1
This pdf of your paper in Puspika Volume 4 belongs to the publishers Oxbow Books and
it is their copyright.
As author you are licenced to make up to 50 offprints from it, but beyond that you may
not publish it on the World Wide Web until three years from publication (August 2020),
unless the site is a limited access intranet (password protected). If you have queries about
this please contact the editorial department at Oxbow Books (editorial@oxbowbooks.com).
A puṣpikā (‘little flower’) is the scribes’ way of marking the end of the main text and
the beginning of the colophon. The present logo is an artistic impression by Shubhani
Sarkar based on such a scribal flourish seen on a Nepalese manuscript.
an offprınt from
P uṣpıkā
Tracing Ancient India,
through Texts and Traditions
Contributions to Current Research in Indology
Volume 4
proceedıngs
of the
seventh ınternatıonal ındology graduate research symposıum
(leıden, 2015)
Edited by
Lucas den Boer
Daniele Cuneo
Paperback Edition: ISBN 978-1-78570-756-8
Digital Edition: ISBN 978-1-78570-757-5 (epub)
© Oxbow Books 2017
Oxford & Philadelphia
www.oxbowbooks.com
Published in the United Kingdom in 2017 by
OXBOW BOOKS
The Old Music Hall, 106–108 Cowley Road, Oxford OX4 1JE
and in the United States by
OXBOW BOOKS
1950 Lawrence Road, Havertown, PA 19083
© Oxbow Books and the individual contributors 2017
Paperback Edition: ISBN 978-1-78570-756-8
Digital Edition: ISBN 978-1-78570-757-5 (epub)
A CIP record for this book is available from the British Library
All rights reserved. No part of this book may be reproduced or transmitted in any form or by
any means, electronic or mechanical including photocopying, recording or by any information
storage and retrieval system, without permission from the publisher in writing.
For a complete list of Oxbow titles, please contact:
UNITED KINGDOM
Oxbow Books
Telephone (01865) 241249, Fax (01865) 794449
9146
Email: oxbow@oxbowbooks.com
www.oxbowbooks.com
Oxbow Books is part of the Casemate Group
UNITED STATES OF AMERICA
Oxbow Books
Telephone (800) 791-9354, Fax (610) 853Email: queries@casemateacademic.com
www.casemateacademic.com/oxbow
Contents
Preface
Peter C. Bisschop
150 Years of Sanskrit Studies in the Netherlands: The Karṇapurāṇa
vii
1
Dániel Balogh
The Abominable Yati: Or, An Intriguing Relic of the Prehistory of Sanskrit Verse
14
Małgorzata Sulich-Cowley
What do Sanskrit Adpositions Really Do and What Do They Mean?
The Analysis of Prati
33
Martina Palladino
Welcome with Open Arms: Iranian Loanwords in the Purāṇic Lexicon
51
Patrick McCartney
Speaking of the Little Traditions Agency and Imposition in ‘Sanskrit-Speaking’
Villages in North India
62
Judith Unterdörfler
Nature and Character Emotions in the Śrīgovindavilāsamahākāvya, Sarga 1
88
Lidia Szczepanik-Wojtczak
How to Teach Sanskrit Grammar: The Case of the Perfect System in
the Bhaṭṭikāvya
105
Prakash Venkatesan
Vaṇṇam: Tracing an Ancient Tamil Music Tradition
119
Katarzyna Skiba
Kathak as a Śāstrīya Nṛtya: The Rediscovery of the Nāṭyaśāstra and the
Invention of Classicism in Indian Dance
132
Jooyoung Lim
Charting the Sāṃkhya Commentarial Literature: A Study of the Lists of
Types of Supernatural Power (aiśvarya)
155
Lucas den Boer
Guṇaratna’s Refutation of the Cārvāka Perspective on the Soul: An Argumentative
Analysis of Tarkarahasyadīpikā 49.98–134
172
J. M. A. Eijsermans
The Benign Overlord and the Restorer of Dharma: Two Cases of Viṣṇu as
Political Expression in Ancient Khmer Inscriptions and Images
200
Elizabeth A. Cecil
Power and Piety in ‘Emplaced Polities’: Temple Patronage and Donative Practice
Under the North Konkan Śilāhāras (Ninth to Twelfth Century CE)
213
e Abominable Yati
Or, An Intriguing Relic of the Prehistory
of Sanskrit Verse
Dániel Balogh
Classical Sanskrit metrical verse paerns prescribe obligatory positions
for caesurae (yati). e testimony of preserved texts shows that poets
painstakingly observed these restrictions most of the time, yet in a number
of instances the rules seem to have been interpreted with some laxity. is
paper demonstrates, through a study of the use of yati in some samples of
Sanskrit poetry from dramas and epigraphic texts, that a large proportion
of these ‘abominable yatis’ actually follow rules of their own. Namely, a
caesura masked by vowel sandhi is neither an error (yatibhraṃśa), nor a
form of licence which poets resort to in dire straits or for literary effect.
Rather, such yatis consistently occur only at specific metrical positions,
and must thus be occasioned by their metrical context. It also seems that
the (regreably understudied) traditional recitation of Sanskrit verse oen
involves a prolonged or slurred syllable at these very same positions, which
allows the performer to restore hiatus in the text. ese two phenomena of
the musical and textual tradition may both be linked to an underlying cause.
e reason for the presence of unusual caesurae in poetical praxis and of
unusual articulation in performance may be sought in the abstract metrical
schemata underlying the fixed metres of classical Sanskrit verse: both occur
in metrical positions where the foot-based schema for the syllable-based
classical metre involves catalectic (incomplete) or syncopated (off-beat) feet.
The Caesura in Sanskrit Metrics*
e term ‘caesura’ in this paper refers to a point that separates one metrical segment
(colon) from another within a periodic unit (line) of verse. e Sanskrit term for
* e research behind this paper is linked to my doctoral thesis on Viśākhadaa’s play the
Mudrārākṣasa, supported in part by the Rabindranath Tagore Research Fellowship awarded by the
e Abominable Yati
what in English I call a caesura is yati, ‘restraint’. eoretically, the caesura is
expected to fall at a boundary between two words, allowing the reciter of a poem
to realise it as an audible (śravya) pause (virāma); another common term for the
yati is viccheda, ‘split’.1 ese synonymous terms are employed in Sanskrit poetics
not only for in-line caesurae, but also for the ends of verse quarters (pādas), which
thus seem to be regarded as qualitatively non-different from the former. Even if
this is so, there is a perceivable quantitative difference between the two, of which
the theoreticians are also aware. A brief text called the Yatyupadeśopaniṣad (cited
in its entirety by several commentators on poetical śāstras2) begins with a list of
the positions where a yati can occur. is treats the ends of quarters separately
from in-line caesurae and emphasises that a yati is especially required at the ends
of half-verses.3 e distinctiveness of the half-verse yati is traditionally said to
consist of the non-application of saṃdhi over this boundary and the prohibition of
compounds crossing it.4 Correspondingly, in the praxis of Sanskrit poets, the yatis
delimiting verse quarters (which I shall call line breaks hereaer) indeed appear to
be more marked than those separating one colon of a line from another (to which I
refer hereaer as caesurae).5 In fact, there seems to be at least one practical feature
in addition to the absence of saṃdhi and the scarcity of compounds stretching
over these boundaries. e established metres always call for a heavy syllable at
Indian Council of Cultural Relations. e topic of the ‘abominable’ yati was omied from the dissertation, but I have published about it in Hungarian (B 2015a). My research on epigraphic
texts was done under the aegis of the Asia Beyond Boundaries project (hps://asiabeyondboundaries.
org/), supported by the European Research Council. roughout this paper, Sanskrit is transliterated
according to IAST conventions. I use hyphens and spaces in Sanskrit text to separate compound
members and independent words (respectively), unless the boundaries are fused in vowel saṃdhi. For
prosodic notation I use the standard symbols ⏑ to represent light syllables and – to represent heavy
syllables. e sign ⏓ stands for an unregulated (anceps) syllable, while ○ indicates an unfilled metrical
position that may be realised as a pause in recitation. A colon (:) represents the caesura in metrical
formulae.
1 us Hemacandra’s Chando’nuśāsana 1.15 (V 1949: ९४): śravyo virāmo yatiḥ; and Halāyudha’s commentary on the Chandaḥsūtra of Piṅgala 6.1 (K̄̄ 1938: 100): vicchidyate vibhajyate pada-pāṭho ’sminn iti vicchedo viśrāma-sthānaṃ, sa ca yatir ity ucyate.
2 e text of the Yatyupadeśopaniṣad (a mere four ślokas) is cited (from Halāyudha’s commentary on the Chandaḥsūtra) and discussed in W 1863: 462–466. For several commentaries (on the
Vṛaratnākara of Kedārabhaṭṭa) citing and expounding this text (or a variant of it), see S et al.
1969: 23–31. For an in-depth discussion, see P 1977: 20–36.
3 Yatyupadeśopaniṣad 1a–c, yatiḥ sarvatra pādānte ślokārdhe tu viśeṣataḥ / samudrādi-padānte ca …
Here samudrādi refers to the coded numbers (bhūtasaṃkhyā) by which traditional metrical definitions identify the length of each colon of verse. For instance, samudra is a code for 4, since there are
held to be four oceans.
4 us, for example, Halāyudha on Chandaḥsūtra 6.1 (K̄̄ 1938: 101) says, ślokārdhe tu
viśeṣataḥ ity atra saṃdhi-kāryābhāvaḥ spaṣṭa-vibhaktitvaṃ ca viśeṣaḥ.
5 P (1977: 37) advises against using the word caesura in the context of Indic poetry, and
employs ‘pause’ for the ends of lines and ‘break’ for in-line yatis. While I accept his reasoning that
this feature is not identical to the caesura of classical European poetry, I see no problem in applying
familiar terminology in a slightly modified sense.
the ends of lines, but not infrequently such line-final syllables are actually light,
yet counted as heavy by a rule corresponding to the tenet of brevis in longo. is
applies in particular to the ends of the even quarters (i.e. to the half-verse break
and the end of the stanza); short syllables at the ends of odd quarters appear to be
rare in the longer lyrical metres, though not uncommon in some popular shorter
verse forms.6 Additionally, brevis in longo is to my knowledge never invoked before
(in-line) caesurae: where a metre prescribes a heavy syllable at the end of a colon
(which is almost always the case7), that syllable is invariably heavy by nature or
by position.
‘Abominable’ Ya s
e lack of a word boundary at a caesura is regarded by Sanskrit theoreticians as a
serious defect in poetry (yatibhraṃśa).8 But even such a lack may come in several
flavours. To illustrate diverse cases of yati transgressions, I reproduce here two
stanzas in the pṛthvī metre. e formula for this metre is ⏑ – ⏑ ⏑ ⏑ – ⏑ – : ⏑ ⏑ ⏑ –
⏑ – – ⏑ ⏓, with a caesura prescribed aer the eighth syllable in several traditional
treatises on Sanskrit metrics,9 yet there are numerous instances in actual poetry
where this prescription is ignored. us, a notorious example from the Nītiśataka
of Bhartṛhari10 goes,
labheta sikatāsu tai:lam api yatnataḥ pīḍayan
pibec ca mṛga-tṛṣṇikā:su salilaṃ pipāsārditaḥ /
kadācid api paryaṭañ: śaśa-viṣāṇam āsādayen
na tu pratiniviṣṭa-mū:rkha-jana-ciam ārādhayet //
6 In the Mudrārākṣasa, Viśākhadaa only makes use of brevis in longo at the ends of odd pādas in
the metres anuṣṭubh, upajāti and vasantatilaka, in which he avails of this licence 16, 2 and 14 times
(respectively), i.e. in 37, 25 and 37 per cent (respectively) of the total number of odd pādas in these
metres.
7 A light syllable before a caesura occurs in only three Sanskrit metres, each very rare (P
1977: 99).
8 us Daṇḍin in Kāvyādarśa 3.125–126 (D 2011: 396) says this should be avoided:
… yati-bhraṣṭaṃ … iti doṣā … varjyāḥ kāvyeṣu sūribhiḥ. Later on (3.152; D 2011: 424) he
explains that yatibhraṃśa is jarring to the ear, śravaṇodvejanaṃ. It is possible that some of the early
theoreticians, including Bharata, regarded the yati as optional, but others, including Piṅgala, viewed
it as compulsory and this laer view gained the upper hand in the long run (V 1949: 19).
9 Such as the Chandaḥsūtra of Piṅgala 7.17 (K̄̄ 1938: 158): pṛthvī jsau jsau yalau g
vasu-navakau.
10 Number 319 in K 1948. e stanza is also cited as an illustration of the non-observance of
yati in pṛthvī by P 1977: 81 and G 1978: 631.
e Abominable Yati
Although the third pāda of this stanza seems to observe the prescribed caesura,
this is probably mere happenstance; in quarter b the theoretical break falls
between a nominal stem and a declensional ending, while in a and d it is smack
in the middle of a morphologically indivisible word.11 It is thus fairly certain that
whoever composed this particular stanza followed a school of versification that did
not stipulate a (compulsory) caesura here. G (1978) presents an overview of
śāstras that do or do not require this caesura and discusses specimens of classical
Sanskrit poetry that refrain from observing it. He concludes that several great poets
including Kālidāsa occasionally break the caesura in the pṛthvī metre, even though
they hold to it most of the time in this metre and observe caesurae fastidiously in
other metres.
If a poet generally does use a caesura at a particular point, yet sometimes omits
it, the rationale for this is most probably a conscious transgression to enhance the
poetic effect of the composition. In contrast to the downright lack of caesurae in
the above example, in verse 9 the Allahabad praśasti of Samudragupta12 we read,
pradāna-bhuja-vikkrama-:praśama-śāstra-vākyodayair
uparyyupari-sañcayo:cchritam aneka-mārggaṃ yaśaḥ /
punāti bhuvana-trayaṃ: paśupater jjaṭāntar-guhānirodha-parimokṣa-śī:ghram iva pāṇḍu gāṅgaṃ payaḥ //
Here the caesurae in pādas a, b and c are more or less in good order (if any appear
otherwise, please bear with me a short while: I shall return to them). In pāda d,
however, the break is conspicuous by its absence. Expected to fall in the middle of
the word śī:ghram, ‘swi’, the omission of a caesura here puts extra emphasis on
the rush of ‘the pale Gangetic water swi upon its release from confinement in the
cavities within Śiva’s dreadlocks’. I see this instance as a premeditated transgression
by the poet Hariṣeṇa,13 enhanced further by the torrential enjambment of the
compound jaṭāntar-guhā-nirodha-parimokṣa-śīghram. What is more, the yatis
in the first three quarters are made even more prominent by initial alliteration14
(clearly manifest in pradāna … praśama and punāti … paśupater, and obscured by
11 e Yatyupadeśopaniṣad (2) actually permits puing caesurae inside a word, provided that both of
the segments are longer than one syllable: kvacit tu pada-madhye ‘pi samudrādau yatir bhavet / yadi
pūrvāparau bhāgau na syātām eka-varṇakau //. is sort of mid-word yati seems to be extremely rare
in poetic praxis, and in any case the pṛthvī stanza cited here stands in violation even of this licence.
12 F 1888: 9, line 30–31.
13 Contrary to my opinion, P (1977: 83) mentions this verse as an example of the noncompulsory nature of the caesura in pṛthvī.
14 My thanks to Stephen Durnford for pointing out this feature of the stanza, which I had overlooked, and to Jean-Luc Chevillard for noting that such word-initial ‘rhyming’ (mōṉai) is regularly
used in Tamil poetry to link words within a colon.
vowel saṃdhi in uparyyupari … ucchritam), and this emphasis is also conspicuous
by its absence in the fourth pāda.
So far I have discussed cases where the outright absence of a word boundary
at the yati can be ascribed to a tradition which does not put stock in a caesura at
that position, and cases where a poet deliberately chooses to ignore the caesura to
heighten the effect of his poetry. However, a fair number of cases are less than clearly
transgressive, since the concept of a word boundary is somewhat fuzzy in Sanskrit.
First, there is a relatively straightforward maer: the issue of nominal compounds
(e.g. vikkrama-:praśama in pāda a of the above stanza). e Yatyupadeśopaniṣad
makes it clear that a word boundary without a manifest declensional case ending
(i.e. a non-ultimate member of a compound) is just as acceptable at a caesura as one
with a case ending.15 is is generally confirmed by the praxis of poets, who have
no qualms about puing their caesurae at the boundaries of compounded words.16
Less than straightforward is the maer of word boundaries (inside or outside
of compounds) obscured by vowel saṃdhi, such as sañcayo:cchritam in the above
example. is can come about in one of two ways. On the one hand, word-final and
word-initial vowels may merge in a single long vowel or diphthong; on the other
hand, word-final ī̆, ū̆ and ṛ change into the approximants y, v and r when followed
by a word-initial vowel. e Kāvyālaṅkārasūtravṛi of Vāmana unconditionally
allows the breaking of words on account of vowel saṃdhi.17 Contrarily, the
Kāvyālaṃkāra of Bhāmaha, though it provides no explicit rule for vowel saṃdhi,
actually illustrates the fault of yatibhraṃśa with such an example (and no other).18
e Yatyupadeśopaniṣad is distinctly aware of this twilight zone of caesurae and
implicitly accepts it as legitimate, stipulating that a vowel joined in saṃdhi is
generally to be regarded as the end of the pre-caesura word (pūrvāntavat), and only
15 Yatyupadeśopaniṣad 1d, vyaktāvyakta-vibhaktike. Halāyudha (K̄̄ 1938: 101) explains:
tatra śrūyamāṇa-vibhakty-antaṃ vyakta-vibhaktikam, samāsāntarbhūta-vibhakty-antam avyakta-vibhaktikam.
16 is is basically a subjective impression, open to refutation on an empirical basis. In the way of
statistical figures, I can say for certain that of the 500 caesurae in the verses of the Mudrārākṣasa, 42
(i.e. 8%) coincide with compound boundaries, in contrast to 0 caesurae within words. Further details
are available in B 2015a: 20–21. Dieter Gunkel (personal communication, September and
October 2015) has expressed doubts about compound boundaries at yati points being equally acceptable to independent word boundaries. A cursory investigation of the data available to me shows that
in-compound yatis may indeed be less frequent than one might expect on the basis of the general
ratio of in-compound boundaries to independent word boundaries. Poets may thus have felt some
degree of aversion to puing a compound boundary at their yatis, but nonetheless did so frequently
enough for the author of the Yatyupadeśopaniṣad to allow this practice explicitly. e maer deserves
further analysis, but for my present purposes I feel we can safely deem in-compound yatis entirely
non-abominable.
17 Sūtra 2.2.4 (J 1971: 62): tad [i.e. yatibhraṣṭam] dhātu-nāma-bhāga-bhede svara-saṃdhy-akṛte
prāyeṇa; vṛi ad loc.: svara-sandhy-akṛta iti vacanāt svara-sandhi-kṛte bhede na doṣaḥ.
18 Kāvyālaṃkāra 4.24–25 (S 1956: 169): yatiś chando-’dhirūḍhānāṃ śabdānāṃ yā vicāraṇā/tadapetaṃ yati-bhraṣṭam iti nirdiśyate yathā // vidyutvantas tamālā:sita-vapuṣa ime vāri-vāhā dhvananti.
e Abominable Yati
occasionally (kvacid eva) as the beginning of the post-caesura word (parādivat),
while an approximant is to be regarded as the beginning of the laer word.19
ese instructions, however, are rather succinct and imprecise. e words
pūrvāntavat and parādivat are evidently used to distinguish between cases of vowels
merged in saṃdhi where the caesura falls, respectively, aer the merged vowel and
those where it falls before.20 But neither the Upaniṣad nor the commentaries I have
studied say anything about the circumstances in which the parādivat variant is
permied, even though the expression kvacid eva implies specific cases rather than
a random rareness. e statement that approximants are (always) parādivat seems
to be a technicality based on the concept of the wrien akṣara, which dictates that
a consonant (including the approximants produced in saṃdhi) joins the following
vowel.21 All in all, the upaniṣad definitely ratifies yatis in which the word break is
obscured by saṃdhi, and seems to say that some of these should be employed only
in special cases.
Case in Point: Anomalous Caesurae in the Mudrārākṣasa
Curious to see how an actual poet handles his caesurae, I examined the verses of the
Mudrārākṣasa of Viśākhadaa from this viewpoint. e total number of theoretical
yatis in the play happens to be exactly 500.22 Not a single one of these occurs inside
a morpheme (as in śī:ghram above), nor are there any instances of caesurae between
a prefix and a stem (for which see later). ere are a fair number (42, to be precise)
of caesurae between compounded nouns (as in vikkrama-:praśama above).23 e
most interesting finding, however, is that caesurae obscured by vowel saṃdhi (as
in sañcayo:cchritam above), though uncommon, appear in numbers that cannot,
19 Yatyupadeśopaniṣad 3, pūrvāntavat svaraḥ saṃdhau kvacid eva parādivat / draṣṭavyo yaticintāyāṃ yaṇ-ādeśaḥ parādivat //. e expression yaṇ-ādeśaḥ refers to Aṣṭādhyāyī 6.1.77 (iko yaṇ aci),
i.e. the formation of approximants from final vowels in saṃdhi.
20 As illustrated, for example, by Halāyudha commenting on Piṅgala 6.1 (K̄̄ 1938: 102),
citing asthāno:pagatayamunā (Meghadūta 1.54) for pūrvāntavad-bhāva and mahiṣa:syāhito (untraced)
for parādivad-bhāva.
21 As explained more specifically in what seems to be an alternative version of the Yatyupadeśopaniṣad in Trivikrama’s commentary on Vṛaratnākara 1.11 (S et al. 1969: 24), which does not
mention yaṇādi as a separate case but says that a consonant aached to a vowel at the beginning of
the post-caesural word is to be treated as the beginning of that laer word even if it belongs to the
end of the former: pūrvāntavat svaraḥ saṃdhau parasyādiḥ svarāt paraḥ // parādi-varṇa-saṃbaddhaṃ
vyañjanaṃ tu parādivat / prāk-padānta-sthitaṃ sādhu vyavahāreṣu dṛśyatām //.
22 Here and throughout this paper I rely on the text of the Mudrārākṣasa as edited by Alfred
Hillebrandt (H 1912). Verses of the Mudrārākṣasa are referred to by act and verse number
since the consecutive verse numbers in Hillebrandt’s edition are incorrect from the beginning of Act
5 onward. See B 2015b: 242–246 for a correspondence of Mudrārākṣasa verses in Hillebrandt’s
and Telang’s editions.
23 See B 2015a: 19–23 for further statistics.
Table 1: Caesurae obscured by vowel saṃdhi in the Mudrārākṣasa
Metre
Stanzas
Caesurae
1. Late
2. Early
3. Slip
count
%
count
%
count
%
hariṇī
3
24
0
0%
0
0%
1
4%
mālinī
4
16
0
0%
0
0%
0
0%
mandākrāntā
1
8
0
0%
0
0%
0
0%
praharṣiṇī
3
12
0
0%
0
0%
0
0%
pṛthvī
1
4
0
0%
0
0%
0
0%
rucirā
2
8
0
0%
0
0%
0
0%
śārdūlavikrīḍita
39
156
1
1%
2
1%
3
2%
śikhariṇī
18
72
0
0%
0
0%
0
0%
sragdharā
24
192
5
3%
1
1%
1
1%
suvadanā
1
8
0
0%
0
0%
0
0%
96
500
6
1%
3
1%
5
1%
Total
in my opinion, be wrien off as the result of chance or sloppy penmanship. ese
anomalous caesurae (of which there are 14 in the Mudrārākṣasa) form the principal
subject of the remainder of this study.
To see if there is a paern underlying caesurae obscured by vowel saṃdhi, I
have classified them into three distinct types and tallied the occurrences of each
type separately by metre. In Table 1 above, the second and third columns show
how many stanzas of a particular metre occur in the Mudrārākṣasa and how many
caesurae this entails (since some metres have two caesurae per line, while others
have just one). e next three pairs of columns indicate the number and proportion
of anomalous caesurae broken down by type. To make notable figures more
conspicuous, the cells for caesura types that do not occur in a particular metre are
shaded in grey. My classes of anomalous caesura are as follows:
1. Late (pūrvāntavat, in the terminology of the Yatyupadeśopaniṣat): a
caesura aer a vowel produced by the saṃdhi merging of a final and an
initial vowel (e.g. vinayā:laṃkṛtaṃ);
2. Early (parādivat): a caesura before a vowel produced by saṃdhi merging
(e.g. hatakasy:ātyantika);
3. Slip (yaṇādeśa): a caesura aer an approximant produced by saṃdhi from
a final vowel before an initial vowel (e.g. upahasaty: ekānta-bhīrūn).24
24 e Yatyupadeśopaniṣad says that this type of vowel saṃdhi should be treated as parādivat, which
implies that both the saṃdhi-derived semivowel and the preceding consonant join the next word
e Abominable Yati
Table 2: Details of anomalous caesurae in the Mudrārākṣasa
Metre
late
1.13
bhartur ye pralaye ’pi pūrva-sukṛtā:saṅgena niḥsaṅgayā
sragdharā
3.10
nālaṃkurvanti rathyāḥ: pṛthu-jaghana-bharā:krānti-mandaiḥ prayātaiḥ
sragdharā
3.24
sā mayy eva skhalantī: prathayati vinayā:laṃkṛtaṃ te prabhutvam
sragdharā
3.27
āruhyārūḍha-kopa-:sphuraṇa-viṣamitā:grāṅgulī-mukta-cūḍāṃ
sragdharā
3.27
loka-pratyakṣam ugrāṃ: sakala-ripukulo:ccheda-dīrghāṃ pratijñām
sragdharā
4.7
sotsedhaiḥ skandha-deśaiḥ: khara-kavika-kaśā:karṣaṇātyartha-bhugnaiḥ
śārdūla
2.6
devaḥ svarga-gato ’pi śātrava-vadhen:ārādhitaḥ syād iti
2.16
sā Viṣṇor iva Viṣṇugupta-hatakasy:ātyantika-śreyase
early śārdūla
slip
Stanza Text
śārdūla
sragdharā
3.19
cūḍā-ratnāṃśu-garbhās: tava caraṇa-yugasy:āṅgulī-randhra-bhāgāḥ
śārdūla
2.13
skandhe dakṣiṇayā balān nihitayāpy: aṅkaṃ patantyā muhuḥ
śārdūla
3.5
śūrebhyo ’bhyadhikaṃ bibhety upahasaty: ekānta-bhīrūn aho
hariṇī
4.2
katham idam ihety: unnidrasya: prayānty aniśaṃ niśāḥ
sragdharā
4.12
daivāt pūrṇa-pratijñaḥ: punar api na karoty: āyati-jyāni-bhītaḥ
śārdūla
7.20
kāryāṇāṃ gatayo vidher api na yānty: ālocanā-gocaram
I must note here that my categories are not mutually exclusive. Some of the instances
I classified as ‘Slip’ could have been assigned to the ‘Late’ category instead, since
they involve the enclitics api and iti, reduced to consonants and thus possessing no
prosodic value when their initial vowel merges with a preceding final vowel, while
their final vowel becomes an approximant on account of a following initial vowel.25
Although the total number of obscured caesurae is relatively low, the table
definitely gives the impression that their distribution among metres is far from
random, as they occur in but three of 10 metres, and two of these three metres exhibit
all three types of anomalous caesura. e picture becomes even sharper if we take
into account the exact location of the anomalies. Of the three metres concerned, the
śārdūlavikrīḍita (formula: – – – ⏑ ⏑ – ⏑ – ⏑ ⏑ ⏑ – : – – ⏑ – – ⏑ ⏓) has only one caesura
per line, but the other metres have two. In hariṇī (⏑ ⏑ ⏑ ⏑ ⏑ – : – – – – : ⏑ – ⏑ ⏑ – ⏑
⏓), it is the first of these two caesurae that behaves ‘abominably’, although this may
not be significant on account of the small number of cases studied. However, the
sragdharā (– – – ⏑ – – : ⏑ ⏑ ⏑ ⏑ ⏑ ⏑ – : – ⏑ – – ⏑ – ⏓) is represented by no less than
24 specimens in the Mudrārākṣasa, and all of the seven anomalies found in these
because that is where they find a vowel to which they can aach. Nonetheless I prefer to avail of
the opportunities afforded by an alphabetic script and segment such caesurae as upahasaty: ekāntabhīrūn rather than upahasa:tyekānta-bhīrūn.
25 ese cases are Mudrārākṣasa 2.13 and 4.2 (q.v. Table 2 below). us nihitayāpy: aṅkaṃ (Slip)
could, alternatively, be analysed as nihitayā:py aṅkaṃ (Late) and ihety: unnidrasya (Slip) as ihe:ty unnidrasya (Late). I have chosen to classify these as ‘Late’ caesurae because they involve enclitics, which
are by nature closely linked to the preceding word. is is also the opinion of the Yatyupadeśopaniṣad
(4ab), nityaṃ prāk-pada-saṃbandhāś cādayaḥ prāk-padāntavat.
occur at the second caesura of that metre. Table 2 above presents the anomalous
caesurae individually, grouped by category and giving a one-line context for each.
Caesurae (including the irrelevant ones) are represented by a colon in the text, and
the immediate context of each anomalous caesura appears in darker bold type.
More Sighঞngs
Before studying the habits and hypothesising about the evolution of the
abominable yati, we need to confirm that this elusive creature is not endemic to the
Mudrārākṣasa. Since first becoming aware of the phenomenon, I have repeatedly
noticed caesurae obscured by vowel saṃdhi in sragdharā and śārdūlavikrīḍita
verses, but a systematic study of a more substantial corpus of texts is yet to be
accomplished. For the purposes of this paper, I have extended my research into
two more textual domains, both of which are quite close in literary space-time to
Viśākhadaa: the three plays aributed to King Harṣa, and the Mandasor pillar
inscription of Yaśodharman.26
For the former, I have relied on a previous study by Roland S (1997:
243–253). In his analysis he classified caesurae into categories similar (though not
identical) to mine. His findings differ from what I have found in the Mudrārākṣasa
in that he reports (ibid. 245–246) six instances of a caesura without vowel saṃdhi
occurring between a prefix and a stem, a phenomenon that is entirely absent
from the Mudrārākṣasa, but permied by some theoreticians.27 However, he also
finds numerous (no less than 42) anomalies of the type that interests me here:
caesurae occluded by vowel saṃdhi. One of these (the only specimen of my ‘Slip’
type in the Harṣa corpus) may not be quite relevant to the present analysis, since
it involves a stanza in the pṛthvī metre which, as noted above, does not follow
caesura rules very stringently.28 Of the remaining 41 obscured caesurae, all but one
26 e striking resemblance of some stanzas of Yaśodharman’s inscription to some verses of the
Mudrārākṣasa was first pointed out by A 1922, but ignored by subsequent scholars. As I discuss in B 2015b: 228–230, Viśākhadaa may well have been connected to the Aulikara court
of Daśapura around, though not necessarily during, Yaśodharman’s reign. As for Harṣa, his metrical
profile—i.e. the proportions in which he employs various metres in his dramas—is strikingly close to
that of Viśākhadaa; see B 2015b: 217 for details.
27 e Yatyupadeśopaniṣad implicitly permits breaks between a polysyllabic prefix and a stem by
the more general rule that intra-word breaks are permied so long as both segments are longer than
one syllable. e Kāvyālaṅkārasūtravṛi of Vāmana (vṛi to sūtra 2.2.4, J 1971: 62) permits all
breaks so long as they are not within a verbal or nominal stem: yati-bhraṣṭaṃ dhātu-bhāga-bhede
nāma-bhāga-bhede ca sati bhavati / … dhātu-nāma-bhāga-pada-grahaṇāt tad-bhāgātirikta-bhede na
bhavati yati-bhraṣṭatvam /.
28 e caesura in question is in pāda b of Ratnāvalī 2.16: caturbhir api sādhu sādhv: iti mukhaiḥ
samaṃ vyāhṛtam. My reason for classifying this as a ‘Slip’ caesura (sādhv: iti) rather than, as S
(ibid. 247–248) sees it, a truly abominable transgression (sā:dhv iti) is that all other pṛthvī lines in
Harṣa’s plays have tidy caesurae throughout (three regular lines in this stanza and four in Ratnāvalī
e Abominable Yati
Table 3: Anomalous caesurae in Yaśodharman’s inscription
Stanza Text
late
early
slip
3
sa śreyo-dhāmni samrāḍ: iti manu-bharatā:larkka-māndhātṛ-kalpe
4
ye bhuktā gupta-nāthair: nna sakala vasudhā:kkrānti-dṛṣṭa-pratāpair
4
nājñā hūṇādhipānāṃ: kṣitipati-mukuṭā:ddhyāsinī yān praviṣṭā
4
vīryāvaskanna-rājñaḥ: sva-gṛha-parisarā:vajñayā yo bhunakti
5
ā lauhityopakaṇṭhāt: tala-vana-gahano:patyakād ā mahendrād
6
nīcais tenāpi yasya: praṇati-bhuja-balā:varjjana-kliṣṭa-mūrddhnā
7
nirddeṣṭuṃ mārggam uccair: ddiva iva sukṛto:pārjjitāyāḥ sva-kīreḥ
7
tenākalpānta-kālā:vadhir avanibhujā: śrī-yaśodharmmaṇāyaṃ
6
cūḍā-puṣpopahārair: mmihirakula-nṛpeṇ:ārccitaṃ pāda-yugmam
7
stambhah stambhābhirāma-:sthira-bhuja-parigheṇ:occhritiṃ nāyito ra
3
rājasv anyeṣu pāṅsuṣv: iva kusuma-balir: nnābabhāse prayuktaḥ
occur in the two metres in which Viśākhadaa also repeatedly employs obscured
caesurae: śārdūlavikrīḍita and sragdharā. e single exception is a ‘Late’ caesura
in a śikhariṇī stanza (Nāgānanda 3.8d, cited in S 1997: 246). One further
difference from Viśākhadaa’s praxis is that Harṣa seems to use such caesurae at
the first yati of the sragdharā metre at least as oen as at the second (4 instances
at the first, 2 instances at the second caesura in the Harṣa corpus as reported by
Steiner).
e Mandasor pillar inscription of Yaśodharman29 is a praśasti composed entirely
in sragdharā (discounting a colophon consisting of a single anuṣṭubh verse) and, as
expected, it yields a bounty of ‘abominable’ yatis. As shown in Table 3 above, in a
mere seven stanzas one finds eight specimens of ‘Late’ caesura, two of the ‘Early’
type, and one ‘Slip’. One of the ‘Late’ caesurae (in 7c) and the single ‘Slip’ is at the
first yati point of the metre, while all others are at the second. ere are in addition
two instances (both in verse 1) of a caesura falling between a verbal prefix and the
verb, an anomaly found in the Harṣa corpus but not in Viśākhadaa’s play, and not
involving vowel saṃdhi (hence not listed here).
It can thus be amply demonstrated that the caesura obscured by vowel saṃdhi
is frequent enough in the works of at least some authors to be a consequence of
something other than chance negligence or an idiosyncrasy of Viśākhadaa.
4.17). Furthermore, the examples G (1978) cites for the non-observance of the caesura in pṛthvī
almost always involve a ‘hard’ transgression (i.e. the expected caesura falls inside a morpheme by all
counts), whereas this example can be explained otherwise if we recognise saṃdhi-obscured caesurae
as a special case.
29 F 1888: 142–148.
A Tentaঞve Explanaঞon
S (1997: 246–247) theorises that anomalous caesurae could have been
realised as a pause in recitation without impeding listeners’ understanding of the
text, since they generally involve the separation of common and easy-to-grasp
morphological elements such as prefixes, common enclitics or the final vowel of
a two-syllable declensional ending (-ena, -āya or -asya). He also observes (ibid.
248–249) that anomalous caesurae occur more frequently in certain metres than in
others. My findings provide the basis for making this observation more concrete:
a particular type of anomalous caesura (namely that obscured by vocalic saṃdhi)
occurs more frequently (or perhaps, solely) at specific points of specific metres. For
this reason I am convinced that the key to understanding this not-so-abominable
but definitely intriguing species of yati lies in its metrical context rather than in its
morpho-syntactic or semantic/pragmatic environment.
A clue pointing this way can be found in the recitational tradition associated
with classical poetry. To be sure, the way present-day paṇḍits and other recipients
of a traditional education chant Sanskrit verses may not have much in common with
the performance techniques that were in vogue a millennium or two ago. However, it
does appear that people from diverse linguistic and geographic backgrounds within
India oen recite a given classical metre using a remarkably similar tune (or chant
intonation), which makes it at least plausible that their recitational practice goes back
to a single tradition that is both antique and widely prevalent. Modern studies of such
performance practices are sadly lacking, and my own knowledge of the field is deplorably
deficient,30 but it is my impression that most performers who chant a śārdūlavikrīḍita
stanza prolong the syllable preceding the caesura in that metre, and typically elaborate
it with a melisma (a slur, a legato shi to a different note in the prolonged syllable).
Some performers do the same before the second caesura of sragdharā.31 e slurring
or prolongation if the pre-caesural syllable is not, however, ubiquitous: many metres
involving caesurae are recited without this feature. Fortuitously, such prolongation
(with or without a legato shi) provides an opportunity for the performer to resolve
saṃdhi (see line B5 in Table 4 below for an illustration), and I think it is hardly an
accident that caesurae which can only be realised in performance by restoring a vowel
lost in saṃdhi occur precisely at such points.32
30 What lile insight I do have, I owe to Ferenc Ruzsa who long ago shared with me and a few colleagues what he in turn had learned from Sadananda Das at his Spoken Sanskrit course in Heidelberg.
31 For audio examples of prolongation of the precaesural syllable with a melisma in śārdūlavikrīḍita
(recited by Sadananda Das), see subhāṣitas 12–15 at the website D n.d. For prolongation (without
a melisma) in both śārdūlavikrīḍita and sragdharā (recited by Ashwini Deo), see D n.d. Further
illustrations are readily available on YouTube.
32 I must admit that I do not know whether a representative of any currently living recitational
tradition would instinctively resolve saṃdhi at these points if asked to chant a stanza with such a
caesura. is is relatively easy to test, and my prediction is that at least some of those proficient not
only in recitation but also in Sanskrit grammar would indeed do so. A confirmation of this would
e Abominable Yati
Even if we can take it as established that there is a connection between
the slur in performance and the slurred caesura in a text, this is still but an
interesting correlation rather than a causal relationship. I contend that the
reason why both these phenomena occur at particular metrical positions and
not at others must be sought in the deep structure of the metres of classical
Sanskrit poetry.
A ground-breaking paper by Ashwini D (2007) has shown that these
classical metres can in fact be viewed in terms of generative metrics as particular
instantiations of foot-based schemata alike in nature to those of Western classical
poetry. In the laer tradition poets selected a schema to follow in a composition,
but retained some freedom to implement each foot in a variety of metrical formulae
(employing for instance an arbitrary mixture of dactyls and spondees in the first four
feet of a hexameter line). But according to Deo the Sanskrit tradition—presumably
at a very early stage—solidified at the level of actual instantiations. is hypothesis
is appealing in its elegance and intuitively convincing (to me at least). It does,
admiedly, go counter to practically all traditional Sanskrit metrical theory, which
never admits any flexibility in syllabo-quantitative metres33 and usually describes
metrical schemata as consisting of arbitrary units of three syllables (gaṇas or trikas)
rather than of feet of specific moraic quantity but variable syllable count. It must also
be admied that the minutiae of the derivation of a particular classical metre from
a certain foot-based schema remain unproven and for the most part unprovable.
Nonetheless, various metrical oddities—including the obscured caesura—may serve
as circumstantial evidence to substantiate Deo’s hypothesis and may even be useful
for reconstructing the schemata hidden behind classical metres.
As to the maer at hand, D (2007:102–103) describes two independent
functions of the caesura in Sanskrit metres. One of these is to split up cola
based on abstract schemata consisting of feet with a different mora quantity;
the other indicates ‘gaps’ where a particular metrical position remains unfilled
because the derivation of the metrical formula from the abstract schema
involves catalectic (incomplete) or syncopated (off-beat) feet. Any particular
caesura may fulfil either or both of these functions. Having described herself
as ‘a fluent participant’ in the recitational tradition (ibid. 70), D also notes
that in recitational practice the syllable preceding caesurae occurring at unfilled
metrical positions is typically prolonged, whereas caesurae performing only the
supply further evidence for my hypothesis, though a negative finding would (conveniently) still not
falsify my speculation, which pertains to reciters and poets a millennium or two ago.
33 With very few exceptions, such as the variation of heavy or light initial syllables in upajāti and
vaṃśamālā. For exceptions in praxis (not, to my knowledge, provided for in theory), P (1977:
210) cites some very interesting early śārdūlavikrīḍita stanzas from the Lalitavistara, in which the
first (heavy) syllable of each line appears to alternate freely with two light syllables. is alternation
seems to be an important piece of evidence in favour of the idea that classical syllabo-quantitative
paerns are derived from earlier foot-based paerns. Another phenomenon that may be worth studying from this aspect is the occasional nine-syllable line in epic anuṣṭubh.
Figure 1: Two ways of parsing śārdūlavikrīḍita.
first function are realised as a pause which ‘is brief and never alternates with
[…] vowel lengthening’ (ibid. 103).
It thus follows logically that the slurred caesurae observed in texts may
correspond to empty metrical positions. What remains to be done is to show that
such empty positions actually occur in the abstract schemata for the metres where
we empirically find slurred caesurae. Deo presents reconstructed schemata for a
number of popular Sanskrit metres, yet her proposed parses for śārdūlavikrīḍita or
sragdharā do not include empty metrical positions. Nonetheless, D (ibid. 93) is
the first to admit that ‘there is no sure-fire solution’ to the problem of determining
whether the derivation of a classical Sanskrit metre involves syncopation or
catalexis, since a particular metrical template may be the surface instantiation
of several alternative foot-based schemata.34 e schema which D (ibid. 109)
proposes for śārdūlavikrīḍita consists of iambic feet, tetramoraic before the caesura
and pentamoraic aer it, with an extrametrical syllable (anacrusis) at the start and
a truncated foot (catalexis) at the end. In order to account for the slurred caesura, I
propose an alternative reconstruction based on trochaic feet. As in Deo’s parsing,
the feet are tetramoraic before and pentamoraic aer the caesura; however, this
schema involves anacrusis at the start of the second colon and catalexis at the end
of the first colon, with the added twist of syncopation at the end of the third foot
of the first colon.
ese alternative ways of deriving the śārdūlavikrīḍita paern from an abstract
schema are illustrated in Figure 1, with the upper half showing Deo’s iambic
34 I must, however, stress that this does not imply that a wide variety of arbitrary abstract schemata
can be constructed for any given syllabo-quantitative formula. e constraints proposed in Deo’s
paper may be incorrect in some details, but in general they are plausible and seem to conform well
to the observed variety of metrical templates. Staying within these constraints, the array of possible
parsings for a given template is severely limited even with quirks like syncopation, anacrusis and
catalexis.
e Abominable Yati
schema and the lower half, my proposed trochaic schema. e metrical template
(shown in the middle of the figure) is identical except for the hypothesised pauses—
before the caesura in my version and at the end of the line in Deo’s. e numerals
at the top and boom of the figure show the number of morae in each foot. Feet
are separated by vertical lines, a double line indicating the caesura and a doed
line showing the boundary of feet joined by syncopation. Columns shaded in grey
indicate extrametrical syllables. Between the numbers and the metrical formula,
trees show how the feet branch into metrical positions, each of which may be
instantiated as a syllable or subdivided into two syllables.35 e thicker line in a
tree indicates the strong metrical position and syllable (also indicated by an accent
in the prosodic formula).
As far as I can tell, the two derivations are equally plausible in theory. In fact,
the seven-syllable metrical phrase preceding the caesura of śārdūlavikrīḍita (– ⏑
– ⏑ ⏑ ⏑ –) is also found in the rare metre called candravartma, for which D
(2007: 95–96) actually presents a syncopated trochaic derivation like mine here. It
is in my opinion generally desirable to aempt to present identical schemata for
identical sequences in the surface instantiations, so proposing this derivation for
śārdūlavikrīḍita makes this application of the theory more parsimonious.
As for practice, Deo’s iambic schema certainly fits a regular śārdūlavikrīḍita
line (illustrated in line A1 of Table 4), but explains neither the melisma or
prolongation observable in traditional performance (including Deo’s own), nor
the anomalous caesurae in poetic praxis. Lines A2 and A3 of Table 4 show the two
ways in which a caesura obscured by vowel saṃdhi might be realised as a break in
performance. e first mutilates a word, while the second is extremely awkward
to pronounce.
In my alternative trochaic schema (illustrated in part B of Table 4), the
catalectic last foot of the first colon provides the unfilled metrical position
that is in my view the underlying cause of both the slurred extension in
performance and the slurred caesura in the texts. Early in the prehistory of
Sanskrit versification, an empty metrical position before a caesura may have
been realised in performance as a bimoraic pause (line B1) or, alternatively, as a
prolongation of the preceding syllable, which would thus have been recited as a
tetramoraic melisma (line B2).
A saṃdhi-obscured yati at an unfilled metrical position realised through such
a melisma (line B4) remains as abominable as one realised through a bimoraic
pause (line B3). However, the extension of the precaesural syllable provides an
opportunity for the resolution of saṃdhi by reciting two syllables in place of a
single melismatic syllable (line B5). I hypothesise that this perceived opportunity
opened up a way for poets to compose śārdūlavikrīḍita stanzas with a merged
35 I follow D (2007: 75–78) in the method of generating syllabic templates from the metrical
schemata.
Table 4: Śārdūlavikrīḍita schemata and examples
In both parts of Table 4, columns represent feet in the abstract metrical schema. A shaded column
indicates extrametrical syllables. Vertical lines show the boundaries of feet; a double line between
columns marks the location of the caesura, while a doed vertical line indicates feet joined because
of syncopation. e text samples both come from verse 3.5 of the Mudrārākṣasa.
A. Schema in iambic feet
A
–
–
⏑⏑
⏑⏑
⏑⏑
–⏑
A1
śrī
r labdha
prasare
va veśa
vanitā
duḥkhopa caryā bhṛ
–⏑
⏓○
śam
A2
śū
rebhyo
’bhyadhikaṃ bibhety u pahasa
ty ekānta
bhīrūn a
ho
A3
śū
rebhyo
’bhyadhikaṃ bibhety u pahasaty ekānta
bhīrūn a
ho
B. Schema in trochaic feet
B
–
⏑⏑
B1 śrīr la bdhaprasa
⏑
⏑⏑⏑
○
–
⏑–
⏑⏓
reva ve
śavani
tā _
duḥ
khopaca
ryā bhṛśam
ryā bhṛśam
B2 śrīr la bdhaprasa
reva ve
śavani
tāā
duḥ
khopaca
B3 śūre
bhyo ’bhyadhi
kaṃ bibhe
ty upaha
sa _
ty e
kāntabhī rūn aho
B4 śūre
bhyo ’bhyadhi
kaṃ bibhe
ty upaha
saa
ty e
kāntabhī rūn aho
B5 śūre
bhyo ’bhyadhi
kaṃ bibhe
ty upaha
sati
e
kāntabhī rūn aho
vowel at the caesura position and for audiences to appreciate such stanzas as
metrically correct. If in such cases the reciter restores hiatus and does not realise
the caesura as a pause,36 then this sort of yati becomes an interesting variation
rather than an abomination.
I believe that it should also be possible to reconstruct a schema for sragdharā in
such a way that it accounts for the occurrence of slurred caesurae at the second yati
position of this metre (or possibly at both positions). Despite some effort expended
in this direction, I have not succeeded in producing such a derivation, and D
(2007: 110) also notes that her own schema for sragdharā is problematic. For now
I can only say that there may be more to sragdharā than meets the eye, and the
reconstruction of its schema should be aempted with a view to other metres with
related templates, such as mandākrāntā, suvadanā and citralekhā.
Conclusions
To reiterate the main points of my somewhat convoluted argument, there is
a twilight zone between clear-cut cases of correct and transgressive caesurae
36 A momentary pause of no substantial moraic duration would be acceptable in the ‘Early’ and
‘Slip’ classes of obscured caesura, but it would fall inside the post-caesura word before the caesura in
the ‘Late’ category (e.g. pūrvasukṛta-ā:saṅgena).
e Abominable Yati
in Sanskrit syllabo-quantitative verse. is zone is inhabited, among others, by
a type of yati where the word break is obscured by vowel saṃdhi. is type of
yati is theoretically permied by the core text on caesurae, and instances of it are
found in a variety of poetic texts. An exact study of a small sample of texts and a
general impression of a much larger body of literature definitely shows that the
frequency distribution of the saṃdhi-obscured yati peaks in a narrow range of
metrical templates rather than being even throughout the gamut of metres. is
indicates that the employment of obscured word breaks is not something poets
viewed as a transgression of the caesura (done out of clumsiness or to heighten
the literary effect of a poem by omiing an expected break). It likewise indicates
that we are not dealing with a general sort of licence that allows fuzzy caesurae.
In both these cases, saṃdhi-obscured yatis would not differ markedly in frequency
from metre to metre. erefore there must be something special in the immediate
metrical context of the caesurae that allow such slurs. I theorise, adopting the
views of D (2007), that this something is an unfilled metrical position preceding
the caesura in the abstract schema of the metrical template. e unfilled position
may be realised in performance as a pause with measurable moraic duration, but
it also may (and oen is) be realised as a prolongation of the pre-caesural syllable.
I believe, though this is probably impossible to prove, that it was this recitational
practice of prolonging the pre-caesural syllable that allowed poets to employ covert
caesurae in metres that possess such an unfilled position, but not in other metres. In
performance, then, the caesura would have become overt when the reciter restored
hiatus instead of prolonging the pre-caesural syllable.
It is in my opinion beyond doubt that the distribution of the obscured yati across
metres is not random. Even if further research should render all my other claims
invalid, this in itself is a considerable finding. It shows that there was a paern
or rule to versification which (at least some) classical Sanskrit poets, consciously
or instinctively, employed in their works, but which was not articulated in any
treatise on poetics, traditional or modern, that I am aware of.
If, in addition, I am correct in my assumption that the obscured caesura is
linked to the slurred performance of pre-caesural syllables in certain metres, then
this gives us an insight into the process of the composition of classical Sanskrit
poetry. It is a concrete manifestation of the primarily aural (śravya) nature of this
poetry inasmuch as it shows that poets, though they may have commied their
products to writing immediately aer or already in the course of the creative
act, were influenced by performative practice. ey probably hummed the lines
they were working on and possibly not even noticed consciously that there was
anything out of the ordinary with obscured caesurae in metres that allowed for
the restoration of hiatus by co-opting a prolonged syllable to this purpose. One
can even imagine a sort of coevolution in which the development of the obscured
caesura on the textual side and the prolonged or melismatic chanting style on the
performative side reciprocally catalysed one another.
Further, if I am right to believe that the root cause of the obscured caesura is
an unfilled metrical position in the abstract metrical schema (which may be true
or false regardless of whether I am correct about interaction with performance),
then we have a practical implication of Deo’s claims about the verity of such
abstract schemata. e reasoning is admiedly a bit cyclical: if Deo is right, then
her hypothesis can be used to explain the obscured yati, and if such an explanation
works, then Deo’s hypothesis is corroborated. However, it is very hard to conceive
of concrete evidence for a metrical system that, if it ever existed, was already
forgoen by the time of the earliest extant śāstras dealing with classical metres.
erefore, even tenuous evidence merits consideration. ere certainly remains
much work to do on reconstructing schemata for at least the common classical
metres. Such work should pay close aention to metrical phrases that occur with
lile or no variation in a number of established metres.37 Should it turn out to be
possible to (re)construct schemata that include an empty metrical position in all or
most metres which exhibit the obscured caesura in poetical praxis, and to do so in
such a way that these schemata are consistently applicable (mutatis mutandis) to
other metres involving identical or related constituent phrases, this interconnection
would be evidence in favour of both the general correctness of Deo’s hypothesis
and of my hypothesis about the origin of the obscured caesura.
e phenomenon of the obscured caesura should be examined in a much
larger corpus of texts, first with an eye to see if it occurs with any regularity in any
other metres beside śārdūlavikrīḍita and sragdharā. On the basis of my studies so
far, pṛthvī and hariṇī may be good candidates for such metres, though the case of
the former has the additional complication that it was not universally regarded to
require a caesura. Another likely candidate is the mandākrāntā, which shares much
of its metrical structure with sragdharā and which is frequently used (in addition
to śārdūlavikrīḍita and sragdharā) in commentaries to illustrate cases of saṃdhiobscured yati. Once the metres prone to exhibit obscured yatis are identified more
securely, further study of a larger body of texts could be used to determine whether
their use is quite universal or peculiar to certain authors, ages, regions or genres.
e texts dealt with in this paper suggest that it was quite common in North India
in early post-Gupta times. My cursory investigation of Kālidāsa’s dramas, however,
has yielded a mere two instances of obscured caesurae in the Abhijñānaśākuntala38
and none in the poet’s other plays. (ere are, however, a number of instances in the
37 Deo has already explored what she calls ‘the Sanskrit trochaic tetrameter’ (D 2007: 72–73) and
the indravajrā family (ibid. 82–89), but not the constituent phrases of the more elaborate ‘courtly’
metres. Velankar’s compilation of ‘vṛa-ghaṭakas’ (V 1951) may be a good starting point
for studying promiscuous metrical phrases, but needs to be extended to paerns with a shared foot
structure.
38 Verses 1.18 and 1.32 in K 1980. Neither of these is a śārdūlavikrīḍita or sragdharā verse, but
both involve the separation of a prefix from a stem. Such separation occurs elsewhere in Kālidāsa’s
plays in contexts not involving vowel saṃdhi, so this is probably a phenomenon quite distinct from
the case of the abominable yati.
e Abominable Yati
mandākrāntā metre in the Meghadūta, some of which are even cited in commentaries
as examples of the phenomenon.) Obscured caesurae also seem to be uncommon,
though not unheard of, in imperial Gupta epigraphs.39 It remains to be seen whether
this type of yati is more or less frequent in geographically and temporally more
distant texts. A larger dataset may also shed light on whether the three varieties
of obscured caesura behave identically or in different ways—perhaps even give
meaning to the proviso ‘kvacid eva’ for the parādivat variety. Another direction for
further research would be to examine whether the slurring of the yati ever occurs at
the ends of odd pādas, and if it does, whether it does so in a systematic way.40
References
A, V. J. (1922). e date of the Mudrā-Rākshasa. Indian Antiquary, 51: 49–51.
A, V. S. (1957–59). e practical Sanskrit-English dictionary (3 vols.) Poona: Prasad
Prakashan.
B, D. (2015a). rdög úr esete a cifra cezúrákkal avagy a szanszkrit időmértékes
verselés sormetszetei Viśākhadaa költészetében’. [Strange Caesurae in Sanskrit
Metrical Verse: A case study in Viśākhadaa’s poetry]. In Kakas, B. & Szilágyi,
Z. (Eds.), Kéklő hegyek ala lótuszok tava – Tanulmányok Bethlenfalvy Géza
tiszteletére [Lotus Lake Under Blue Mountains – Essays in Honour of Géza
Bethlenfalvy], (pp. 25–43). Budapest: L’Harmaan.
—–. (2015b). A Textual and Intertextual Study of the Mudrārākṣasa (Unpublished doctoral
dissertation). Eötvös Loránd University, Budapest, Hungary. Retrieved from hp://
doktori.btk.elte.hu/lingv/baloghdaniel/diss.pdf.
D, S. (n.d.). Recited Subhashita files. Devavānī. Retrieved from hp://www.devavani.
org/hablado2eng.html (last accessed September 2016)
D, A. (n.d.). Classical Sanskrit meters. Retrieved from hp://pantheon.yale.
edu/~asd49/meters.html.
39 For instance, verses 5 and 8 of the Allahabad inscription of Samudragupta (F 1888: 6, line
9, amanuja-sadṛśāny: adbhuta° and line 16, kavi-mati-vibhavo:tsāraṇaṃ), verse 1 of the Bilsad pillar
inscription of Kumāragupta I (F 1888: 44, line 10, sphaṭika-maṇi-dalā:bhāsa-gaurāṃ), and verses 1
and 3 of the Kahaum Pillar of (the time o) Skandagupta (F 1888: 67, line 4, varṣe rinśad-daśaiko:
araka-śatatame and line 12, giri-vara-śikharā:gropamaḥ kīri-karā). All of these specimens are in
sragdharā, and all but one at the second caesura of that metre.
40 e Mudrārākṣasa has just a single such instance, found in 5.3, a śikhariṇī stanza whose quarter
c ends with a ‘Slip’, bahu-prāpita-phalety / aho. I have heard verses of this metre recited with a legato
extension of the last syllable, and D’s derivation (2007: 106–107) of the formula involves catalexis
in the last foot, so the finding, though isolated, fits the hypothesis proposed in this paper. Commentators’ illustrations of saṃdhi-obscured line breaks (lumped in with caesurae under the umbrella term
yati) also include stanzas in anuṣṭubh (Halāyudha on Chandaḥsūtra, K̄̄ 1938: 102), upajāti
(Halāyudha on Chandaḥsūtra, K̄̄ 1938: 102; Sulhaṇa on Vṛaratnākara, S et al. 1969:
28), mālinī (Sulhaṇa on Vṛaratnākara, S et al. 1969: 28) and śārdūlavikrīḍita (Somacandra on
Vṛaratnākara, S et al. 1969: 31).
—–. (2007). e metrical organization of Classical Sanskrit verse. Journal of Linguistics,
43(1): 63–114.
D, D. (2011). Śabdālaṃkāradoṣavibhāga. Die Unterscheidung der Lautfiguren
und der Fehler. Wiesbaden: Harrasowitz.
F, J. F. (1888). Inscriptions of the early Gupta kings and their successors (Corpus
Inscriptionum Indicarum, vol. 3) Calcua: Government Press.
G, A. (1978). Caesura in Pṛthvī metre. Annals of the Bhandarkar Oriental Institute,
58/59, 627–634.
H, A. (Ed). (1912). Mudrārākṣasa by Viśākhadaa. Edited from mss. and
provided with an index of all Prākrit words. Breslau: Marcus.
J, B. (Ed.). (1971). Kāvyālaṅkāra Sūtra of Āchārya Vāmana. Benares: Chowkhamba
Sanskrit Series Office.
K, D. K. (Ed.). (1980). A reconstruction of the Abhijñānaśākuntalam. Calcua:
Sanskrit College.
K̄̄ (Ed.). (1938). e Chhandas Śāstra by Śrī Piñgalanāga: With the commentary
Mṛitasañjīvanī by Śrī Halāyudha Bhaṭṭa. Bombay: Nirṇaya Sāgar Press.
K, D. D. (Rd.). (1948). e epigrams aributed to Bhartrhari, including the ree
Centuries. Bombay: Bharatiya Vidya Bhavan.
P, S. (1977). Aspects of versification in Sanskrit lyric poetry. New Haven, CT:
Americal Oriental Society.
P, Alan. (1989). Metrical forms. In: Kiparsky, P. & Youmans, G. (Eds.), Rhythm and
meter (pp. 45–80). San Diego: Academic Press.
S, C. S. R. (Ed). (1956). Kāvyālaṅkāra of Bhāmaha, vol. 1., paricchedas 1 to 6.
Madras: Sri Balamanorama Press.
S, A., D, K. P, D. G. (Eds.). (1969). Vraratnakara of Sri Kedara
Bhaa with four commentaries. Hyderabad: Osmania University.
S, R. (1997). Untersuchungen zu Harṣadevas Nāgānanda und zum Indischen
Schauspiel. Swisal-Odendorf: Indica et Tibetica.
V, H. D. (1949). Jayadāman (A collection of ancient texts on Sanskrit prosody and
a classified list of Sanskrit metres with an alphabetical index). Bombay: Haritosha
Samiti.
—–. (1951). e Vṛa-Ghaṭakas. Journal of the Bombay Branch of the Royal Asiatic
Society (New Series), 26, 150–157.
W, A. (1863). Ueber die Metrik der Inder. Berlin: Harrwitz und Gossmann.
Dániel Balogh