically determined Daughter
A cophonology makes for an interesting contrast with Vedic perfects, which also
show reduplication: the 3.sg.perf.act.ind. of dvac is uviica, which has the reduplicant the reduplicated present would have produced without ONSET.This contrast,
then, perhaps provides evidence of intralinguistic morphologically conditioned
phonology.
4.4. Conclusions
Section 4.2 demonstrated that a model that defines reduplication in terms of
phonological correspondence between the constituent morphemes, such as
BRCT, cannot account for significant parts of the Vedic data. In contrast, MDT
allows one to describe the formation of the Vedic reduplicated present as an essentially morphological operation, subject to relatively minor alteration at the
phonological level. This outcome bolsters the thesis of Inkelas and Zoll (2005)
that reduplication is fundamentally morphological and semantic in nature, rather
than phonological.
Although the approach described in 54.3 captures the data of athematic reduplicated presents, certain thematic reduplicated presents, on the other hand, flatly
contradict this model, e.g., 3.sg.pres.mid.ind.jighnate to w a n , where the model
of 54.3 would predict "jaghnate. Given that the thematic reduplicated presents
are absolutely consistent in the form of the reduplicant regardless of the root
phonotactics, one may simply treat them as instantiations of fixed-segment reduplication, as discussed under 53.2.1. Thematic reduplicated presents reflect a
morphophonological process distinct from athematic reduplicated presents.
The question of historical origins nevertheless remains: how and why would
Vedic have developed the synchronic process that it employs for forming reduplicated presents, in the face of data from other Indo-European languages?
5. Reconstructing the Indo-European reduplicated present
This section will summarize arguments concerning earlier reconstructions of the
PIE reduplicated present, then finally make a new attempt, employing MDT to
model the reduplicative process.
248
Ryan Sandell
5.1. One or two reduplicatedpresents?
In the main, I am sympathetic to the intuitions of Frotscher and Hill (2010:2) that
"the assumption of just one stem formation is much more economical" and that
"mixed paradigms are better explained ... if i- and e-reduplication originally
belonged to one paradigm," e.g., sbakti vs. sdscati, jigiiti vs. jbgat-, or jahati
vs. jihite. Models that adopt two types of reduplicated present must appeal to
extra-paradigmatic analogy to explain such mixed paradigms, and also have a
more difficult burden of proof with respect to semantics.l4
The converse approach, namely, to derive all athematic reduplicated presents
in Vedic from a single Indo-European type, goes back at least to Hirt 1900:1903. More specifically, Hirt suggested that the consistent absence of full grade in
the desinence of the 3.pl.pres.act.ind. of reduplicated presents, e.g., bi-bhr-ati or
dd-d-ati, in contrast to the other Vedic athematic present types, which always
show full grade of the desinence in the S.pl.pres.act.ind., e.g., s-dnti or yu-fi-jdnti, points to original accentuation and full grade in the reduplicant; thus Vedic
dd-d-ati would continue PIE *d&-dhs-~ti.
Watkins 1969:36 makes a similar observation, though Watkins ultimately attributes the full grade of the reduplicant
alongside the zero grade of the ending in the 3.~1.to analogy with Vedic intensives of the form 3.~1.vdr-vrt-ati (cf. example (7) above).15 Kortlandt (1987:222)
makes essentially the same observation as Hirt, and Kulikov (2005:445-6) follows and expands upon Kortlandt.
Frotscher and Hill (2010) also aim for a single PIE type, but because the
Vedic accent is interpreted as absolutely indicative of a PIE fall grade, Vedic
forms like bibharti are derived from PIE *bh&bhelor-ti.That the reduplicated present took *o-grade of the root in its strong stem seems improbable, insofar as a
stem *bh&bhor-would then be very similar to the stem of the PIE perfect. Furthermore, the non-operation of Brugmann's Law in the 3.sg.pres.act. subjunctives
bi-bhar-a-t and ba-bhas-a-t, where an original *o-grade should have produced
'bi-bhiir-a-t, 'ba-bhiis-a-t, also speaks against the reconstruction of *o-grade in
the root!
For the stem to have been *bh&-bherwould be even stranger, given that
14 The semantic difference that Tichy (2006: 113-4) sees between bibharti and dddhati is probably more a feature of lexical semantics than of the reduplication itself.
15 Watkins' implication seems to be that fixed-segment *i-reduplication was original to the IndoEuropean reduplicated present, though a means of accounting for the thoroughgoing areduplication of dadati and dddhati is then lacking.
16 Admittedly, however, the data on the inflection of the active subjunctive singular of the reduplicated present is conflicting as to whether full grade or zero grade of the root was original.
The Morphophonology of Reduplicated Presents
249
PIE did not license the co-occurrence of two e-grades in a single word. Hence,
the approach of Hirt, Kortlandt, and Kulikov, in which *i-reduplication is an outcome of zero-grade ablaut, while *e-reduplication is the outcome of full-grade
ablaut, seems most promising.
5.2. Reconstruction andformalization in MDT
I reconstruct the paradigm of the PIE reduplicated present active indicative as
follows, using *bher-as a sample root:
Thus, one may identify a strong stem in the singular, a weak stem in the first
and second plural, and a weakest stem in the third plural and the participle.
The root accentuation of the singular is preserved in a few cases, e.g., bibhhrti
and iydrsi, as well as in all roots with zero grades in u . The l/2.pl., in
general, faithfully preserve the original accent and ablaut, whereas the singular
has shifted the accent onto the reduplicating syllable (so also Whitney 1889:243),
possibly to enhance contrast between the reduplicated present and the perfect.
Thus, 3.~1.s2cati and pres.part.act. sdscat-l8 preserve PIE *si-skw-gtiand *sis P - ~ t - respectively,'9
,
while jdgat-20 preserves PIE *g%-gwh2-pt-.Note also the
Frotscher and Hill (2010:3) try to dismiss the accentuation of bibharti (RV 4.50.7, AVS
19.26.1, 2) and i y h i (RV 10.37.4), but both forms are lectiones difficiliores, given that the
accent throughout the paradigms of those stems is otherwise on the reduplicant. To attribute
the accentuation to perfect-derived forms like cikksi seems difficult, since such forms are relatively rare (bibhkti is not yet present in RV or AV).
Jasanoff (2003: 129) contends that 3.~1.sdscati is a nonce formation, based on the "substantivized participle sascat- 'pursuer'." That the substantivization of the participle occurred is true
enough, but a certainly participial usage, datsg. sdscate, occurs at RV 2.16.4b. Jasanoff does
not account for why saicat- itself should have a-vocalism in the reduplicant. To simply take
saicati at face value seems more straightforward.
That the */c" of *sk-sku-ptiand * sk-sP-pt- seems to show a palatal outcome (Skt. c) in a nonpalatalizing environment is unproblematic, because Indo-European root-final (labio-)velars
appear as palatals in Skt. when preceding a vowel, regardless of the provenance of that vowel.
Put another way, the underlying form of the IE root *seP- has become /sac/ in Skt., so 3.sg.
skakti is the result of internal sandhi from Isisactil. Compare the RVic paradigm of the Class
VII present to dprc 'mix': 2.sg.act.ind. p ~ & i , 3.pl.act.ind. prficanti, 3.sg.act.impv. pmaktu.
An informal count of roots from Whitney 1885 finds seven roots in final k versus twenty-two
250
Ryan Sandell
paradigmatic split of the PIE root *ihehl-into Vedic active jahati (cf. PPP htina-)
versus middlejihite (cf. PPP hind-). Furthermore, reduplicated media tantum presents that have been reinterpreted in Vedic as root presents, such as h e < *h$h&d-t6i, are precisely in line with the inflection of the weak stem found in the
first and second plural of the active.
If one accepts a PIE paradigm containing both *i- and *&vocalism in the reduplicant, then the *i-vocalism falls out as a representation of the zero-grade
form of the root; more specifically, this *i-vocalism may well reflect an anaptyctic vowel, inserted into the reduplicant at the level of Mother node phonology
in order to prevent a geminate consonant from surfacing, following the view that
PIE did not allow geminate consonants (26).
(26)
Morphosyntactic operation: construct the 3.sg.pres.act.ind. for 'bear'
Output: * bhi-bh&Suffix: *-ti
+ *bhibh6rti
-
-
roots in final c; most of the former seem to be late introductions into the Classical language,
while the latter are largely of Indo-European origin.
20 jagat- is, however, fossilized as a noun, 'moving world', and has been replaced in Vedic with
jigat-, in accordance with the rules of derivation given in 94.3.
I
I
The Morphophonology of Reduplicated Presents
25 1
For the PIE construction, the inputs are taken to be the full-grade form of the
root. In accordance with the common view that *e-grade in PIE bore the word
accent, the constraint *E[_stressl
prohibits an unaccented *e from surfacing; for my
purposes here, I assume that the accent is assigned at a level of derivation preceding what is shown in (26) and (27). Thus, the reduplicant takes zero grade, but *R
disallows a syllabic resonant, so this potential syllable nucleus is r e m ~ v e d . ~The
'
anaptyctic [i] appears at the level of the Mother cophonology in order to satisfy
both *GEMINATE
and MAX-10, given that DEP-V is ranked lower than those two
constraints. Limitations of space here do not permit further comment on the nature of anaptyctic vowels in PIE, but the reader may find helpful discussion in
Vine 1999 and Byrd 2010: 108. The corresponding weakest stem, in the 3.pl., easily falls out from these constraints as well:
(27)
Morphosyntactic operation: construct the 3.pl.pres.act.ind.for 'bear'
(*bher-)
Output: *bh&bhrSuffix: *-nti
+ *bh6bhrpti
21
*F$ seems well motivated insofar as no IE language attests to a syllabic resonant or a clear
vocalizedreflex thereof in a reduplicant. For more on special phonological properties limited
to reduplicants, see Inkelas and Zoll2005:ch.3.
252
Ryan Sandell
5.3. From PIE to Vedic
The PIE athematic reduplicated present derivations seen in (26) and (27) furnish
morphologically identical inputs as the daughters of the reduplicative construction, whereas Vedic, as demonstrated in $4.3, employs morphologically distinct
inputs. The final question, then, is: what drove this basic change in the underlying
morphological constituency of the reduplicated present between PIE and Vedic?
As per Kulikov 2005:445-6, the crucial step appears to lie with laryngeal-final
roots (Vedic Cd-roots): following the vocalization of the laryngeals to Vedic Iil
when such roots were in the zero grade, speakers began to associate the inherited
i-vocalism in the reduplicant with the zero-grade allomorph of the root. Consequently, the zero-grade vocalism spread throughout the individual paradigms,
thus leading to u-reduplication in roots with zero grades in u (such as juhhti) and
a-reduplication to roots in which a phonological full grade had come to fill the
morphological zero grade, as in roots ending in coronals (such as b d b h a ~ t i ) ddii
.~~
and ddhii, meanwhile, may have developed a paradigm with a-reduplication due
to early loss and non-vocalization of the root-final laryngeal; contrast 1.pl.pres.
act.ind. Jiiimasi with dadmas or 3.sg.pres.mid.ind. Mite with dattd. ddii and
ddhd thereby escaped the pattern of other Cii-roots, but ultimately fell victim to
the larger pattern that the Cd-roots promoted (cf. n12 above).
References
Albright, Adam. 2002. Islands of Reliability for Regular Morphology: Evidence from
Italian. Language 78.684-709.
Alderete, John, Jill Beckman, Laura Benua, Amalia Gnanadesikan, John McCarthy, and
Suzanne Urbanczyk. 1997. Reduplication with Fixed Segmentism. http://roa.rutgers
.edu/files/226-1O97/226-1097-ALDERETE-0-O.PDF (1 1 February 20 10).
Amano, Kyoto. 2009. MaitrayamSamhita 1-11. Bremen: Hempen.
Bates, Dawn, and Barry F. Carlson. 1997. Spokane (Npoqinikn) Syllable Structure and
Reduplication. In Ewa Czaykowska-Higgins and Dale Kinkade (eds.), Salish Languages and Linguistics: Theoretical Descriptive Perspectives (Trends in Linguistics:
Studies and Monographs 107), 99-1 23. Berlin: Mouton de Gruyter.
Black, Deidre. 1996. The Morphological and Phonological Structure of Spokane Lexemes. Ph.D. diss., University of Montana.
22 The constraints and derivational process given in figures (26) and (27) would in fact predict
that PIE roots with zero-grade u- or i-vocalism would have such vocalism in the reduplicant. If
one insists that PIE athematic presents with zero grade in u never exhibited u-vocalism in the
reduplicant (note Hittite kukuizi < *geys-),then (26)and (27)require minor modification.
The Morphophonology of Reduplicated Presents
253
Byrd, Andrew. 2010. Reconstructing Indo-European Syllabification. Ph.D. diss., University of California, Los Angeles.
Cardona, George. 1992. On the Development of Presents like bibhkti. In Ravindranatha
N. Srivastava (ed.), Language and Text: Studies in Honour of Ashok R. Kelkar, 1-13.
Delhi: Kalinga.
Cohn, Abigail. 1989. Stress in Indonesian and Bracketing Paradoxes. Natural Language
and Linguistic Theory 27.167-2 16.
Crowley, Terry. 1998. An Erromangan (Sye) Grammar (Oceanic Linguistics Special Publication 27). Honolulu: University of Hawai'i Press.
E WA = Mayrhofer, Manfred. 1986-2001. Etymologisches Worterbuch des Altindoarischen. Heidelberg: Winter.
Frotscher, Michael, and Eugen Hill. 2010. The Accentuation of Oind. 3rd Class (Reduplicated) Presents. Paper presented at the Arbeitstagung der Indogermanischen Gesellschaft, 13-1 5 September 2010, Los Angeles.
Hirt, Hermann Alfred. 1900. Der indogermanisc
t: Vornehmlich in seinem
Verhaltnis zur Betonung. Strassburg: Trubner.
Hurch, Bernhard. 2005. Graz Reduplication Database. http://reduplication.uni-graz.at/(2
February 201 1).
Inkelas, Sharon, and Cheryl Zoll. 2005. Reduplication: Doubling in Morphology.
Cambridge: Cambridge University Press.
Insler, Stanley. 1972. Vedic mamatsi, dmamadur, and Gate. Zeitschrift fur vergleichende
Sprachforschung 86.93-103.
Jasanoff, Jay H. 2003. Hittite and the Indo-European Verb. Oxford: Oxford University
Press.
Joachim, Ulrike. 1978. Mehrfachprasentien im &Veda.Frankfurt am Main: Peter Lang.
Kortlandt, Frederik. 1987. Archaic Ablaut Patterns in the Vedic Verb. In George Cardona
and Norman H. Zide (eds.), Festschrift for Henry Hoenigswald: On the Occasion of
His Seventieth Birthday, 216-23. Tubingen: Nan.
Kulikov, Leonid. 2005. Reduplication in the Vedic Verb: Inheritance, Analogy, and Iconicity. In Bernhard Hurch (ed.), Studies on Reduplication, 431-54. Berlin: Mouton
de Gruyter.
Kummel, Martin Joachim. 2000. Das Perfekt im Indoiranischen. Wiesbaden: Reichert.
LIV= Rix, Helmut. 2001. Lexikon der indogermanischen verben. Wiesbaden: Reichert.
Lubotsky, Alexander. 1998. A &Vedic Word Concordance. New Haven: American Oriental Society.
Macdonell, Arthur Anthony. 1916 [1993]. A Vedic Grammarfor Students. Reprint: Delhi:
Motilal Banarsidass.
Marantz, Alec. 1982. Re Reduplication. Linguistic Inquiry 11.487-509.
McCarth~,John, and Alan Prince. 1995. Faithfulness and Reduplicative Identity. In Jill
Beckman and Susan Urbanczyk (eds.), Papers in Optimality Theory (University of
254
Ryan Sandell
Massachusetts Occasional Papers in Linguistics IS), 249-384. Amherst: Graduate
Linguistics Student Association, University of Massachusetts at Amherst.
Narten, Johanna. 1972. jcigat- im Rigveda. In Jacob Ensink and Hans Peter Gaeffke
(eds.), India maior: Congratulatory Voluine Presented to J. Gonda, 161-6. Leiden:
Brill.
Orun, Noeurng, and John Haiman. 2000. Symmetrical Compounds in Khmer. Studies in
Language 24.484-5 14.
Prince, Alan, and Paul Smolensky. 1993. Optimality Theory: Constraint Interaction
in Generative Grammar (Cognitive Science Center Tech Report TR-2). New
Brunswick, NJ: Rutgers Center for Cognitive Science.
Raimy, Eric. 2000. The Phonolosy and Morphology of Reduplication. Berlin: Mouton de
Gruyter.
Schaefer, Christiane. 1994. Das Intensivum im Vedischen (Historische Sprachforschung
Erganzungsheft 37). Gottingen: Vandenhoeck and Ruprecht.
von Schroeder, Leopold. 1881. Maitriiya@amhitci. Leipzig: Brockhaus.
Tichy, Eva. 2006. A Survey of Proto-Indo-European. Bremen: Hempen.
Vine, Brent. 1999. Greek pica 'root' and 'schwa secundum'. In Vjaeeslav V. Ivanov and
Brent Vine (eds.), UCLA Indo-European Studies I, 5-30. Los Angeles: University of
California, Program in Indo-European Studies.
Vogelin, Charles F. 1958. Working Dictionary of Tiibatulabal. International Journal of
American Linguistics 24.221-8.
Watkins, Calvert. 1969. Indogermanische Grammatik 11111: Geschichte der indogermanischen Verbaljlexion. Heidelberg: Winter.
. 2003. Hittite ku-ku-us-zi, KUB 10.99 i 29. In Gary Beckman, Richard Beal, and
Gregory McMahon (eds.), Hittite Studies in Honor of Harry A. Hoffner Jr.: On the
Occasion of His 65th Birthday, 390-2. Winona Lake: Eisenbrauns.
Whitney, William Dwight. 1881. Index Verborum to the Published Text of the AtharvaVeda. Journal of the American Oriental Society 12.1-383.
. 1885 [1963]. The Roots, Verb-Forms, and Primary Derivatives of the Sanskrit
Language. Reprint: Delhi: Motilal Banarsidass.
. 1889 [1960]. Sanskrit Grammar. Reprint: Cambridge: Harvard University Press.
. 1905. Atharva-Veda SamhitG: Translated with a Critical and Exegetical Commentary. Cambridge: Harvard University Press.
Wilbur, Ronnie Bring. 1973. The Phonology of Reduplication. Ph.D. diss., University of
Illinois, Urbana-Champaign.