Study 1: Orthographical Sanskritization


Prof. Franklin Edgerton in the Introduction to his Buddhist Hybrid Sanskrit Grammer and Dictionary (BHSG 1.56) made the significant statement that “it is ... certain that some Sanskrit-appearing features [in BHS] are orthographic only; the words were pronounced as in Middle Indic.This is not something that affects BHS only, as is well-known the Vedic texts have also been Sanskritised according to the rules of the later language, see the extract from Arnold's book on Vedic Metre, elsewhere on this website.01

This first study attempts to show in detail how a metrical analysis of the verses of Udānavarga can help us to restore the way the verses were pronounced, over and against the way that they have been written in the text as it appears now. A disclaimer is necessary at the outset: the metre is only a help in the reconstruction of the pronounication. In regard to certain features of the language of Udānavarga the metre is of no help at all.

Take the Śloka pādayuga: śīle pratiṣṭhito bhikṣuś cittaṁ prajñāṁ ca bhāvayet [156ab], a number of questions arise that the metre cannot help us with: were the sibilants distinguished by sound as in Sanskrit or assimilated to s as in Pāḷi? was the pronounciation bhikṣu or bhikkhu? was the vowel before anusvara in prajñāṁ pronounced as long or short? was the word-final consonant in bhāvayet pronounced or not?

The metre cannot help us resolve these questions. But there are many other places where it does in fact help, and can therefore give us an idea of what the sound was, and this paper is concerned with those matters.

This will be done by considering 4 ways that the text has been rewritten to make it appear more Sanksritic than it really was, they can be enumerated as follows:

1: Expansion of syllables - this is the case where syllables that have been contracted in Middle Indic have been rewritten in their full Sanskrit form.

2: Syllables that must be excluded m.c. - in this case by-forms that appear in Middle Indic, have been regularised to their Sanskrit form.

3: Syllables that do not make position - normally in Indian prosody a conjunct consonant will make the preceding vowel heavy in metrical weight, but there are many occasions where they fail to make position in this text.

4: Svarabhakti vowels - in Middle Indic the insertion of epenthetic vowels to facilitate pronounciation is a well known phenomena. In the text as we receive it some of these vowels have been excluded from the orthographic representation of the text.

There are a couple of other matters that have to be dealt with in the Introduction before we can move on to a consideration of the metres themselves, they are:

5: Hiatus and Incomplete Sanskritisation - there are places where the Sanskritisation of the text has not been completed, which as it stands leaves the metre incorrect.

6: Resolution - quite frequently one presumed heavy syllable has been resolved into two light ones, this is perfectly acceptable and within the parametres that were allowed in composition.


1: Expansion of syllables

i: -ava- = o

In the Udānavarga we can see that there are many cases where bhava- (⏑⏑) has been written, but the metre demands that the syllable be scanned as bho- (). In fact it seems that the words with this component were favourites in critical positions, where we can be sure of what the weight is in these positions.

In the text I have scanned these expansions as being equivalent to one heavy syllable, and have identified them by marking them in green and underlining them, thus:

apramādaratā bhavata suśīlā bhavata bhikṣavaḥ | [119ab]

which, for the metre to be correct we need to read as:

−⏑−⏑¦⏑−−¦¦⏑−−¦⏑−⏑−    Pathyā Śloka
apramādaratā bhota suśīlā bhota bhikṣavaḥ.

Here is a list of the words and places where these expansions are found:

bhavati 63c, 104e, 105bd, 106bd, 134c, 135c, 144c, 146a, 147c, 207c, 209d, 214e, 217d, 221a, 227c, 235d, 243a, 279a, 281cd, 346d - 351d (6 verses), 352c, 368d, 373d, 421.iid, 464a - 467a (4 verses), 475d, 546c, 548c, 655c, 657f, 674c, 700c, 703a, 765d, 875a, 858b, 881c, 890ad, 966a, 970 cd. bhavata 109ab, 369d. This gives a total of 55 occurrences.

I have therefore read bhavati as having 2 syllables (bhoti) in non-critical positions in the following places: 14c, 564d, 691d.

Note that there are a number of places where bhavati must be scanned as having 3 syllables for the metre to be correct. In Pāḷi the distinction of the written form is usually maintained, where bhavati may appear as hoti, for instance, but the distinction in this text has been lost because of the Sanskritisation. There is not one occurrence of the form bhoti (for bhavati) in Udānavarga, which as we can see must have been the actual pronounication in the majority of cases of this word.03

We should read bhavati (3 syllables) at the following verses: 13c, 147d, 269b, 280iia, 281b,, 291.iid, 421.iib, 434a, 898d, 905d (a total of 11 occurrences). -bhava- also counts as having its full complement of syllables in the following compounds: yathāvibhavato 227b; bhavatṣṇāpradālanam 248d; prabhavati 740a; vikṣīna-bhavatṣṇasya 916a, 917a; ucchinnabhavatṣṇasya 918a, 919a.

Occasionally we find that even in the same verse the word is scanned as 2 syllables in one place, and one in another: This phenomena also occurs in Pāḷi, e.g. Jā 487, vs 6 Kathaṁ bho brāhmaṇo hoti, kathaṁ bhavati kevalī...In Pāḷi it even occurs in prose, e.g. Atittāva, bhikkhave, khattiyaparisā hoti, atha rājā cakkavattā tuṇhī bhavati; perhaps for emphasis.04

⏑⏑−−¦⏑−−¦¦⏑−⏑⏑¦⏑−⏑− Pathyā
asatāṁ na priyo bhavati satāṁ bhavati tu priyaḥ || 147cd

cf. 281bcd, 421iibd.

Another instance of the same type of expansion is abhijñāvyavasito 1008b, which must be read as abhijñāvyosito.


ii: -aya- = e

We occasionally find that -aya- must also be scanned as an expansion from -e-, as in the following: deśayati = deśeti 190c; bhāvayati = bhāveti 267d; bhāvayata = bhāveta 344b.

I therefore read varjayate = vrajete 463c; deśayati = deśeti 993e, which makes more sense metrically


2: Syllables that must be excluded m.c.

There are many other Sanskritisations where a word which has a by-form in Middle Indic, has been regularised to its Sanskrit form in the text, against the metre. Again here we can see that this is merely an orthographic device, and cannot have represented the way the words were pronounced.

There are various cases where this occurs: api = pi: 30b; 106a; 765c; 855; iti = ti 290b; 680d - 683d (3 verses); iva = va 56a; 68f; 112d; 120d; 213b; 214b; 370d; 398c; 438b; 602b; 702b; 721a. There are sometimes alternative ways of understanding these forms - see section 6 below and the commentary to the text.05


3: Syllables that do not make position

We also see orthographical Sanskritization in words where there are syllables that do not make position. For details about conjuncts not making position in Pāḷi, see An Outline of the Metres in the Pāḷi Canon 1.5 elsewhere on this website.06 In Indian verse composition rhythmical patterns are built up through the alternation of light and heavy syllables. To define the weight of the syllables it is necessary to take into consideration 2 factors: the natural length of the vowel, and what follows it.

If the vowel is long (ā, ī, ū, , e, ai, o, & au) normally the syllable will count as heavy. If the vowel is short (a, i, u, , & ), then metrical weight depends on what follows it. If it is followed by another vowel, or by a simple consonant, it will be be light in weight; if it is followed by a conjunct or by niggahīta normally it will be heavy.

However, certain conjunct consonants do not make the preceding syllable heavy as expected, they are then said to be not making position. In the Udānavarga the most frequent cases of this are found when a consonant is in conjunction with the semi-vowel sound -r- .

In this section I have examined all the conjuncts with -r- that may be liable to fail in this regard, and have made a list of the result. In the lists which follow the words are given in Indian alphabetical order.07 In the commentary to the text, however, to prevent a profusion of notes I only normally identify the conjuncts that have failed to make position, unless there is a need to explain why I have identified one variation over another.


i: br

As in Pāḷi br- regularly does not make position as we can see when it appears in metrically critical positions, where the weight of the syllable is assured. In the text that follows we can see that br- in brāhmaṇ- and brahma- regularly fails to make position. There is no other conjunct with br in the text which is in critical position, so that the status of the conjunct cannot be generalized from this text alone.08

I count br- as failing to make position in the following words: e.g. brāhmaṇa 1022ib; brāhmaṇam 1017d; brāhmaṇaḥ 362b, 724d, 964b, 965b, 1024d, 1026b, 1038d; śramaṇabrāhmaṇāḥ 957b, 958b, 959b, 1038d;

I have also therefore, when entering the metrical markings in non-critical positions, counted it as not making position in the following words: coṣitabrahmacaryaḥ 968c; brahmacaryavān 244b, 891b; brahmaṇaḥ 956d, 969d, 978d; brāhmaṇaṁ 970af, 971d - 977d (7 verses), 979d - 992d (14 verses), 993f, 994d - 1001d, 1002f, 1003d - 1012d, 1013a-d; 1014a-d; 1015a-d; 1016d, 1017id, 1017iid, 1020d, 1021id; brāhmaṇasya 1026a; brāhmaṇasyedśam 1039a; brāhmaṇā 1018d, 1019if, 1019iif.


ii: śr

śr makes position in: ananuśrutaṁ 261a; alpaśruto 465a, 466a; bahuśrutaṁ 468a; bahuśrutam 562b, 974b; bahuśrutasya 795b, 796b; bahuśrutāḥ 780b; bahuśrutaiḥ 539b; bahuśruto 464a, 467a; miśro 653b.

It seems therefore that words from √ śru normally make position, But see the exception just below.09 therefore I also count it making position in the following non-critical positions: aśrutvā 461c; bahuśrutyena 903b; śrutaṁ 458a; śrutamātratā 582b; śrutvā 104c; śrūyate 474.iia.

śr fails to make position in the following words: gautamaśrāvakāḥ 313b - 330b (18 verses); śrāvakaḥ 692d; śrutvā 541c; śreyaḥ 537d; śreṣṭhaṁ 218a; śraiṣṭhyaṁ 542c; śrotriyau 724b, 1024b, 1025b; samyaksaṁbuddhaśrāvakāḥ 379d. It seems from this that words with śrāvak- (written sāvak- in Pāḷi) do not make position normally, even though they are derived from √ śru.

As it appears that śr is likely to fail to make position in most words, I also count it as not making position in the following verses, where it occurs in non-critical positions: aśraddhaś 723a; aśrāddhebhiḥ 538a; miśrā 743b; vijñānaśrutaṁ 475c; viśreṇayitvā 244c, 891c; śuśruṣur 219c; śraddhā 222a; śraddhāsyanti 453a; ; śrotreṇa 474.ia; śramaṇaḥ 956d, 969d; śramaṇo 964a, 965a; śreṣṭham 290b; saddharmaśravaṇe 225b.


iii: pr

The places where pr in this text makes, or fails to make, position is not so easy to define, and it seems that it may have been counted or not counted as making position according to convenience. According to the analysis made here Edgerton's rather categorical statement (BHS Reader, p. vi) that “An initial consonant cluster never makes position (in BHS)” is incorrect - it is true that usually they do not make position, but there are many exceptions where they must in fact be making position for the metre to be correct. See e.g. 53c; 67c; 104c; 147d; 211c; 367d; 368d; 538c; 684c; 685c; 751c; 863d etc. etc.10 In this regard compare the verses found at 684 & 685, where pr fails to make position in line a; but does make position in line c! A similar situation occurs at 262cd.

Roughly, however, we can see from the analysis that follows that pr normally makes position when it is in the middle of a word (including compounds), which is what we would expect, as these words produce gemination even in Pāḷi when in medial position.

There are a significant number of occasions, however, where it unexpectedly makes position at the beginning of a word also, where simplification of the cluster, and consequent loss of metrical weight, would be the rule in Pāḷi. We may regard these forms as being completely Sanskritised during the redaction of the verses: In what follows I have seperated medials from initials.11

a: pr makes position in the following places:

Medials: aprakampayaḥ 366b; apramattaḥ 307b; apramatto 841a; apriyam 198b; kāyapradoṣaṁ 169a; cirapravāsinaṁ 141a; niḥsaraṇaprajño 473c; notkuṭukaprahāṇaṁ 955c; buddhapratibhāsite 692c; viprasannaḥ 306b; viprahāya 344a, 790a, 994a;

vairaprasaṅgo 297c; saṁyojanavipramuktaḥ 1046d, 1047d; satpuruṣapraśastāḥ 216b; saprajñaḥ 158c, 539c, 551a, 553a, 555a; saprajño 153a; sugatapradeśitaḥ 262d; hetuprabhavaṁ 354a; I therefore take -pr- as making position in sarvaprāṇeṣu 853c also.

Initials: praceṣyate 368d; prativaśāś 271c; pratītaḥ 211c; prapañcitaṁ 751a; pramodate 685c; praśocati 684c; prasahate 711ie - 716 (7 verses); prājñaḥ 538c; priyaḥ 147d; prītisukhaṁ 863d.

b: From what follows we can see that the places where pr does not make position are predominantly at the beginning of a word, but sometimes it fails unexpectedly in middle position also:

Medials: apramattasya 89d; asthiprākāraṁ 353a; nirvāṇaprāptaye 195b, 219b; puṇyaprekṣī 537b.

Initials: pramattacāriṇas 21a prajāḥ 604b; prajñā 250a; pratirūpaṁ 878a; pratyūhaśatāni 767a; pravrajitaḥ 564c; prāptaye 262d, 267d, 856d; prāṇinām 82d; pretya 684a, 685a.

I also take prathamaṁ 482a, 483a; pravrajyā 234c; praharen 1026a; & prāptavyaṁ 606a; as not making position to satisfy the metre.


iv: other conjuncts

Because of the well-known weakness of the semi-vowels y, l, & v, I made a cursory inspection of the conjuncts with these consonants in critical position, In this examination I have counted the cadence of Śloka; and the opening and cadence of Triṣṭubh, and Mātrācchandas lines; and the gaṇas in Gaṇacchandas metres as critical.12 but was only able to find 4 places where it seems certain that they fail to make position, they are found at: tv in 132c, giving jagaṇa in an Old Gīti verse; kl- in kliṣṭaṁ at 684d in a Vaitālīya opening; dv- in dvesaṁ 935a, in a Aupacchandasaka cadence; sn- in sneham 371a, in a Vaitālīya opening.

All of the other conjuncts seem to make position. Below is a list of their occurrence divided according to the conjunct:

tr: atra 184c; ubhayatra 684b, 685b; tatra 354b, 370c; duratikramāḥ 799d. paratra 118d, 181d, 217d, 286d, 762d, 765d; yatra 53b; śatrubhūtam 767b; sarvatra 757c; sahasrāṇi 185a; sukhatrayam 149b; sucitrā 30a.
dr: bhadrikā 539d; samudramadhye 200a.

dhr: adhruvam 587b.

kr: atikramet 740f; atikrameta 417b; anantavikramaṁ 755c; kriyate 652b; duratikramam 1020ia, 1020iia; duratikramāḥ 72f; dharmacakraṁ 263a; samatikramaḥ 776b; samatikramam 643b; so 'tikramej 306d.

vr: parivrajet 83d, 283d, 878f, 882f, 994b; pravrajitaḥ 564c.

sr: parisravāṇi 298c.

Although it appears that most of the other conjuncts do in fact make position, in a limited number of places, for reasons explained in the commentary it seems better to take them as not making position in the following words (these differ from the 4 instances listed above as the scanning of the syllables is uncertain):

tr- in traividyo 208.iia, giving the pathyā cadence; -dv- in vigatadveṣebhyo in 347b, giving the pathyā cadence; dv- in dveṣaś 355c, giving the pathyā cadence; sv- in svake 563a, in a Triṣṭubh break, to avoid the heavy 6th syllable.


v: cch

A further case where we have orthographical Sanskritization is in the writing of cch, which makes the previous syllable heavy, where ch is required for the metre to be correct, as in: cchandam 52d & hiraṇyacchannaḥ 702b; cch is also occasionally written unnecessarily at the beginning of a pāda, as in: cchinnākṣaḥ 100c & cchāyā 834f; it is also written after anuśvara: cchrāmaṇyārthasya 105d.

I therefore take it as not making position in cchitvā 51c; cchannaṁ 162c; cchandraṁ 671c, 672c; vastucchinnasya 912b - 919b (8 verses); & ucchinnabhavatṣṇasya 918a, 919a.


4: Svarabhakti vowels

Sometimes svarabhakti (epenthetic) vowels that have been excluded from the orthographic representation of the text also need to be reinstated, and then given their full value. For more on svarabhakti see An Outline of the Metres in the Pāḷi Canon 1.6.13 I have identified svarabhakti vowels at the following places: utpādaviyadharmiṇaḥ 3b; arahatām 187a, 219a; diviyaṁ 216c; brahmacariyaṁ 235d; siyān 305c.


5: Hiatus and Incomplete Sanskritisation

In the Middle Indic dialects hiatus between words is perfectly acceptable, and quite often found, as a glance at nearly any Pāḷi text will show. In Classical Sanskrit, however, haitus is normally avoided. In this section I have examined all the cases of external vowel sandhi to find out how this relates to the pronounciation of the text.

Hiatus occurs in the following places: ka ānanda 4a; garbha eke 9a; rathakāra iva 56a; kṣīrapaka iva 68f; ya etāṁ 74a; chinnākṣa iva 101d; dṣṭadhārmika eko 109a; tīvra
utpadyate 127c; ya ihākṣeṇa 184b; bhasmācchanna ivānalaḥ 213d; śreṣṭha uktaḥ 218d; nānurakta iti 231a; sa ucyate 243d; sthavira ucyate 244d; ya āryam 267c; jñātvālpajñāta iti 280.iic; careta eṣaṇām 284d; sa imaṁ 302e, 335d, 337d, 339d; sa imāṁ 336d, 338d, 340d; ṣaṣṭha adhipatī 352a; 'lpamatsya iva 357d; hrada iva 366c; ka imāṁ 367a; kṣīrapaka iva 370d; ya udācchinatty 394a, 933a, 945a, 946a; abalāśva iva 398c; ca asajyamānam 417c; bāla eva 559d; bhavyarūpa iva 602b; sa eko 607d; ya evaṁ 608c; vipāka iti 680d, 681d, 682d, 683d; madhya ivodakasya 689d; lohārdhamāṣa iva 702b; śrāddha ārabdhavīryavān 711.iid; nāga iva 721a; ya icchet 759d; iṣukāra iva 817d; yaśa iha 840c; ya imāṁ 1001a; śuddha ity 1025d.

Most of these examples are within the normal limits of the metre, but there are a number of forms included in the list where the metre is incorrect as it stands, and could be corrected by the simple expedient of completing the Sanskritisation of the text. Examples of this are:

sa imāṁ 336c, 338c, 340c = semāṁ
vipāka iti 680 - 3d = vipāketi
lohārdhamāṣa iva 702b = lohārdhamāṣeva
nāga iva 721a = nāgeva
iṣukāra iva 817d = iṣukāreva


There are a few more examples where the metre could be corrected by completing the Sanskritisation, but it would also be possible to correct the metre by dropping the Sanskritisation:

rathakāra iva 56a = rathakāreva or rathakāra 'va
nānurakta iti 231a = nānurakteti or nānurakta 'ti
kṣīrapaka iva 370d = kṣīrapakeva or kṣīrapaka 'va
abalāśva iva 398c = abalāśveva or abalāśva 'va
bhavyarūpa iva 602b = bhavyarūpeva or bhavyarūpa 'va


6: Resolution

As in the Pāḷi texts a common feature of the verses of the Udānavarga is the resolution of one (presumed) heavy syllable into 2 light syllables.

Resolution cannot occur in 2nd or 8th position in a Śloka, but apart from that it can occur anywhere else, the most common position being the opening syllables of the line.

Below is a table listing the occurences of resolution found in this text according to the position and metre:



74d; 131a; 147a; 168c; 235c; 243a; 258c; 277c; 337a; 466c; 467c; 538b; 545a; 711.ia; 711iia; 712a; 713a; 714a; 715a; 716a; 746a; 801c; 816b; 878a.








203e; 256c; 257a; 480c; 575c; 623a; 632a; 633a.


160c; 401.iia; 402a, 405a; 407a, 409a; 411a; 413a.

Triṣṭubh & Jagatī


51d; 65c; 298c; 563d; 596.iabcd; 696b; 702a; 845b; 846b; 847b; 848b; 876d; 1045c; 1046c.





It can be seen from this that in this text resolution at the 1st is quite common in both of the main metres; and at the 6th it is common in the Śloka. Anywhere else resolution can only be counted as incidental. The number of verses given as having resolution at the 7th, apart from the first one, are probably all to be corrected - see the notes to 401.iia (which covers the repetitions which follow).14


7: Examples

It may be useful to show how all of this affects the text, and we can best do that by giving some examples with the written and pronounced form side by side. That is, as far as we can reconstruct it, see note 1 above.15 I include the metrical markings so that we can see how the metre can help guide us in this reconstruction of the pronounciation of the text:



etad api cchitvā tu parivrajanti [51c]


−−⏑−¦−,⏑⏑¦−⏑−−     Triṣṭubh
etam pi chitvā tu parivrajanti [51c]


alpam api cet sahitaṁ bhāṣamāṇo

dharmasya bhavati hy anudharmacārī

rāgaṁ ca doṣaṁ ca tathaiva mohaṁ

prahāya bhāgī śrāmaṇyārthasya bhavati [106]


−−⏑−,¦⏑⏑−¦−⏑−−     Triṣṭubh
alpam pi cet sahitaṁ bhāṣamāṇo

dharmasya bhoti hy anudharmacārī

rāgaṁ ca doṣaṁ ca tathaiva mohaṁ

prahāya bhāgī śrāmaṇyārthasya bhoti [106]


apramādaratā bhavata

suśīlā bhavata bhikṣavaḥ [119ab]


−⏑−⏑¦⏑−−−    Pathyā Śloka
apramādaratā bhota

suśīlā bhota bhikṣavaḥ [119ab]


asatāṁ na priyo bhavati

satāṁ bhavati tu priyaḥ [147cd]


⏑⏑−−¦⏑−−−     Pathyā Śloka
asatāṁ na priyo bhoti

satāṁ bhavati tu priyaḥ [147cd]


arthaṁ dharmaṁ ca deśayati [190c]


arthaṁ dharmaṁ ca deśeti [190c]


śraddhādhano hy arhatāṁ

dharmaṁ nirvāṇaprāptaye [219ab]


−−⏑−¦⏑⏑⏑−    navipulā
śraddhādhano hy arahatāṁ

dharmaṁ nirvāṇapāptaye [219ab]


bhāvayati mārgaṁ hy amtasya prāptaye [267]


−−⏑−¦−,⏑⏑¦−⏑−⏑−    Jagatī
bhāveti mārgaṁ hy amtasya pāptaye [267]


pāpadharmāpi ced bhavati

sa teṣāṁ bhavati pūjitaḥ [281cd]


−⏑−−¦⏑−−−    Pathyā Śloka
pāpadharmāpi ced bhoti

sa teṣāṁ bhoti pūjitaḥ [281cd]


śrutvā hy anarthaṁ varjayate [463]


−−⏑−¦−,−−−     Pathyā Śloka
śrutvā hy anarthaṁ varjete [463]


hy adhyavasitā yatra prajāḥ prasaktāḥ [604]


−−⏑−,¦−⏑⏑¦−⏑−−     Triṣṭubh
hy adhyositā yatră prajāḥ prasaktāḥ [604]


bhikṣur na tāvatā bhavati

yāvatā bhikṣate parān

veśmāṁ dharmāṁ samādāya

bhikṣur bhavati na tāvatā [890]


−−⏑−¦⏑−−−     Pathyā Śloka
bhikṣur na tāvatā bhoti

yāvatā bhikṣate parān

veśmāṁ dharmāṁ samādāya

bhikṣur bhoti na tāvatā [890]


bhovādī nāma sa bhavati

sa ced bhavati sakiñcanaḥ [970]


−−−−¦⏑⏑−−    savipulā
bhovādī nāma sa bhoti

sa ced bhoti sakiñcanaḥ [970]


na brāhmaṇasya praharen

na ca muñceta brāhmaṇaḥ [1026ab]


⏑−⏑−¦⏑⏑⏑−    navipulā
na bāhmaṇasya paharen

na ca muñceta bāhmaṇaḥ [1026ab]