Pàràyanavagga Home Page    Next Section

The elaborate composition now known to us as Pàràyanavagga is found in two places in the Khuddakanikàya of the Suttapiñaka. It forms the 5th and final section of Suttanipàta; and it is found again together with its ancient commentary in the Cullaniddesa. The text is essentially the same in both places, except in regard to some small, but nevertheless important, readings. ¹

It has been noted before that the Vatthugàthà are much later in composition than the Pucchà, this can be shown to be the case on linguistic and doctrinal grounds. On the other hand the Pucchà are regarded by scholars as containing some of the earliest recorded sayings of the Buddha. ²

In this paper I intend to examine the metre of Pàràyanavagga in some detail. The first section concerns certain matters that have to be noticed in regard to the scansion of the text. In this section I also demonstrate that there is an hitherto unrecognized rule that was used in Pàëi metrical composition, which I call the rule of resolution. The second section examines the lines written in the Siloka metre as they appear in the Pucchà and in the Vatthugàthà; with the results from that it has been possible to throw light on the date of the epilogue. The third section studies the Tuññhubha lines, and shows that the parametres of the early Tuññhubha are wider than has been previously supposed, and that we must accept that there are one or two secondary openings, and an unusual variation in the break ³.

In BJT & ChS the text is first stated in full at the beginning of the books. Then the Pucchà and epilogue are repeated together with Niddesa's commentary ⁴. The comparison has normally been made with the first statement of the text in these cases. Euro & Thai have the text from the Pucchà onwards only as it stands embedded in the commentary, the textual comparison therefore has been made with the verses as they are recorded there.

Pàëi metrical composition, of course, did not arise in a vacuum, but as part of a continuum with its cultural environment, and initially takes over and continues the metres that were current in the Buddha's day ⁵. It is essential therefore, if we wish to understand Pàëi prosody that we have some idea of what these metres looked like in the Vedas, and how that relates to the Pàëi period. My prime reference for information on the Vedic metres has been:

b) Sarabhatti, or partial vowels, are vowels that have been written, but which have to be ignored when determining the metre of the text. When the sarabhatti vowel is ignored it means that there is then a conjunct consonant to be taken into consideration. For example, the word Araha looks like it should be scanned as ÛÛÛ , but when we take into account the sarabhatti vowel we find that it needs to be scanned as ÜÛ . In the text as presented here I have counted sarabhatti vowels in the following words:

d) In another work ⁸ I have stated that there is an hitherto unrecognised rule in Pàëi metrical composition, which says that when syllables are resolved it is normally the first 2 syllables of a word that are concerned in the resolution ⁹. In the text I have marked what I consider to be unambiguous cases of this rule by underlining in the metrical markings the syllables concerned, they can be seen in the following places, with resolution at the 1st syllable: 984a, 990a, 995cd, 1011b, 1014d, 1015a, 1016a, 1047c, 1050c, 1065a, 1081f, 1086a, 1092f, 1097a, 1102a, 1110c.

at the 3rd: 1003d, 1116c; 4th: 1084c, 1135c; 5th: 995d, 1131a; 6th: 1023a, 1130c; 7th: 1002a, 1149d.

1) The scansion of the line Avijjàya nivuto loko at 1033a looks like this ÛÜÜÛÛÛÜÜÜ . The rule of resolution would say that if there is resolution here it is at the 5th (i.e. nivuto). If that is so then it leaves the metre slightly wrong, as there is no vipulà showing the necessary pattern (ÅÜÜÛÜÜÜÅ), though it is only a little different from the mavipulà, having a light 4th syllable instead of a light 3rd. In his notes in GD II (pg 367), Norman puts the resolution at the 4th, if that is so then what we have is an acceptable Siloka pathyà variation. The line however is in reply to the question Kenassu nivuto loko? It may be, therefore, that the line here should simply be regarded as containing 9 syllables, a situation which has arisen because this is an answer that has been given in imitation of the question line.

2) At line 1040e we find the following: Kaü bråsi Mahàpuriso ti? ¹⁰ The scansion reads ÜÜÛÛÜÛÛÜÛ . According to the rule the resolution must be at the 6th in puriso, but again that leaves the metre slightly wrong (ÜÜÛÛÜÜÜÛ), as the mavipulà, which it closely resembles normally has a heavy 4th syllable. Norman (GD II, pg 368) counts the syllable as resolved at the 3rd, which would give an acceptable savipulà.

There is another line which involve resolution, which leave the line irregular at 1140a. In this case, however, there is nothing that goes against the rule as such, but rather even with its aid we are still left with a slightly irregular line, in this line the resolution cannot be at any other syllable. There are a couple of other ambiguous cases, at 1122c, & 1146c. But again we are not dealing with alternative explanations. Either there is resolution, or the metre is incorrect. If there is resolution in these cases we have to count a normally long syllable as short in order to make the resolution, by reading na t›yhaü ¹¹, & Õvam respectively.

Given that there are 27 unambiguous cases ¹² where we can count the rule of resolution as upheld; and a further 4 where resolution has probably taken place (and if so, it has taken place in line with the rule); and only 2 (or 3, if we count the virtual repetition of 1040e at 1042c), which may give slightly better interpretations if we go against the rule, I take it that the rule can be considered well founded on the evidence that has been presented here ¹³. The rule is of quite some importance it should be noted, as when it is taken into consideration it can help to identify correctly the variation that is being used in Siloka lines, and the underlying gaõa structure in the bar metres.

There are two metres used in Pàràyana, which are in fact the main metres used in Pàëi prosody. They are the Siloka and the Tuññhubha ¹⁴, we will examine them in this order.

Now it is very interesting from an historical perspective that in Hemakamàõavapucchà the metre very much looks like an early form of the Anuññhubha. Of the 18 lines which make up this section no fewer than 14 show the Anuññhubha structure. Two further lines are pathyà (the normal form of the prior line in the Siloka verses), which is also the most normal variation in the Vedas ¹⁵.

Of the openings 5 show the normal Vedic form ÅÜÅÜ (1084ce, 1085bd, 1086d), and the others show syncopated forms that are also common in the Vedas (1084bd, 1085c, 1086bcd, 1087abc ¹⁶). So that it appears that all the evidence would suggest that we should probably count this Pucchà as being in Anuññhubha metre.

In the Vedas after some time variations from this basic pattern started to emerge, which eventually gave rise to a new addhasamavutta metre, the Siloka. In canonical Pàëi the metre is normally described ¹⁷ as having a pathyà structure, and 7 variations. They are:

In the Pucchà there are 200 lines in the Siloka metre, of which I count 102 as prior lines. The distribution of the variations in the prior line as presented in the text that follows is like this ¹⁸:

As can be seen from this by far the most important variation is the Anuññhubha, which constitutes some 15% of the total ¹⁹. The second most important variation is savipulà, with some 5% of the total ²⁰.

These figures are of importance because in the later development of the metre in the canon the Anuññhubha and savipulà are normally avoided in the Siloka prior lines ²¹, which then sees a corresponding increase in the occurrence of the pathyà, and the first 4 variations. With the help of these characteristics it is possible to determine whether verses belong to the early or to the middle and late periods.

When we examine the Vatthugàthà, which on linguistic and doctrinal evidence can be shown to be late in composition ²², we can see that there has been an evident shift in the way the metre is composed. In this section there are 222 Siloka lines, exactly half of which are prior lines. The statistics for the variations are as follows:

When we compare the Siloka lines in the epilogue we can see that they clearly belong to the early period. Of the 41 prior lines in that section, 5 are Anuññhubha, 1135ce ²³, 1137c, 1139c, 1141c. The other statistics for this section are as follows:

In the Pucchà and epilogue of Pàràyanavagga there are 228 lines in Tuññhubha metre ²⁴. When we compare these with the Vedic models we will find that there are some interesting continuities and also discontinuities.

As we can see from the description given above in the Vedic period there are 2 main forms of the opening, they are ÅÜÛÜ, & ÅÜÜÜ ²⁵. In the Vedas these openings are roughly equal in number. Both of these openings also occur in Pàràyanavagga. The first is by far the most common form, but the second, though becoming rare ²⁶, is found on a significant number of occasions, see 1045e, 1052b, 1070a, 1071b, 1072b, 1073a, 1082a, 1096d, 1120c, 1149d ²⁷. As this is continuous with the Vedic form of the metre, it seems that we should accept it as a genuine form of the metre. In the early period therefore, readings that meet the requirements of this opening should be regarded as legitimate, and when establishing texts there is no pressing need to take alternative readings that have almost certainly been introduced by scribes who were seeking to regularize the metre according to classical norms.

There is a third form of the opening that occurs in the Pucchà, which shows the following pattern ÜÛÛÜ , these can be found at the following lines: 1050c, 1082b, 1096c, 1097e, 1120e, 1122f; see also 1056b, 1104b.

Of the instances that are listed here 4 concern the compound jàtijaràya (ÜÛÛÜÐÛ), one other has the compound santipadaü (ÜÛÛÜ); and another has the reading upadhinidànà (ÛÛÛÛÜÜ); in these cases it would not be difficult to rectify the metre by reading jàtã, santã, & upadhã, it is therefore perhaps of some significance that no such reading is found in any of the editions. The other 2 readings open with the word bhikkhu, where to read bhikkhå, would violate the context, producing a plural where a singular is needed.

In the Vedic period this opening is regarded as irregular ²⁸. But in the Pàëi verses it appears fairly frequently, and shows a definite form, always beginning with a long syllable, so that I feel that we have to regard it as a genuine syncopated variation of the opening in the Pàëi period.

The most important difference in the early Pàëi period is the dominance of the bhagaõa break ÜÛÛ which in fact becomes the norm in the classical period. In the tables that follow it will be seen that it is this characteristic, and not the position of the caesura, that distinguishes the metre (only the regular breaks are listed in this first table, breaks that are considered to be irregular in form will be listed later).

 caesura at the 4th:

,ÜÛÛ (57 lines, constituting 24% of the total)

,ÛÛÜ (21 lines, 10% of the total)

,ÜÛÜ (19 lines, 8% of the total)

,ÛÛÛ (8 lines, 3.5% of the total)

,ÜÜÛ (4 lines, 2% of the total)

Û,ÛÛ (13 lines, 6% of the total)

Ü,ÛÜ (6 lines, 3% of the total)

Ü,ÛÛ (34 lines, 15% of the total)

Ü,ÜÛ (1 line)

ÜÛ,Û (40 lines, 18% of the total)

The irregular lines are also of some interest in helping to define the parametres of the prosody. The first break listed below is particularly interesting, as a long 6th has been normally considered to be wrong in terms of the metre. As there seems to be a regular form to this break ²⁹, with the caesura after the long 6th, followed by a short syllable, it appears that the break should be regarded as acceptable:

ÜÜ,Û (8 lines, 4% of the total)

1048a, 1049d, 1050d, 1055d, 1073bd, 1077c, 1103c ³⁰

Û,Ü 1068d,

Ü,Ü 1064c, 1068c, 1076b, 1103d

Ü,ÜÛÛ 1044d,

Û,ÜÛÛ 1047c, 1081f,

Û,ÛÛÜ 1120a, 1123a

irregular ³¹:

It is characteristic of the early Vedic period that the mixing of Tuññhubha and Jagatã lines was normally avoided. Of the 228 lines counted above, only 16 are in the Jagatã metre, which therefore constitute less than 7% of the total, this shows that although by the early Pàëi period Jagatã lines were allowed in what were otherwise Tuññhubha verses, nevertheless they were not part of the normal parametres of the prosody ³², and seem to have been allowed only as an expedient.

with the syncopated opening ÜÛÛÜ sometimes appearing, and the break ÜÜ,Û occasionally giving a long 6th syllable. It is on the basis of this description that the present text has been established. ³³

1 I do not propose to discuss these differences in readings here, which do not affect the metre, being mainly of doctrinal importance. I have chosen to establish the text from the Cullaniddesa edition because in almost every case Niddesa's readings are, in fact, to be preferred to Suttanipàta's, and they probably represent an earlier strata of the text. This may be because once the text was embedded in its commentary, which must have been at an early date, that it stopped developing, whereas the Suttanipàta text lacked the restrictions in regard to its readings that a commentary imposes.

2 All scholars are agreed that this text contains some of the most ancient teachings of the Buddha; see e.g. Rhys Davids, Buddhist India (London 1903), pg 122; G. C. Pande, Studies in the Origins of Buddhism (2nd rev. ed. Delhi, 1974), pg 51ff; Abeyanayake, A Textual and Historical Analysis of the Khuddaka Nikàya (Colombo, 1984), pg 75ff, etc.

3 Although in this paper I am mainly concerned with the metre as it appears in Pàràyanavagga, I have given cross references to Aññhakavagga (as it appears in the PTS edition of Mahàniddesa) when they can help to confirm the suggestions made herein.

4 It should be noted that Cullaniddesa does not comment on the Vatthugàthà, but they are nevertheless included at the opening in all editions of CNidd.

5 Later there is a great innovation in Indian prosody with the emergence of the so called Musical metres (mattàchandas and gaõacchandas) which brought new vigour into Indian verse composition. In literary terms these seem to have arisen first in Pàëi, but in the first and earliest period of Pàëi prosody, which is what we are concerned with here, they are as yet unknown.

6 Although the text which follows is of Pàràyanavagga as it appears in Cullaniddesa, I have given references to the verse numbers as they appear in the European edition of Sn to facilitate comparison with PJ II & GD II.

7 Please note that in this paper and in the text that follows sarabhatti vowels are normally written in superscript as here.

8 See my Outline of the Metres in the Pàëi Canon.

9 This should not be taken as affirming that it is always the initial syllables that are affected, as the word in which resolution takes place may be the second member in a compound, or preceded by one or more prefixes, and quite often is itself a prefix. To give an example, in the word pañivasati, according to the rule we may take the 1st & 2nd syllable as resolved, or the 3rd & 4th; however, there cannot be resolution of the 2nd & 3rd, or the 4th & 5th. There seems to be a small exception to the general rule, in that the negative particle na occasionally appears to form the first half of a resolved syllable, perhaps because there was felt to be a close semantic affinity between the negative and the word it modifies.

10 The line is virtually repeated in the answer at 1042c, exchanging Taü for Kaü, which makes no difference metrically.

11 Tuyhaü is normally read as a long syllable, cf. 1030a & 1061b

12 In Aññhakavagga I count the following as unambiguous: At the 1st: 774b, 783b, 787a, 791a, 796a, 827c, 830d, 831b, 832b, 862b, 863b, 869c, 870c, 876d, 888c, 889c, 890c, 893c, 895b, 900d, 901b, 907d, 908d, 909d, 963a, 965a, 969cd, 970d, 971d, 973ad; 3rd: 967a; 4th: 823a, 938d; 5th: 790d; 6th: 823c, 959a; 7th: 841a; to give jagaõa ÛÜÛ in the Old Gãti verses: 920d, 922d, 923b x 2, 924d, 925b, 929b x 2 (total: 47 instances).

13 No metrical rule is upheld 100%, of course, but we may take it that there is a rule if it is normally adhered to. I may add that as far as I have been able to discern the rule holds even better in medieval metrical compositions, but I am unable to bring the evidence forward at this point. There is a compliment to this rule, which I call the rule of replacement. This states that when replacement takes place it always does so after a caesura (cf. 1064c, 1068cd, 1076b, 1103d in the text). This shows that the presumption is that the first two syllables of a word are the ones that are being replaced. This further helps to confirm the rule, of course.

14 That is, the metres that are otherwise known as øloka and Triùñubh. In this paper, which examines the distinctive characteristics of early Pàëi metrical composition, I have preferred to use the Pàëi names for the metres. Although, as I will show herein, these are clearly related to the Sanskrit metres, they nevertheless have to be carefully distinguished from the latter, otherwise there is a danger of forcing Pàëi forms into classical Sanskrit models to which they do not, in fact, adhere.

15 The other two lines (1084f & 1086a) as they stand are unclear. If we read abhãramiü in the first of these lines (with Smith PJ II pg 660), that would once again give an Anuññhubha line. If we correct the second line by excluding -a vi¤¤àta- (with Norman GD II, pg 378) that would give savipulà.

16 See Arnold's charts on pg 153 of VM.

17 See Warder PM, pg 172ff. Warder refers to this metre as Vatta presumably on the basis of Vuttodaya, and in my Outline of the Metres in the Pàëi Canon I followed him. However it should be noted that the Vatta metre is described in Vuttodaya as having the same cadence in all 4 lines. Only the Pathyàvatta has alternating cadences.

18 It should be noted that lines can sometimes be scanned in different ways, producing different variations. On a small number of occasions I have had to make a choice as to which variation we are dealing with and assign the line accordingly.

19 This figure includes the lines in Hemakamàõavapucchà. If we count these as belonging to the Anuññhubha metre, and exclude them from the Siloka statistics, then the Anuññhubha variation would amount to 6% less. Although this has a bearing on what follows, the overall position is not affected one way or the other.

20 According to my calculations in Aññhakavagga Anuññhubha constitutes no fewer than 20% of the total (out of Siloka 116 prior lines): 781a, 814c, 815c, 850c, 851a, 853c, 854a, 855a, 856c, 859ac, 860a, 861c, 938c, 941c, 945c, 947a, 949c, 950c, 953a, 954c, 958a, 960c. There however savipulà is negligible (one case at 940b); while mavipulà takes 7% of the total 817c, 818c, 858a, 937a, 938c, 939c, 947c, 952c.

21 They do not seem to be ever totally excluded. Note that also the 6th (if it is a variation) is also normally avoided in the later works.

22 Good sources for an examination of the linguistic and doctrinal evidence are N. A. Jayawickràma, A Critical Analysis of the Sutta Nipàta, which was reprinted in the Pàli Buddhist Review 1, 3 (1976), and the same author's The Sutta Nipàta: Pucchàs of the Pàràyana Vagga, published in the University of Ceylon Review (photocopy only available to the present writer). See also Norman's extensive discussion of the linguistic materials in GD II.

23 This is counting 1135e as a prior line Anuññhubha, but it may be an even line, in which case we would have to leave it out of the reckoning. In any case the overall figures are not affected.

24 This total includes 15 lines that are written in the Jagatã metre (in the lists that follow Jagatã lines are placed in brackets). As the variations that occur in the one also occur in the other, I take them together here. The relationship of the two metres will be considered in what follows.

25 See Arnold's charts of pgs 188 & 194 of VM.

26 Eventually in the middle and late Pàëi periods it is normally avoided.

27 This opening is also found in Aññhakavagga, cf. 795d, 869c, 870d, 873bc, 875b, 878c, 881c, 889a, 884c, 894b, 908d, 966b, 971b, which lends good support to the argument as that section is also considered to be old.

28 Cf. Arnold VM pg 194, where its occurrence is listed at 2%. However it appears quite often in Aññhakavagga, cf. 836c, 842c, 845a, 870a, 874b, 902a, 964b, 970a, 975b. As in Pàràyanavagga the form always opens with a long syllable.

29 Unfortunately Warder's tables on pg 207 & 209 of PM do not bring this out clearly. Arnold's tables do not record caesuras at the 6th, so that it has not been possible for me to check it in the Vedic period. It appears that many scholars have not countenanced a caesura at the 6th when making their analyses. But it seems to the present writer that in reciting verse in Pàëi it often falls quite naturally after the 6th. According to the above analysis some 24% of the breaks can be counted as producing this caesura.

30 See also 1056b, 1101a, & 1133ad, where the same pattern turns up after a word-break at the 4^th. The comments in the text apply to this form also. To this we can add the following references in Aññhakavagga: 776d, 93d, 799a, 802d, 846d, 862d, 864a, 866a, 871a, 894b, 901d, 964cd, 969a. The same break, but with caesura at the 4^th: 829d, 887c, 894a, 913ad. Also cf. 913c, 972d for the same break but with a different caesura.

31 If we follow the suggestions given in the notes to the verses in the text some of these lines are not so irregular, but as there is some ambiguity about them, I have counted them seperately here.

32 On the other hand, in the middle and late Pàëi periods mixing is normal.

33 It will be noticed from the variant readings that the Burmese editors had in mind another, more classical, model, that seeks to avoid the long 3rd and 6th. It has long been recognized that the Burmese editions have often been `corrected' to make the metre fit into what is, in fact, a later prosody; see e.g. Helmer Smith's remark in PJ II, pg 637. Judging by Norman's comments in GD II, he was also working with this more classical model of the metre.