Sāptapadīnaṃ Sakhyam || A. 5.2.22 ||

1. Introduction

1There is a lot of metrical variety among the Tamil language specimens that have been handed down to us by history, but the shortest way to mastering that complexity is not to confidently read Tamil metrical treatises, as if they contained the whole truth concerning Tamil literature, because those treatises are themselves incomplete attempts at capturing the ideal forms of Tamil literature which were meant as a help for those who had the desire to create, preserve, recite and/or understand “poetical compositions” (ceyyuḷ).1 Additionally, they often contradict each other, having been composed in different circumstances, both from the linguistic point of view (the languages which they describe are not completely identical) and from the point of view of the literature which they characterise (the poetical genres which they describe are not the same). This is a difficulty for the student, who will find it hard to find the equivalent of a compass for navigating this maze, if he (or she) has to read several treatises,2 but an opportunity for the historian, because, given the fact that for most technical texts we have very little external objective evidence for determining where they stand in history (i.e. no original datable manuscripts), the study of “contradictions” is often the only method we have for reconstructing the internal chronology of those fields of knowledge.3

2For instance, the Yāpparuṅkalak Kārikai (henceforth YK), a short treatise which is possibly datable to the 10^th century through epigraphy,4 and the Yāpparuṅkalam (YA), a probably slightly earlier treatise, of which the YK was probably an abridged version5 tell us that Tamil poetical compositions are complex combinations of several levels of metrical entities, among which the elemental (or primary) ones are called eḻutt^u,6 acai, cīr, taḷai (or pantam), aṭi and toṭai,7 and the resulting ones, obtainable, as final output, with the primary ones as “matière première”, are called pā and pāviṉam. Although the names of most of those categories seem identical to technical terms found in the Ceyyuḷiyal of the Tolkāppiyam (TP-Cey), a much earlier treatise, a close examination of the definitions (or characterisations) given for those terms in the YA/YK, on the one hand, and the TP-Cey, on the other hand, reveals that there are profound differences between them, although the commentaries on those treatises seem to navigate with some degree of ease between those contradictions. Therefore, if one wants to understand their history and their relationships, (real) literature and technical literature8 have to be studied, in an agnostic way, partly independently from each other, although one must ask oneself, all the time, how they fit together. Because, however, time is always limited and we cannot deal simultaneously with every aspect, the present article will concentrate on one specific example, which illustrates unequivocally, in a striking way, the contradictions I have evoked. I draw this example from technical literature, trying however not to forget the question of its applicability to (or of its relevance for) the intepretation of (real) literature.9 The example chosen centres around the number “six hundred and twenty-five”, which is declared by the TP-Cey (in TP357i) to be the number of possible templates for certain categories of metrical lines. Interestingly that count is arrived at, as we shall see, in two different ways by the three commentators10 of the TP-Cey, and it is also discussed (and obtained in a third manner!) in the final “fourre-tout” section of the Yāpparuṅkala Virutti (YV), which is the commentary on the YA. Very briefly, we can say that:

• For Iḷampūraṇar, 625 = 5 × 5 × 5 × 5 (see section 2)
• For the YA commentator, 625 = 261 + 232 + 132 (see section 12)
• For Pērāciriyar and Nacciṉārkkiṉiyar, 625 = 324 + 181 + 12011 (see section 14)

3The lack of agreement between the three is clear. However, before I am able to present those very technical explanations in more detail, in such a way that they be understandable and appear (hopefully) relevant, I shall first of all have to introduce a number of basic elements, and this is what will be done in several opening sections. A number of difficulties can however be expected, which are due to the fact that understanding the historical relationships between successive theories (theory A, theory B, theory C, etc.), which share some technical terms, is more difficult than understanding one single theory, which can be presented in the logical (or pedagogical) order. One difficult technical term in this respect, which is central to the topic examined in the article, is the term taḷai, better leftuntranslated at this preliminary stage. One of the conclusions of this study (see section 17) is that we must distinguish between conceptions such as taḷai ^A, taḷai ^B, taḷai ^C, etc.12 But, in such a situation, the needs of pedagogy can enter into conflict with the logical order for historical investigation, which requires parsimony, because a reader familiar with taḷai B, which is in a sense the dominant theory (being expounded in the YA and the YK), does not have the same needs as a reader to whom it must first be presented. For that reason, taḷai will be evoked in a recurrent manner, although perhaps not sufficiently clearly, before being discussed again at the end of the article. Similar problems are met with frequent technical terms such as āciriyam, vañci, veṇpā and kalippā, which can be very roughly defined as the names of the main metres, although it should be made clear from the onset that the manner in which they have been understood has varied in the course of history.

2. Prologue

4While expressing his views on the Classical Tamil poetry current in his time, the author of the Ceyyuḷiyal (TP-Cey), which is today for us the penultimate chapter in the Poruḷatikāram (TP), the third book of the Tolkāppiyam (T),13 has bequeathed to posterity several enigmatic cūttiram-s, the precise interpretation of which has been a matter of dispute between medieval metricians. Among those cūttiram-s, we can count the one which is variously referred to as TP357i,14 TP362p or TP-Cey50n, depending on whether we try to read and understand it with the help of the commentaries composed by Iḷampūraṇar (see sections 2 and 3), or by Pērāciriyar and Nacciṉārkkiṉiyar (see sections 14 and 15),15 and which we can also examine, with a slightly variant text, as a citation inside another treatise (i.e. the YV, see sections 11 and 12). That cūttiram and three possible (but incompatible) translations for it are:

(1) aivakai yaṭiyum virikkuṅ kālai
meyvakai yamainta patiṉēḻ nilattu16
meḻupat^u vakaiyiṉ17 vaḻuvila vāki
yaṟunūṟ ṟirupat taintā kummē (TP357i/TP362p/TP-Cey50n)

(2a) if one expands [the enumeration of] the five-typed “metrical lines” (aṭi), they become six hundred and twenty-five (subsub-types), being of the unescapable (vaḻuvila) seventy- varieties, inside the seventeen [sub-types, loci] (nilam) established on the basis of the [syllabic elements (eḻuttu-s)] which are the body (mey) [of those metrical lines] (Translation 1, geared towards Iḷampūraṇar, see section 2 and section 3).

(2b) if one expands [the enumeration of] the five-typed “metrical lines” (aṭi), they become six hundred and twenty-five (sub-subtypes), [on the condition of] being without the seventy-varietied faults (vaḻu), inside the seventeen [sub-types, loci] (nilam) established on the basis of the [syllabic elements (eḻutt^u-s)] which are the body (mey) [of those metrical lines] (Translation 2, geared towards the YV commentator, see sections 11 and 12).

(2c) if one expands [the enumeration of] the five-typed “metrical lines” (aṭi), they become six hundred and twenty-five (sub-subtypes), [on the condition of] being with the seventy fault-less (vaḻu-v-ila) [cīrs] inside the seventeen [sub-types, loci] (nilam) established on the basis of the [syllabic elements (eḻutt^u-s)] which are the body (mey) [of those metrical lines] (Translation 3, geared towards Pērāciriyar’s interpretation, See sections 14 and 15).

5As can be seen, several technical terms (such as aṭi, mey, nilam, vaḻu and eḻutt^u) are first of all to be explained, if the three translations are to be reasonably informative to an outsider, and the immediate context in which this cūttiram appears should also be clarified. It can be said, as a preliminary explanation, that the cūttiram TP357i which is here examined is part of a group containing forty-three cūttiram-s,18 all devoted to the various types of aṭi “metrical line”, which I shall characterise here as the Level-4 metrical units.19 That group is preceded inside the TP-Cey by two shorter sections devoted to the various types of acai (Level-2 metrical units) and to the various types of cīr “metrical foot” (Level-3 metrical units). Those two sections contain, respectively, eight and twenty cūttirams, and are themselves preceded by two cūttirams, one (TP310i) being a table of content of the TP-Cey in which acai, cīr and aṭi are the third, fourth and fifth element, and the other (TP311i) dealing with the first two elements in the table of content,20 namely māttirai “measurement, duration” and eḻutt^u (Level-1 metrical units),21 the underlying logic in this progression from Level-1 units (i.e. eḻutt^u) to Level-4 units (i.e. aṭi) being that:

• An aṭi “metrical line” (literally “step, foot”) normally contains four cīr-s22 (but there are other, exceptional, cases)
• A cīr “metrical foot” is a combination of two or three acai-s (but may contain only one)
• An acai is a combination of several eḻutt^u-s (but sometimes contains only one)

6Coming back to (1) and to its three possible translations, (2a), (2b) and (2c), I shall first point out that among the four figures which appear in them, namely (a) “five [varieties]” (ai-[vakai]), (b) “[in the] seventeen [loci]” (patiṉēḻ [nilattum]), (c) “six hundred and twenty-five” (aṟunūṟṟirupattaint^u) and (d) “seventy” (eḻupat^u), the first two, (a) “five” and (b) “seventeen”, pertain to the possible types and sub-types of metrical lines, and are already implicitly known, whereas the third figure (c) is a count of the sub-sub-types of lines, conditioned by a count (expressed by the figure (d) “seventy”) either of the unescapable patterns (as in translation 2a) or of the faulty patterns to be avoided (as in translation 2b), so that both (c) and (d) represent new information. The first two figures (a & b) are already known because they refer to a classification scheme where Level-4 units have been given additional names23 based on the number of “countable” Level-1 units which they contain. That classification has been done from TP344i to TP348i (alias TP348p–TP352p)24 and is summarised in the following chart:

7In this chart, the names appearing in the left column of the chart are conventional designations and the figures appearing in the right column are the criteria for the designation. A metrical line is baptised neṭilaṭi “long line” if it contains between 15 and 17 “countable eḻutt^us”, this expression referring to the elements in which a consonant is combined with a vowel, long or short (see section 4), because the specifier “countable” (as stated in TP351i)25 excludes consonants which are not combined with a vowel, as well as consonants which are combined with an over-short vowel. That being the context at the time when TP357i is enunciated, what is the new information which it brings, when the figures (c) “six hundred and twenty-five” and (d) “seventy” are uttered?26 As we shall see in the course of this presentation, different commentators seem to have had very different opinions concerning the matter, and an elucidation of their answers requires, as a preliminary, a presentation of additional elements concerning the classification of cīrs and of acais (see section 5) as well as a presentation of what the over-short u (kuṟṟiyalukaram) was (see section 9).

3. How Iḷampūraṇar explains “six hundred and twenty-five”

8I shall now first of all examine Iḷampūraṇar’s commentary on TP357i. In the course of this examination, new technical terms will appear which I shall explain only in ulterior sections.27 The immediately most important among them are those pertaining to the several categories of cīr. Iḷampūraṇar starts his exposition (under TP357i) by referring to a count of eighty-four possible cīrs which has already been provided by him while commenting on TP339i. The figure “eighty-four” (eṇpattunāṉku) obtains as a grand total for the three sub-totals (16, 64 and 4) associated with each of the three groups appearing in the left-side column inside the chart which follows:

9The elements appearing in the second or in the third column of this chart are designations which have been introduced in cūttirams for which the reference is given in the fifth column.28 Finally the fourth column contains abbreviations (or concatenations of abbreviations), each abbreviation referring to one of the four possible acai-s enumerated by the TP-Cey (in TP312i and TP313i) and iconically29 named nēr (= E), nirai (= I), nērp^u (= EP) and niraip^u (= IP). As an example, the abbreviation EP-E refers to the cīr that is referred to as “nērp^u-nēr” by commentators.30 I shall not however at this stage explain what the four acais are31 because our immediate goal is to explain how Iḷampūraṇar arrives at the figures “seventy” and “six hundred and twenty-five”, and because his explanation is based on the fact that, according to the TP-Cey, there are five main types of cīrs, the names of which appear in columns 2 and 3 as: (a) iyaṟcīr,32 (b) āciriyavuriccīr, (c) veṇpāvuriccīr, (d) vañciyuriccīr and (e) ōracaiccīr.33 On the basis of those five designations, the figure for which Iḷampūraṇar’s explanation is the shortest is “six hundred and twenty-five”, and I shall start with it. His opinion is that this figure simply represents, using the language of combinatorics anachronistically, the possible sequences of four items (in this case cīrs), selected inside the set {a, b, c, d, e}. This task amounts, expressed in modern vocabulary, to computing the number of functions from the set {1, 2, 3, 4} into the set {a, b, c, d, e}, which is 5⁴ = 625. In Iḷampūraṇar’s own words, this is expressed thus:34

(3a) iṉi aṭi aṟunūṟṟirupattaintāmāṟu: acaiccīr iyaṟcīr āciriyavuriccīr veṇcīr vañciyuriccīr eṉṉum aintaṉaiyum niṟutti ivvaintu cīrum varuñcīrāka vuṟaḻumvaḻi irupattaintu vikaṟpamām. avvirupattaintaṉ kaṇṇum mūṉṟāvatu aintu cīraiyum uṟaḻa nūṟṟirupattaintu vikaṟpamākum. an nūṟṟirupattaintaṉ kaṇṇum nāṉkāvatu aintu cīraiyum uṟaḻa aṟunūṟṟirupattaintām eṉṟavāṟu.

(3b) “[Let us] now [present] the method for having the six hundred and twenty-five metrical lines. Install (niṟuttutal “cause to stand”) [as “standing cīr”] the five which are called acaiccīr, iyaṟcīr, āciriyavuriccīr, veṇcīr and vañciyuriccīr; when these same cīrs alternate (uṟaḻtal) as “coming cīr-s” (varuñ cīr) [after the “standing cīr”], they become (ākutal) twenty-five alternatives (vikaṟpam); when the five [cīrs] alternate as third [cīr] at the [right side of] the twenty-five [possible combinations of two cīrs], they become hundred and twenty-five alternatives; when the five [cīrs] alternate as fourth [cīr] at the [right side of] the hundred and twenty-five [possible combinations of three cīrs], they become six hundred and twenty-five alternatives;” (from Iḷampūraṇar’s commentary on TP357i).

4. How Iḷampūraṇar explains “seventy”

10I shall now turn to an examination of the way Iḷampūraṇar justifies the count of “seventy” which appears in TP357i. The first point to be mentioned is the word-for-word gloss he provides for the third line in TP357i, in the following way:

(4a) eḻupat^u vakaiyiṉ vaḻuvila vāki eṉpat ^u — eḻupat^u vakaippaṭṭa uṟaḻcciyiṉ vaḻuvutal iṉṟi eṉṟavāṟ^u.
(4b) “The passage eḻupat^u vakaiyiṉ vaḻuvila vāki inside the cūttiram means ‘without deviating from the various possibilities which are subdivided into seventy’”
(5a) [eḻupat^u] vakaiyāvat^u — iraṇṭ^u cīr tammuṭ puṇarum puṇarcci eḻupat^u vakaiyām eṉṟavāṟ^u.
(5b) [The use of] theexpression [eḻupat^u] vakai ([seventy] subdivisions) is a way of saying that ‘the union (puṇarcci) by which two cīrs unite (puṇarum) has seventy subdivisions (vakai)’”

11After stating this, Iḷampūraṇar points out the total count of eighty-four possible cīrs that has already been given (see Chart 2) and introduces seven types of lines that he names after the five types of cīrs I have discussed above, using some additional criteria which pertain to sequences of two cīrs and which I shall explain on the basis of the following figure:

12The logic which is at play is that if in a sequence of two cīrs, S and C, we categorise S and C only on the basis of the five possible types enumerated in 3a and 3b (and summarised in Chart 2), we have only twenty-five possible sequence types. However, if we add two additional parameters which are the type of f (i.e. the nature of the final acai inside S) and the type of b (i.e. the nature of the acai which stands at the beginning, inside C), we have a much greater number of possibilities. Using the abbreviations introduced in the previous section (namely: nēr [= E], nirai [= I], nērp^u [= EP] and niraip^u [= IP]), we can say that the possible sets of alternative envisioned by Iḷampūraṇar are the following:

• S1: “f = E” (when “S = iyaṟcīr”).
• S2: “f = I” (when “S = iyaṟcīr”)
• S3: “f = EP” (when “S = āciriyavuriccīr”)
• S4: “f = IP” (when “S = āciriyavuriccīr”)
• S5: “f = E” (when “S = veṇcīr”)
• S6: “f = I” (when “S = vañciyuriccīr”)
• S7: “f = EP” OR “f = IP” (when “S = vañciyuriccīr”)
• S8: “S = acaiccīr”
• C1: “b = E” (while C = iyaṟcīr), OR “b = EP” (while C = āciriyavuriccīr), OR “b = E” (while C = veṇpāvuriccīr), OR “b = E” (while C = āciriyavuriccīr), OR “b = E” (while C = ōracaiccīr).
• C2: “b = I” (while C = iyaṟcīr), OR “b = IP” (while C = āciriyavuriccīr), OR “b = I” (while C = veṇpāvuriccīr), OR “b = I” (while C = āciriyavuriccīr), OR “b = I” (while C = ōracaiccīr).

13On the basis of such a categorisation, and because it would be premature to give more details when some of the basics have not yet been explained, I shall now content myself with stating that Iḷampūraṇar arrives at a total of seventy by considering seven situations, each of which is subdivided by him into ten cases. They are:

• Union 1: S1 combined with C1 (5 cases) OR S2 combined with C2 (5 cases)
• Union 2: S3 combined with C1 (5 cases) OR S4 combined with C2 (5 cases)
• Union 3: S1 combined with C2 (5 cases) OR S2 combined with C1 (5 cases)
• Union 4: S5 combined with C1 (5 cases) OR S5 combined with C2 (5 cases)
• Union 5: S6 combined with C2 (5 cases) OR S6 combined with C1 (5 cases)
• Union 6: S7 combined with C1 or with C2 (10 cases)35
• Union 7: S8 combined with C1 or with C2 (10 cases)36

14After this enumeration, Iḷampūraṇar undertakes to compare such a classificatory system (i.e. his own interpretation of TP357i) with the one adopted by later metricians, such as the authors of Yāpparuṅkalam (YA) and Yāpparuṅkalak kārikai (YK), in which the relationship between two successive cīr-s, which is called taḷai “linkage” is said to always fall under one of seven possible values, which he enumerates and to which we shall come back, once we have presented, in the following sections, some elements concerning the chronology of theories and some elements concerning the important item called kuṟṟiyalukaram (see section 9).

5. The sounds of Classical Tamil and the formation of acai according to the Tolkāppiyam

15I shall now present in more detail the elements which lie at the basis of the items examined in the previous sections. The items to be examined here are eḻutt^u and acai. In a previous section, I have defined the eḻutt^us as the Level-1 elements for metrics, on the basis of which the Level-2 elements, called acai, are formed. I have also alluded there to the fact that Tamil terminology is ambiguous and that the meaning of eḻutt^u, as “Level-1 (metrical) units”, used in the TP-Cey is only one of the two meanings of eḻutt^u, the other one being “Level-0 (phonological) units”, on the basis of which the Level-1 (metrical) units are formed.37 This meaning (i.e. “Level-0 units”), which we may consider as more primary,38 is used in the TE,39 when the phoneme inventory of Classical Tamil is introduced or referred to. That inventory is shown in chart 3, in which the thirty-three (Level-0) eḻutt^u-s of Tamil (following TE1i)40 are mentioned. They are:

16The Level-0 elements contained in Chart 3, combine into the following sequences (delimited by “{“and”}”), which are to be considered as Level-1 metrical elements:

• {V} and {V̅}41 (found only in word-initial position)42
• {CV̅} and {CV}43
• {C} (such items are called “oṟṟu”)44
• {P u}45 (such items are found in word-final position, with additional constraints)46
• {P i}47
• {ḵ} (this item occurs only in very specific environments but a detailed account is not possible here)

17Among those, both {V̅} and {CV̅} can be targeted by the first occurrence of the expression “neṭil” found in TE312i,48 and I shall represent the joint possibility by the symbol {(C)V̅}. Similarly, the expression “kuṟil”, as used in the same cūttiram, can target both {V} and {CV} and I shall represent that possibility by {(C)V}. Using such a notation, we are now in a position to give the patterns for the four types of acai-s which are defined by TP312i and TP313i. They are:

• {(C)V} (referred to as “kuṟil” in line 1 of TP312i)
• {(C)V̅} (referred to as “neṭil” in line 1 of TP312i)
• {(C)V}{CV} (referred to as “kuṟiliṇai” in line 1 of TP312i)
• {(C)V}{CV̅} (referred to as “kuṟiṉeṭil” in line 1 of TP312i)

18The first two among these four are assigned to the pattern nēr and the last two are attributed to the pattern nirai, because of the distributive predicative relationship which maps line 1 towards line 3 inside TP312i. To those four items must be added four items, which occur thanks to the second line, in which an optional {C} (i.e. an “oṟṟ^u”) is added as a tail to the eḻutt^us already present in the pattern.49 The result is the following four cases, two of them illustrating nēr and two illustrating nirai:

• {(C)V}{C}
• {(C)V̅}{C}
• {(C)V}{CV}{C}
• {(C)V}{CV̅}{C}

19We end up thus with four formulas for a nēr and four formulas for a nirai.

20A word of warning is however necessary at this stage, which is a consequence of a feature of the sūtra style, where no statement can be considered as final, until we have observed that it is not (partially) negated by an ulterior statement.50 This applies to one of the formulas for nēr, namely {(C)V} (referred to as “kuṟil” in line 1 of TP312i). That item can be a nēr only if it is immediately followed by a word boundary. This overgeneration51 feature of TP312i will be made clear by two further cūttirams, namely TP313i and TP315i. We shall now examine TP313i, in which the patterns for the two remaining acais, namely nērp^u and niraip^u, are specified. That cūttiram is itself interesting from the point of view of overgeneration because, while its third line eliminates an unwanted consequence of TP312i, its first line, probably for the sake of parsimony (in words), generates more forms than are necessary, and this will have to be taken care of by a later cūttiram, namely TP317i. The problem comes from the fact that TP313i deals simultaneously with two types of nērp^u and niraip^u: those that contain a {C^u} coda with “^u” (kuṟṟiyalukaram) and those that contain a {Cu} coda with “u” (muṟṟukaram). While the first type occurs relatively frequently in ancient Tamil literature, as we shall see while examining some statistics (see Chart 7), the second type appears more marginal,52 although cūttiram TP313i seems to treat them on the same footing, when it says that:

(6a) iruvakai yukaramō ṭiyaintavai variṉē // nērpu niraipu māku meṉpa // kuṟiliṇai yukara malvaḻi yāṉa. (TP313i).

(6b) “if they (i.e. nēr and nirai) come combined with [a coda containing] the two types of ukaram, (i.e. kuṟṟiyalukaram and muṟṟukaram), we have the advent of nērp^u and niraip^u, provided we do not consider the ukaram which is in a kuṟiliṇai.

21Since we have explained earlier kuṟiḷiṇai as referring to the formula {(C)V}{CV}, this seems to mean that what is avoided by the precautionary clause inside TP313i are formulas of the types {(C)V}{C^u} and {(C)V}{Cu}. Given the fact that the sequence {(C)V}{C^u} cannot be found inside the initial section of a (short) word,53 but can nevertheless be found in the final section of a long word, we can conclude from the formulation of TP313i that words such as ñāyiṟ^u “sun” and valiyat^u “that-which-is-strong” have to be scanned as “nēr-nirai” and “nirai-nirai” (and not as “nēr-nērp^u” and “nirai-nērp^u”). The final result is that we obtain fourteen formulas54 (see Chart 5) for nērp^u and niraip^u:

22Thirteen of those fourteen formulas are illustrated by Iḷampūraṇar, in his commentary to TP313i: he uses the following fourteen words (which I reorder, in order to better illustrate the chart):

• kāt u kaṉṟ^u, kāṟṟ^u, kāvu, kallu and cārvu (illustrating nērp^u)
• varak^u, malāṭ^u, arakk^u, paṉāṭṭ^u, katavu, urumu, viṉāvu and puṇarvu (illustrating niraip^u).55

6. Counting eḻutt^u-s in acai-s and in cīr-s

23In Charts 4 and 5 the last column indicates the lengths of the various sub-types, as measured in eḻutt^us. As already indicated (see n. 25), among the Level-1 units that come in the making of a Level-2 unit (i.e. an acai), those that have a duration inferior to one māttirai are not to be counted. As a consequence, the words which are given by Iḷampūraṇar for illustrating nērp^u and niraip^u have the following lengths:58

• kā- t^u, kā- ṟ-ṟ^u and ka -ṉ-ṟ^u have a length of one eḻutt^u.
• kā - vu , cā- r -vu and ka -l- lu have a length of two eḻutt^u-s.
• va-ra -k^u, a-ra -k-k^u, ma-lā -ṭ^u and pa-ṉā -ṭ-ṭ^u have a length of two eḻutt^u-s.
• ka-ta-vu, pu-ṇa- r- vu, u-ru-mu and vi-ṉā-vu have a length of three eḻutt^u-s.

24This type of calculation also extends to the Level-3 metrical units (i.e. the cīr-s), which we have introduced in Chart 2. It is therefore necessary to further subdivide the categories given there, if we want to be able to specify the lengths (measured in eḻuttus) of the various cīrs. If we limit ourselves to the groups (a), (b), (c) and (d) introduced in Chart 2, we have the possibilities listed in Chart 6 (see page 262):

25It should be clear, on the basis of the explanations given for Charts 4 and 5, why, if we select one example from Chart 6, the two words given as illustrations for EP-E (nērp^u-nēr), which are pō- t^u -pū and mē-vu-cī -r, have respective lengths of two eḻutt^u-s and three eḻutt^u-s. We shall now turn to the task of providing some statistics on the “real use” of the various types of cīr-s presented in Chart 6, drawn from one of the works belonging to the core corpus of Tamil literature. This appears necessary if we want to put things into real perspective, because the reading of texts written by theoreticians (grammarians) may sometimes present in an apparently symmetrical way phenomena that are in fact quite unequally distributed.

7. Counting eḻutt^u-s in the cīr-s of veṇpā metre

26The work which I have chosen for illustrating the distribution is the first section of the Kuṟaḷ. That section is called Aṟatt^u p pāl and it contains 380 distichs. Those distichs are called kuṟuveṇpāṭṭ^u by the TP-Cey (and kuṟaḷ veṇpā in more modern parlance) and each one of them contains two metrical lines totalling seven cīrs.62 This provides us with a sample of 760 lines and 2660 cīrs. Their common structure is represented here by the following figure, where each individual cīr receives an identifier according to the philologists’fashion:

27It should be added that those verses are often considered as the most perfect representatives of a metre called veṇpā, which is one of the four Classical Tamil metres (the other three being called āciriyam, vañcippā and kalippā) and that the Tolkāppiyam contains, inside the TP-Cey, several cūttirams containing specifications supposed to apply to that metre. It must be said however that there is not a total agreement between the Tolkāppiyam commentators on the correct interpretation to be given to those cūttirams and, when people refer nowadays to the veṇpā metre, they usually understand it as explained by later treatises such as the YA or the YK.67 Nevertheless, if we take the TP-Cey seriously, it seems an important requirement to examine whether the specifications which it spells out for the veṇpā metre are respected by the literary works which are deemed to be composed in that metre, and especially by the most famous among them, the Kuṟaḷ. That includes, among other tasks, the computing of some statistics on the length (measured in eḻuttus) of the lines contained in those works, because the cūttiram TP364i (alias TP370p or TP-Cey58n) does give precise specifications for those lines, and because those specifications will play an important role, if we really want to understand the differences of opinion between the metrical commentators (of YA and TP-Cey), when they deal with TP357i concerning the interpretation of the figures “seventy” and “six hundred and twenty-five” (see section 1), for which I have already presented Iḷampūraṇar’s interpretation (see sections 2 and 3). This is why Chart 7 gives a precise description of the distribution of the 2660 cīrs found inside the Aṟattup Pāl of the Kuṟaḷ.

28It is of course impossible to summarise in a few lines the data contained in this chart but it is nevertheless possible to observe that the distribution of cīrs found in the Kuṟaḷ is quite different from the distribution of cīrs found inside the Kuṟuntokai, a sample from which has been analysed in Chevillard (2011). The most striking features are the much higher frequency of the cīrs made of three acais and the high visibility of the cīrs containing only one acai. The second feature is easily explained by one of the constitutive constraints which a verse in veṇpā metre must follow, namely that the third (and last) cīr in its last line must be an “acaic cīr”.68 In the case of a kuṟuveṇpāṭṭ^u, which has only two lines, this concerns the cīr labelled as 2c (see Fig. 2). If we remove those obligatory instances from the statistics (there are 380 of them), we are leftwith the following distribution (concerning the 2280 remaining cīr-s):

29In addition to those figures, I shall also now provide the statistics concerning the line lengths of distichs, and connect this with the general problem of the relationship between theory and practice in Tamil metrical literature. I should also mention here that the Appendix B contains four similar additional charts, for two Kalittokai poems. They will be evoked in section 17, when discussing the “mannerism” of the kali metre.

8. Which are the loci (nilam-s) represented in veṇpā lines?

30As we can see in figure 2, the two lines in a Kuṟaḷ distich differ in the number of cīrs which they contain. If we add together, for each of the 380 distichs in the Aṟatt^u p pāl, the lengths (measured in eḻutt^us) of the cīrs labelled 1a, 1b, 1c and 1d, we obtain integers contained inside the interval [10; 16], with the distribution which is given inside the following chart, where the conventional names given to lines by the Tolkāppiyam in the TP-Cey have been added (as spelled out in Chart 1).

31Similarly, doing the same for the second lines (which however have only three cīr-s) of those distichs, we obtain integers belonging to the interval [5; 10], as indicated in chart 12.⁷⁰⁷¹

32Data of this type, compiled on the basis of a real, independently existing literary work, such as the Kuṟaḷ, can now be compared with statements found in metrical treatises, concerning the possible loci (nilam) in which veṇpā lines are found, among the “seventeen loci”, and it seems logical to start the exploration with the TP-Cey, in which there is a cūttiram (TP364i) dealing with this topic. That cūttiram states that:

(7) aḷavuñ cintum veḷḷaik kuriya // taḷaivakai yoṉṟāt taṉmai yāṉa “being either a normal or a short line is appropriate for (a) WHITE (line), when the property (of that line) is UNEQUAL hindrance-type (taḷai-vakai)” (TP364i).

33The translation which I am providing here tries to stick to the literal meaning of the words used in the cūttiram, but it is fair to tell the reader that for a text as loaded with history as the Tolkāppiyam, a commentator has a lot of latitude for projecting all sorts of interpretations on one cūttiram which will be immediately understandable only to those who have understood precisely how that commentator interprets many other cūttirams. A possibly “naive” view of the relationship between TP364i (on the one hand) and Charts 11 and 12 (on the other hand) may induce one to think that if TP364i mentions both cintaṭi and aḷavaṭi as appropriate for veṇpā lines, and if Charts 11 and 12 show that in the first book of the Kuṟaḷ a huge majority (93.2%) of three cīrs lines seem eligible to be called “cintaṭi” (on the basis of eḻutt^u count)72, this means that TP-Cey mentions “cintaṭi” because of three-cīr lines and “aḷavaṭi” because of four-cīr lines. However, the examples provided by Iḷampūraṇar for cintaṭi are four-cīr lines (although none of them comes from the Kuṟaḷ, where he could not have found any, at least in the Aṟatt^u p pāl, as attested by Chart 11) in contradistinction with the examples which he provides for “aḷavaṭi”. And he also gives examples of veṇpā lines which are neṭilaṭi, because they contain fifteen or sixteen countable eḻutt^us.73

9. What is a Kaṭṭaḷaiyaṭi “touchstone line” for Pērāciriyar?

34If we now consider Pērāciriyar’s interpretation of the same cūttiram, we can say, in a nutshell, that this cūttiram (TP364i, alias TP370p) expresses for him, like many other cūttirams in the TP-Cey, the distinction to be made between what he calls a kaṭṭaḷai aṭi and another type of line, which he calls a “cīr vakai aṭi”, on the basis of a distinction which he discusses, among other places, under TP366p (alias, partially, TP361i).74 Before attempting to explain the word kaṭṭaḷai, if it is at all possible,75 a preliminary remark is that it is not a word found in the Tolkāppiyam itself. In ordinary language (if the Caṅkam anthologies can be said to represent “ordinary” language), the word “kaṭṭaḷai” (or its homonym) designates the “touchstone” that a goldsmith uses in order to test the quality of a specimen of gold.76 It is noticeable that Pērāciriyar tends to use the word “kaṭṭaḷai” in his commentary to gloss the word taḷai in the Tolkāppiyam.77 As for Nacciṉārkkiṉiyar, he takes pains to explain that the items he calls “kaṭṭaḷai aṭi” do not have the same properties as the “kaṭṭaḷaik kalittuṟai”78 of later metricians, because the kuṟṟiyalukaram is counted in the line length for the kaṭṭaḷaik kalittuṟai,79 but not for his own kaṭṭaḷai aṭi.

35However, while reading Pērāciriyar’s commentary on the TP-Cey, one often gains the impression that the net result of his description of ideal metrical lines is to present an ideal form of poetry, which does not necessarily have a counterpart in existing literature. For instance, the verses in the Kuṟaḷ which do not fall within aḷavaṭi (and cintaṭi) could never be considered as examples of kaṭṭaḷai aṭi. This is certainly paradoxical even though the intended meaning of kaṭṭaḷai aṭi may not have been “touchstone line” (see n. 75),80 because in latter days the Kuṟaḷ has certainly become one of the touchstones of Tamil poetry against which other works were measured, especially in metrics. At the same time, such a contradiction (i.e. the idea that the Kuṟaḷ might not have originally corresponded, in the idea of some grammarians, to the ideal described by the Tolkāppiyam) might explain, in part, why there have been popular stories concerning a conflict (of pre-eminence) between the Kuṟaḷ and the Caṅkam poets and why it was the Kuṟaḷ which won in front of pedantic grammarians, who had to fall into the Lotus pond.81 One may also wonder who the Nakkīraṉ was, who is said in YV to have composed an aṭi nūl “treatise on [metrical] lines”.82 Miraculous legends/stories may have been invented on the basis of real characters, debating technical points, which most people could not understand, but would be happy to mock.

36Coming back to Pērāciriyar, one of the interesting things about the “six hundred and twenty-five [metrical] lines”, thanks to which we started this exploration of Tamil metrical literature, is the fact that they are for him the kaṭṭaḷai aṭi, at least as far as three of the four classical Tamil metres are concerned, namely āciriyam, which is supposed to claim “three hundred and twenty-four” ideal models of line inside the total which it has to share with the veṇpā metre (claiming “one hundred and eighty-one” models of line) and the kali metre (claiming “one hundred and twenty”).

37One of the most telling passages, however, concerning Pērāciriyar’s view of the world might be the passage inside his commentary on TP363p, where he explains that the kaṭaiccaṅkattār (i.e. the members of the last Caṅkam) were not as gifted (or well trained) as their predecessors for composing verse by making use of the “six hundred and twentyfive” models of kaṭṭaḷaiyaṭi. Why should we not conclude that the flood which supposedly took away the “chef-d’oeuvres” of the first and second Caṅkam poets83 may simply have been the discrepancy between an honest practice of poetry and an increasingly precise use of theoretical speculation, which had transformed some preliminary technical approximations into a threatening procrustean bed, which no longer fitted with the practice of poets, because precise new meanings had been given to old labels, against a background of changing linguistic realities, affecting for instance the topic which I shall discuss in the next session and the pronunciation of chaste poetical Tamil (centamiḻ)?

10. The nature of the kuṟṟiyalukaram

38I have so far mentioned the kuṟṟiyalukaram “over-short u” (literally “short-natured u”) as if it were a self-evident eternal notion, the existence of which is proved by the fact that it has been mentioned by the Tolkāppiyam, as one of the three cārpeḻutt^us (see Chart 3).84 It must however be said that like the other components of the Classical Tamil language, it was (and still is) an item existing in a diglossic situation85 where the conceived (and perceived) ideal realities did not necessarily coincide with the realities which could be met with in everyday life.

39One of the most remarkable properties of the kuṟṟiyalukaram is the fact that it disappears in a number of sandhi situations, and more precisely when a word ending with a kuṟṟiyalukaram is followed by a word starting with a vowel. This can be illustrated by the different behaviour of the three instances of the word virunt^u “guest”, in the following distich taken from the Kuṟaḷ.

(8) celvirun tōmpi varuvirunt^u pārttiruppā ṉalvirunt^u vāṉat tavarkk^u (K86) “He who, having entertained the guests that have come, looks out for others who may yet come, will be a welcome guest to the inhabitants of heaven” (Translation Vaṭivēlu Ceṭṭiyār, 1904/1972).

40In this example, which can be analysed as containing ten items, namely cel, virunt^u, ōmpi, varu, virunt^u, pārtt^u, iruppāṉ, nal, virunt^u and vāṉattavarkk^u, we see that the first instance of virunt^u, which is followed by ōmpi, a word with an initial vowel, does not behave in the same way as the two other occurrences, which are followed respectively by pārtt^u and by vāṉattavarkk^u, two words with initial consonants. This behaviour is described by the Naṉṉūl, a medieval grammar, in a cūttiram which says (refering to the kuṟṟiyalukaram as “ukkuṟaḷ”):

(9) uyirvari ṉukkuṟaṇ meyviṭ ṭōṭum // yavvari ṉiyyā muṟṟumuṟ ṟorōvaḻi “if a vowel comes [after it], the dwarf u, leaving [its] consonant, runs away; if [consonant] y comes, it becomes [dwarf] i; this applies sometimes to full u” (N164vi, see Taṇṭapāṇi Tēcikar 1957)

41Interestingly, inside the Tolkāppiyam itself, this behaviour is described in a much less explicit way. The commentators to its first book, the Eḻuttatikāram (TE), describe the phenomenon while commenting on a cūttiram (TE106i/TE105n) which states that the kuṟṟiyalukaram is like final consonants, apparently implying, because of what has just been stated in TE105i, that it stands with a puḷḷi “dot”. This is further interpreted, explicitly in the case of Nacciṉārkkiṉiyar, as signifying that the kuṟṟiyalukaram, at the end of a word, allows the vowel which is in initial position in the following word to climb onto the consonant which it itself rides.86 Intriguingly, Nacciṉārkkiṉiyar seems to consider (also under TE105n) that although another vowel has climbed on the consonant, the kuṟṟiyalukaram is still present, apparently refuting the formulation found in the Naṉṉūl, where we have the word viṭṭ^u “having left”.87

42Another element to be considered is the fact that inside the chapter division of the TE, the last three chapters, called respectively uyir mayaṅkiyal (chapter 7, TE204i to TE296i), puḷḷi mayaṅkiyal (chapter 8, TE297i to TE406i) and kuṟṟiyalukarap puṇariyal (chapter 9, TE407i to TE483i), which take up more than half the text of TE, are devoted to a detailed examination of the sandhi phenomena which take place when a word B follows a word A. The distribution between the three chapters is the following:

• if A ends with a vowel (uyir), the case is handled in chapter 7 (uyir mayaṅkiyal)
• if A ends with a consonant (mey or puḷḷi), the case is handled in chapter 8 (puḷḷi mayaṅkiyal)
• if A ends with a kuṟṟiyalukaram, the case is handled in chapter 9 (kuṟṟiyalukarap puṇariyal)

43All these elements seem to indicate that, from the point of view of those who originally described the language which is now known to us as Classical Tamil, and which is best represented by a grammar, the Tolkāppiyam, and by a corpus of poetical texts, which Pērāciriyar refers to as Pāṭṭun Tokaiyum (which we normally understand as referring to the Patt^u p pāṭṭ^u and to the Eṭṭ^u t tokai), the kuṟṟiyalukaram was neither a vowel nor a consonant but partook of the nature of both.

44But all this was to change, in a certain measure, as we shall see in the following section, the change being possibly due to a change in the dominant local dialect among the poets who were the main exponents of Classical Tamil. Ancient southern Tamil Nadu poets (from Pāṇṭiya Nāṭu or from Cēra Nāṭu)88 may have been replaced in the dominant position by northern Tamil Nadu poets, who may have been unable to properly utter the ritually correct pronunciation of the kuṟṟiyalukaram while reciting poetry, which may have led to a catastrophic change in the poetical standard.

11. The simplification of Tamil metrics by later metricians

45Whatever the reasons may have been, we do indeed see major changes taking place when the YA, the YK and the VC, along with their commentaries, replace (to some extent) the TP-Cey as a much simpler guide on Tamil metrics. In order to understand the drastic simplification which has taken place, it suffices to compare, inside Chart 13, the possible 2-acai cīr-s conceivable in the framework of those later three treatises with those which we have already discussed several times, while presenting the TP-Cey (see Chart 2).

46As explained by Iḷampūraṇar, while commenting on TP313i, one of the main proponents of the need to simplify Tamil metrics was the theoretician known under the name of Kākkaipāṭiṉiyar. According to Iḷampūraṇar, Kākkaipāṭiṉiyar, from whose lost work only 89 cūttirams have been preserved,90 explained that nērp^u and niraip^u were not necessary for describing Tamil poetry. Iḷampūraṇar nevertheless is quick to point out that in his description of the veṇpā metre, Kākkaipāṭiṉiyar was forced to retain the notion that the last cīr in the last line of a veṇpā must be a 1-acai cīr, and that four possibilities exist for that, which correspond to the old four acais found in the Tolkāppiyam, although this is the only place where they will be seen, in the new theory.91 Ironically, later traditions have made him into a classmate of the author of Tolkāppiyam, stating that he was, along with him, one of the twelve students of Agastya (see Chevillard 2009).92

47Another important change can also be seen in the naming conventions adopted by later metricians for the aṭis (metrical lines). The YA uses the same names as the TP-Cey, but it gives them new significations. For instance, the kuṟaḷaṭi “dwarf line” is no longer an aṭi containing four to six countable eḻutt^us, but it is defined as an aṭi containing two cīrs, whereas the cintaṭi is defined as containing three cīrs (see YA24). The differences between the two systems are given in Chart 14.

48Inside the commentary to the YA-24 and to YA-25 (referred to as YV), some verses from the lost treatise by Kākkaipāṭiṉiyar are given, showing that he used those names with their modern meaning. It is not clear whether the new naming scheme was invented by him, but, if not, he was certainly one of its users.93 Another very important change, which I shall however only discuss later, is the codification of taḷai (see sections 17 and 18).

49It must however be emphasised that although the YA (Yāpparuṅkalam) was certainly originally intended to be a treatise presenting a much simpler metrical theory within its ninety-six cūttirams, it acquired, thanks to the commentary that grew on it, and which is usually referred to as Yāpparuṅkala Virutti (YV), the role of a repository, or an archive, of the whole history of Tamil metrical theories, which uniquely preserves citations from many theoreticians whose work has been lost. It is especially worth noting its third part, the Oḻipiyal “Chapter of the rest”, which contains only three cūttirams (YA94 to YA96), having respectively two, seven and twenty-nine lines, but for which the commentary has become, in part, a totally independent treatise, as can be seen from the fact that, in the 1998 edition (YV-1998), the seven lines of YA95 require 164 pages of digressing commentary (YV-1998, pp. 397–560) while the twenty-nine lines of YA96 still require 60 pages of illustration (See YV-1998: 560–619). That this commentary, especially in the case of the section (nominally) attached to YA95, does not directly comment on YA will be clear if we remark for instance that the YV devotes twenty-four pages (YV-1998: 483–506) to an explanation of the Tolkāppiyam cūttiram (TP357i) which we have been discussing in this article from section 1 (Prologue) onwards, replacing however the word “vakaiyiṉ” in the third line by the word “taḷaiyiṉ”. I cannot give a detailed account of the YV explanations, but the next section will be devoted to a presentation of the point of view which it illustrates, which is all the more important since it has been discussed (and refuted) by Pērāciriyar in his commentary on the same cūttiram (TP362p), as noted by Ganesh Aiyer (1943: 314, n. 3).

12. How the author of YV explains “seventy”

50With the added familiarity that comes from exploring one question from several angles, we now return to the examination of the interpretations given, at several points in time, to the two figures “seventy” and “six hundred and twenty-five”, for which I have presented Iḷampūraṇar’s explanation, inside sections 2 and 3 of this article.

51As already said, the important word in the context is the word nilam “locus”, which appears in the recurrent expression patiṉēḻ nilam “seventeen loci” (see Chart 1). However, the YV author has a horizon of retrospection which differs from Iḷampūraṇar’s and he uses it for stating that the three metres upon which he will base his explanation, namely āciriyam, veṇpā and kalippā, are each entitled to the following counts of loci:

• āciriyam possesses all the seventeen loci (or sub-types) [This agrees with TP359i]
• veṇpā possesses ten loci (or sub-types) [This contradicts TP364i; cf. (7)]
• kalippā possesses eight loci (or sub-types) [This agrees with TP365i]

52Although this statement is partly at variance with the TP-Cey, it is based on an old cūttiram, detailing a count of fifty-one loci (including the thirty-five which I have just mentioned) which the YV author had earlier cited, without giving its author’s name, while commenting on YA49.94 Expressed in simpler language, and drawing on his commentary, those statement are meant to inform us that, if measured in “countable eḻuttus”,

• the length of a line in the āciriyam metre can go from 4 to 20 eḻutt^us (seventeen sub-types)
• the length of a line in the veṇpā metre can go from 7 to 16 (ten sub-types)
• the length of a line in the kali metre can go from 13 to 20 (eight sub-types)

53The reader may think that it is not possible to understand such statements without knowing what precisely those three metres are but, as I am trying to suggest, technical terms such as kali, veṇpā and āciriyam (as well as vañci) should best be understood, in historical perspective, as (almost) pure labels which are capable of surviving a change in meaning, across what might be (daringly) called successive historical “dialogues de sourds” between successive generations. Human beings crave for meaning, i.e. for the belief that the items with which they deal and to which they are attached are meaningful, which means, among other things, for them to have a stable meaning. However, history teaches us otherwise, but it also teaches us that figures, enumerations, formal elements, symbols, etc. have a life of their own and tend to survive reinterpretations.95

54Therefore, concentrating on the formal, mathematical aspect of those (almost) countless enumerations, I shall now try to explain, as clearly as I can, how the YV author justifies the expression “eḻupat^u taḷai vaḻu” which he reads inside what is for him the text of what I have given in (1).96

55As we should remember, the items under discussion are standard metrical lines (aṭi), made of four cīrs (see n. 22), which means that inside such a line there are three meeting points between cīrs because there are three sequences of two cīrs (see figure 1), namely (1) between the first cīr and the second, (2) between the second cīr and the third, and (3) finally between the third cīr and the fourth. According to the standard theory which is represented by the YA, the meeting point of two cīrs is the place where a constraint called taḷai can operate (see YA-1797 and see sections 17 and 18).

56A constraint (or “binding” [French “entrave”]) such as taḷai can be either respected or violated. In the latter case, we have a taḷai vaḻu “fault in the taḷai”. The YV author seems to imagine that, for each of the seventeen āciriyam lines, ten veṇpā lines and eight kali lines, there are two possible faults, consisting of:

• using (at least once) either the taḷai appropriate for veṇpā or the taḷai appropriate for kali inside one of the seventeen lines which should follow only the taḷai appropriate for āciriyam.98
• using (at least once) either the taḷai appropriate for āciriyam or the taḷai appropriate for kali inside one of the ten lines which should follow only the taḷai appropriate for veṇpā.99
• using (at least once) either the taḷai appropriate for veṇpā or the taḷai appropriate for āciriyam inside one of the eight lines which should follow only the taḷai appropriate for kali.100

57If we add all those faults together, we obtain (2x17) + (2x10) + (2x8) = 70. These are the explanations which the YV commentator provides for the figure “seventy” inside his reading of the cūttiram TP357i (see n. 96). It can be announced, in anticipatory manner, that Pērāciriyar will find faults with this explanation, but before trying to understand his objections, we have to examine the explanations concerning the other figure, namely “six hundred and twenty-five”, and this is what we shall do in the next section.

13. How the author of YV explains “six hundred and twenty-five”

58Continuing our exploration of what is, in reality, theoretical metrics and may not have much to do with real poetry, I shall now present the explanation given by the YV author for the figure “six hundred and twentyfive”. In this case, the explanation is based on the list of possible cīrs by which the line can start. The YV author has a separate list of possible cīrs for each of the three metres and he also relies on a set of “precompiled information”, which takes the form of two-line and four-line veṇpās, that he seems to accept uncritically, although they are more dogmatic than explanatory. Since our goal is to understand the logic which is at play rather than the complete detail of the calculation, I shall comment in detail only on one fragment, and summarise the remaining parts. The general idea is that the total of 625 is obtained as the sum of three smaller numbers (261, 232 and 132),101 one of them (namely 261) concerning the metre called āciriyam, the second one (namely 232) concerning the metre called veṇpā and the last one (namely 132) concerning the metre called kali. Each of those smaller numbers is also obtained as the sum of still smaller numbers, which are explained on the basis of the cīrs that stand at the beginning of the lines.

59For instance, the total of 261 for āciriyam is computed as the sum of five smaller numbers, namely 50, 88, 61, 12 and 50. Among these five, the first four correspond to lines starting with an iyaṟcīr and the last one corresponds to lines starting with an acaiccīr. Although those technical terms may appear familiar (because they have been seen in Charts 2, 6 and 7), the reader should be warned that the referent is not exactly the same (although there is much overlap). If one concentrates for instance on the four sub-groups corresponding to the lines starting with an iyaṟcīr, one may be puzzled by the formulations used by the YV author until it is realised that he probably belongs to an age predating some of the distinctions which have been taken for granted (for the sake of easier reading) inside Charts 6 and 7. What I am trying to say should become obvious when reading the following passage in YV (1998: 485–89):

(10a) iyaṟcīr pattumē koṇṭu aṭi vakukkumiṭattu nāṉku nilaimaiyavām. iraṇṭeḻuttuc cīrum, mūṉṟeḻuttuc cīrum, nāṉkeḻuttuc cīrum, ainteḻuttuc cīrum eṉa. “When subdividing [metrical] line (aṭi) [types] by making use of (their initial) ten iyaṟcīrs, they are found to possess four possible statuses: 1. being [lines starting with] a cīr of two eḻuttus; 2. being [lines starting with] a cīr of three eḻuttus; 3. being [lines starting with] a cīr of four eḻuttus; 4. being [lines starting with] a cīr of five eḻuttus” (YV-1998: 485).
(10b) avaṟṟuḷ, īreḻuttuc cīrāvaṉa, nāṉkām. avaiyāvaṉa, ‘pōtupū, pōrēṟu, pātiri, tēmā’ eṉa ivai. “Among them, there are four ‘cīrs of two eḻuttus’. They are pōtupū, pōrēṟu, pātiri and tēmā”.
(10c) […] iṉi, mūveḻuttuc cīrāvaṉa, ēḻu cīr. avaiyāvaṉa, ‘pātiri, puḷimā, viṟakutī, pōtupū, pōrēṟu, pūmarutu, kaṭiyāṟu’ eṉa ivai “[…] furthermore the cīrs of three eḻuttus are seven. They are ‘pātiri, puḷimā, viṟakutī, pōtupū, pōrēṟu, pūmarutu and kaṭiyāṟu’” (YV-1998: 487).
(10d) […] iṉi, nāleḻuttuc cīrāvaṉa aintu vakaippaṭum. ‘avai yāvaiyō?’ eṉiṉ, ‘kaṇaviri, pūmarutu, kaṭiyāṟu, viṟakutī, maḻakaḷiṟu’eṉa ivai. “[…] furthermore the cīrs of four eḻuttus are five. If one asks ‘which ones? [The answer is]: ‘kaṇaviri, pūmarutu, kaṭiyāṟu, viṟakutī and maḻakaḷiṟu’” (YV-1998: 488).
(10e) […] iṉi ainteḻuttuc cīrāvatu ‘maḻakaḷiṟu’ eṉpatu ““[…] furthermore the [unique] cīr of five eḻuttus is ‘ maḻakaḷiṟu’” (YV-1998: 489).

60What may immediately strike the reader is the fact that statements which are apparently contradictory are made. How is it possible for pātiri, pōtupū and pōrēṟu to be said to possess two eḻuttus in (10b) and three eḻuttus in (10c)? How is it possible for viṟakutī, pūmarutu and kaṭiyāṟu to be said to possess three eḻuttus in (10c) and four eḻuttus in (10d)? And finally, how is it possible for maḻakaḷiṟu to be said to possess four eḻuttus in (10d) and five eḻuttus in (10e)? The short answer is that what is referred to in 10b as “pātiri [having] two eḻuttus” will be referred to by later metricians (such as Pērāciriyar and Nacciṉārkkiṉiyar) as “ñāyiṟ^u”.102 And for the six other problematic items, we meet with other designations, summarised in the following chart:

61We see, thus, that the list of ten iyarcīrs to which the YV refers several times is in fact a list of seventeen distinct items,103 if we accept that this is what this metrician had in mind. He seems however to be in fact paraphrasing in prose a series of explanations which had been transmitted in verse form (as veṇpās), perhaps considering them as too cryptic.104

62For the sake of clarity, I have grouped together (in 10b to 10e, cf. above) the statements concerning the lengths of the various iyaṟcīr-s. These statements are however separated by other explanatory segments (or rather prose paraphrases explaining the meaning of several veṇpā-s quoted as authorities, but without real demonstration). Typically, some prose segments tell us which are the minimum length and the maximum length of the metrical lines which start with a given iyaṟcīr. On that basis, the count of nilams which are obtained thanks to that cīr is announced. At the end of the section, the contributions are added together. As an example, I shall now summarise the “explanations” given between 10b and 10c (see YV-1998: 483–84).

• It is said first of all that both for tēmā and for pātiri (i.e. in fact ñāyiṟ^u), there exist āciriyam lines that have them as their first element and have lengths which can range between five and seventeen.105 Therefore, the count of nilam for each of them is “thirteen”.
• In the case of pōt^u pū and pōrēṟ^u, it is said that there exist āciriyam lines that have them as their first element and have lengths which range between six and seventeen. Therefore, the count of nilam for each of them is “twelve”.
• Adding together two times thirteen and two times twelve, one obtains the count of “fifty”,106 announced earlier as being the contribution of that section to the grand total for the āciriyam lines.

63This type of explanation is continued for more than twenty pages (up to p. 506 in the edition which we have been using) but the fragment already given should be sufficient for the reader to understand the nature of the arguments used. I shall now summarise all the statements made by the YV author in Chart 15B, indicating for each cīr template the identity of the line templates which can start with it in each of the three categories concerned (āciriyam, veḷḷai and kali). The identity is specified by a range, i.e. by indicating the length of the shortest and of the longest lines which can start with it. It is immediately followed by the count for that range. For instance, the chart indicates that puḷimā has the range [6; 18] in āciriyam and the range [8; 16] in veṇpā, and is not used in kali. This means that there exist thirteen āciriyam line templates starting with puḷimā (the shortest having six countable eḻutt^us and the longest having eighteen eḻutt^us) while there exist nine veṇpā line templates which start with puḷimā (the shortest having eight countable eḻutt^us and the longest having sixteen coutable eḻutt^us). The chart is as follows.

14. Are these notions applicable to “real” literature?

64The reader may be puzzled by (or drowning in) this avalanche of numbers. A question which must be asked is the following: did the poets really follow such rules? On the one hand, we must not underestimate the capacity of poets, because it is a fact that a significant part of Tamil literature has strict constraints on line lengths, as evidenced for instance by the works composed in the “medieval” metre called kaṭṭaḷaik kalittuṟai and by a significant part of Tamil devotional literature (see Chevillard forthcoming, Studies-2). On the other hand, the theoreticians might be overdoing it (being carried away by their desire for exactitude), and might be describing constraints that have never really been followed by real productive poets. The only way to know is to examine the literature which we have and see whether it was really composed that way.

65The answer may however differ for the three types of metre considered in the chart. While ONE part of the answer might be that, YES, to a great extent, veṇpā may follow many of those constraints, but maybe not all of them, ANOTHER part might be that, NO, classical works which are composed in āciriyam metre cannot possibly be said to follow those constraints, except in very exceptional (and rare or suspicious) cases. And as far as kali is concerned, there is an added difficulty in the fact that the label kali (as illustrated for instance by the anthology called Kalittokai) does not refer to a really homogeneous poetical reality, from a metrical point of view, because there are many types of kali poem, and this is the reason why the section devoted to it inside the Tolkāppiyam is rather long and complex, totalling twentyfour cūttirams (from TP435i to TP458i). And therefore, answering the question of the relevance of the data contained inside the columns that deal with it in Chart 15B is extremely complex. It may be added that the four charts contained in Appendix B, which provide statistics on the cīrs used in two poems from Kalittokai, illustrating two very different types (ottāḻicaik kalippā and kaliveṇpāṭṭu) may contribute to the beginning of an answer.

66Coming back to the simpler part, which is the veṇpā metre, it is not difficult to extract from the database which I have prepared for the sake of making the Charts 7 to 12, a number of lines satisfying specific constraints. A full answer would occupy too much space, but one can easily state that, among the 380 first lines of the Aṟattuppāl distichs:

• 64 lines start with the tēmā templates and their lengths fully cover the [10;14] range
• 42 lines start with the puḷimā templates and their lengths belong to the set {10, 11, 12, 13, 15}
• 39 lines start with the pātiri template, their lengths covering the [10;14] range
• 43 lines start with the kaṇaviri template, their lengths covering the range [11;14]
• 1 line starts with the pōt^upū template and its length is 11
• no line starts with the viṟak^utī pattern • 10 lines start with the pōrēṟ^u template and their lengths cover the range [10;13]
• 9 lines start with the kaṭiyāṟ^u template and their lengths cover the range [11;13]
• 2 lines start with the pūmarut^u template and their length is 12
• 4 lines start with the maḻakaḷiṟ^u template and their lengths cover the range [12;13]
• 49 lines start with the mācelvāy template and their lengths cover the range [11;13]
• 80 lines start with the pulicelvāy template and their lengths cover the range [11;14]
• 8 lines start with the māpaṭuvāy template and their lengths cover the range [12;14]
• 23 lines start with the pulipaṭuvāy template and their lengths cover the range [13;16]
• no line starts with the NĒRP^U template
• 1 line starts with the NIRAIP^U template and its length is 11

67The total does not add up to 380 because there are five special cases which I have leftout as they do not concern us here. But it can be remarked that this sample covers only 45 distinct items among the 232 theoretically possible templates mentioned in Chart 15B.

15. How Pērāciriyar and Nacciṉārkkiṉiyar explain “six hundred and twenty-five”

68We are now nearing the end of this exploration of a fragment of Tamil shastric literature and the last item which we shall examine in detail, before drawing a few conclusions, is the procedure used by Pērāciriyar and by Nacciṉārkkiṉiyar (who closely follows Pērāciriyar) for explaining the technical items already examined several times by us in the course of this study. As I have tried to suggest several times, what is at stake is not an examination of the really attested structure of the Tamil poetical corpus (although sections 6 and 7 were concerned with that question). We are really dealing here with theoretical (or speculative) metrics. But, at the same time, we are also dealing with the pedagogy associated with the transmission of theoretical points of view, which is in a sense independent from the underlying reality, because the students are supposed to first memorise verses and only then to understand what the items which they have memorised refer to. This was rather clear in the case of the school through which the YV was transmitted, along with its collection of embedded veṇpās (see n. 104 and n. 106).

69Because Nacciṉārkkiṉiyar closely follows Pērāciriyar, I shall now present his conclusions, as concisely as possible. The data which follows is summarised from the reading of pp. 314–328 in the 1943 Pērāciriyam edition (by Ganesh Iyer) and from the reading of pp. 42–47 in the 2003 Nacciṉārkkiṉiyam edition (Tamiḻ Maṇ patippakam). Since Pērāciriyar criticised the detail (but not the spirit) of the YV explanation, he must have read it, and since Nacciṉārkkiṉiyar closely follows Pērāciriyar, he must have read him and that probably explains why each of the three texts is shorter and more condensed than the preceding ones. I shall try to be even more concise, because the data which I shall soon provide (in Chart 17) is, in a sense, the “printout” of an Excel file compiled on the basis of Nacciṉārkkiṉiyar’s explanations under TP-Cey43n and TP-Cey50n. These two cūttirams should be considered as a pair because they contain a series of mnemonic verses (many, but not all, being veṇpās) which contain all the necessary information for drawing the chart which follows, and, just like the YV author, Nacciṉārkkiṉiyar’s task simply seems to be to paraphrase the anonymous107 mnemonic verses.

70The key element, for understanding the theoretical metrics calculations given in what follows, is the list of thirty-one cīrs first given under TP-Cey43n and repeated under TP-Cey50n. Among those thirty-one cīrs,

• twenty-seven are said to be found in āciriyam lines
• twenty-seven are said to be found in veṇpā lines
• twenty-four are said to be found in kali lines

71More precisely, four groups of cīrs are considered by Nacciṉārkkiṉiyar, but one of them (the group of iyaṟcīrs) is split into two because of some constraints associated with the Kali metre, in its (strict) kaṭṭaḷai variety (see section 8), the consequence of which is that iyaṟcīrs ending in nēr are not allowed in that variety (see TP333i). The names of those four groups, as seen in Nacciṉārkkiṉiyar’s commentary are:

1. Iyaṟcīr (19 items, expanded from the original 10 items [see n. 103])
2. Uriccīr (4 items: a subset from the set of āciriyavuric cīrs [see (c2) in Chart 6])
3. Acaiccīr (4 items [see (a) in Chart 6])
4. Veṇcīr (4 items [see (d) in Chart 6])

72The four groups of cīrs which are, according to Nacciṉārkkiṉiyar, accepted in the kaṭṭaḷai variety of the āciriyam, veṇpā and kalippā metres have the following number of constituent elements:

73Having given these preliminary explanations, we now proceed to give the initial cīr and the number of possibilities, for each of the three varieties of lines (āciriyam, veṇpā and kalippā). They are as indicated in Chart 17.

74That chart should be interpreted in the following way: each line in the chart indicates whether (metrical) line patterns exist or do not exist, which both Nacciṉārkkiṉiyar and Pērāciriyar would call kaṭṭaḷaiyaṭi, having as a first cīr the cīr which appears in the first column. Those lines follow the constraints either of the āciriyam metre, or of the veṇpā metre or of the kalippā metre. The number of such lines, for a given initial cīr, is given in columns 7, 8 and 9, while the (individual) grand total is given in column 10. To give an example, illustrating the element A [8; 19] which appears in the line 6, that deals with the cīr template kaṇaviri, Nacciṉarkkiṉiyar tells us (p. 44, op. cit.)115 that:

• the shortest possible āciriyam line pattern starting with kaṇaviri is the line pattern “kaṇaviri varak^u vaṇṭ^u vanṭ^u” (which totals 8 countable eḻutt^us).
• the longest possible āciriyam line pattern starting with kaṇaviri is the line pattern “kaṇaviri naraiyurumu naraiyurumu naraiyurumu” (which totals 19 countable eḻutt^us).

75Whether all such lines really exist at all (in “real” literature) is however a question which cannot be answered, at least on the basis of the TP-Cey commentary composed by Nacciṉārkkiṉiyar. But nothing prevents us from checking whether they exist in theory116 and whether they exist in available literature, although the initial prospect seems bleak, for some of the line patterns at least.117

16. How Pērāciriyar and Nacciṉārkkiṉiyar explain “seventy”

76We have not yet examined how Pērāciriyar (followed by Nacciṉārkkiṉiyar) explains the figure “seventy” inside cūttiram TP357i/TP362p. I have already alluded (at the end of section 10) to the fact that Pērāciriyar refutes the interpretation given by the YV author. He does this in two phases. He first of all remarks that some people (whom he does not name)118 believe that the veṇpā metre possesses ten loci (nilam), i.e. possible line length (see section 11). As we know, this is not in accordance with TP364i, which I have translated in (7). He then reproduces an argument, identical with the one seen in YV, which could be a justification for talking about seventy taḷai faults. However, after this faithful account, he continues by stating that the argument is inconsistent (poruntātu) and he proceeds to refute it, by finding a flaw. He points out that in a kuṟaḷ line having a length of four countable eḻutt^us, but having (like every item discussed under TP362p) four cīr-s, the only possibility is that each cīr will possess only one eḻutt^u. This entails that each cīr will be either vaṇṭ^u or n^untai. But in such a line there is no place for either veṇṭaḷai or for kalittaḷai. Therefore this item cannot be an occasion for fault and it is not acceptable to say that eḻupat^u “seventy” refers to eḻupat^u taḷai vaḻu “seventy taḷai faults”.

77If this is not the correct explanation for eḻupat^u, what is, for him, the correct one? The explanation provided by Pērāciriyar (and reproduced by Nacciṉārkkiṉiyar) is the following. It can be followed easily by keeping an eye on Chart 16.

• take the 27 cīr-s used in the āciriyam metre (among which there are the four acaiccīr-s)
• take the 27 cīr-s used in the veṇpā metre (among which there are the four acaiccīr-s)
• take the 24 cīr-s used in the kali metre
• add together these three numbers (27, 27 and 24), and obtain 78
• subtract two times 4 from 78, in order to remove the acaiccīrs from the total
• you have obtained 70 (= 78-4-4)

78What is the justification given for removing the acaiccīrs? The reason given is that this is done because of what cūttiram TP340p (alias TP336i or TP-Cey28n) prescribes when it explains, completing the preceding cūttiram, that what has the “status of acai” (acainilai) and has obtained the “status of cīr” (cīrnilai) must be assimilated (pāṟpaṭuttal) to an iyaṟcīr, in order for the taḷai not to be disrupted (citaital).119 Therefore, when the third line in TP362p reads eḻupat^u vakaimaiyiṉ vaḻuvila vāki, what is hinted at is the fact that the cir-s chosen are apt to be used in taḷai. Therefore, what are faultless are the seventy cīr-s contained in that selection.

17. Poets, theoreticians of poetry and the metaphor of walking

79We are now approaching the conclusion of this journey across time and I have (hopefully) explained the three translations proposed in the prologue (section 1), as 2a, 2b and 2c. In order however to see things in a proper perspective, rather than asking “Is any of the three the correct one?”,120 additional remarks are necessary, centring on the word taḷai, often met, but never really discussed. Those remarks (found in section 17) will be preceded by a short sketch of the pre-history of Tamil metrics, as I speculatively imagine it. As a preliminary remark, however, I must repeat that it is difficult to know exactly what the word taḷai means inside the TP-Cey because there is a natural tendency to read the words one sees in ancient texts as if they already had the precise technical meanings which they will acquire later, and taḷai, which occurs nine times inside the Tolkāppiyam, is probably one of those words.

80The Tolkāppiyam cūttiram (TP357i) which I have chosen as the focus for this discussion probably crystallises some of the tensions that occur in the give-and-take relationship between poets and theoreticians of poetry, as is for instance seen when examining the successive uses of the word taḷai by metricians. But, as I said, such an examination must be preceded by an evocation of the pre-history of metrics.

81In the oldest stages of Tamil poetry there were probably no theoreticians (and therefore no treatises). There were only poets (or bards)121 who recited their compositions in ritual circumstances. Those poets probably used several styles, the traces of which may still be visible in the opposition between (a) akaval and (b) vañci.

(a) The first one possesses the characteristic pattern of having a penultimate shorter line (see Fig. 2 in Chevillard 2011: 134), which is probably there to announce (or to prepare the listener for) the final “chute”.122

(b) The second, making use of much shorter lines, typically consisting of two long cīrs, and being used, not autonomously, in combination with some akaval “lines”,123 especially in the final section, can be seen in many Puṟanāṉūṟu poems (and also in the Paṭṭiṉappālai, alias vañci neṭum pāṭṭu).

82At some point in time, the society in which lived the descendants of those poets/singers/bards, who had been the masters of akaval and of vañci but who belonged to the pre-theoretical age, must have come into contact with a powerful influence, and this must have been the main reason for the community of poets to progressively equip itself with a set of treatises, of which the Tolkāppiyam is probably the first culmination. One of the hypotheses that has been put forward is that the event which triggered the change (from pre-theoretical to theoretical) was the arrival in the southern part of India of Vedic brahmins. In a description found in the Maṇimēkalai, a later text, we see that those Vedic brahmins had, among many other talents, the mastery of an art (or craft) called chandas, which was considered the never-stumbling leg (kāl) of a powerful entity, described in the following way:

(11a) kaṟpaṅ kai can taṅkā l eṇkaṇ (“the kalpa [‘ritual’] [is] the hand (kai), the chandas [‘metrics’] [is] the foot/leg (kāl), and the ‘number’ (eṇ) [jyotiṣa] [is] the eye (kaṇ)”) // (Maṇimēkalai 27, line 100)
(11b) ṭeṟṟe ṉiruttañ cevicik kaimūk (“the Nirukta (niruttam) which comforts is the ear (cevi), and the śikṣā the nose (mūkku)”) // (line 101)
(11c) kuṟṟa viyākara ṇamukam peṟṟu (“the firm vyākaraṇa ( viyākaraṇam) is the mouth [mukha] (mukam); having received [those six]”) // (line 102)
(11d) cārpiṟ ṟōṉṟā vāraṇa vētak // kāti yanta millai […] (“the Āraṇa-Veda which does not appear as dependent [on anything], has neither beginning nor end”) // (line 103–104)

83Those Vedic brahmins, who were divided in various schools (śākhā), each possessing its recitation manual (prātiśākhya), became themselves (or more precisely their descendants became), after a few generations, native Tamil users. A number of poems in Caṅkam literature have been composed by brahmins and the Tolkāppiyam itself seems to show traces of their influence.124 A significant part of Tamil technical terminology seems to consist in a number of loan translations from Sanskrit into Tamil, as if it had been coined by bilingual Sanskrit-Tamil users.125 In the case of metrical terminology, the dominant metaphor seems to be the metaphor of walking (see 11a), as evidenced by the choice of the term aṭi (step, foot) for naming the top-level component inside “song” (pāṭṭu), the main poetical genre, existing probably from the pre-theoretical time onwards.126 This would correspond to some uses of pada127 (and pāda)128 in several Sanskrit texts.

84An important element, in the analysis of Vedic hymns, was the item called akṣara “syllable”. An examination of its characterisation in a treatise like the The Ṛgveda Prātiśākhya seems to suggest that the formulas which we have examined in section 4, especially those given for nēr and for nērp^u in Chart 4 and Chart 5, may have been inspired by a similar analysis for Sanskrit akṣaras, with the kuṟṟiyalukaram in the {C^u} coda part being treated like a svarabhakti.129 This induces one to see acai, the Level-2 metrical unit, as the metrical item corresponding to akṣara. There is however a difficulty in maintaining the parallelism, due to the presence of nirai and niraip^u, which must have been perceived by Tamil metricians (and poets) as interchangeable with nēr and nērp^u, from the point of view of Tamil word morphology.130 Another way of expressing the fact that nirai was perceived as a [metrical] syllable, would be to say that it was perceived as unnatural to have a syllabic break after the first element in a nirai, i.e. after the initial “(C)V”.131 The difference between the two metrical systems (for Vedic Sanskrit and for Ancient Tamil) is seen in the counting strategies, because the first one relies on counting the akṣaras, whereas the second one does not rely on counting acais, but in counting the “countable” eḻutt^us, as we have seen in sections 5 and 6, and therefore a nirai (and a niraip^u) normally counts for two units of measurement, which makes it into an atypical akṣara.

85The aṭis “steps, [metrical] lines” were analysed and were said to be made of homogeneous elements which were at the same time subtly different. Because of the homogeneity, or balance (or “regularity” [cīr]),132 the elements immediately below the level of aṭi (“step/line”) were called cīr.133 Because of the subtle differences, the elements belonging to the level below the cīr were called acai (“movement”?).134^,135 Finally, even the acais needed to be analysed and their constituent elements were called eḻuttu, because they could be written (or painted).136

86The (newly available) treatises were able to say that there existed normal, well-measured “steps/lines” (nēraṭi, aḷavaṭi) and also shorter “steps/lines” (cintaṭi), such as the characteristic penultimate metrical line inside a pāṭṭu composed in akaval (cf. above). There also existed dwarf metrical lines (kuṟaḷ aṭi), such as the ones characteristic of vañci. And there also existed long lines (neṭilaṭi), such as those which several metricians have cited, such as found in Puṟam 235, “ciṟiyakaṭ peṟiṉē...” (see, for instance TP370i, where Iḷampūraṇar explains that this poems contains an aṭi of six cīrs, and see also Chidambaranatha Chettiyar 1943: 71–72).

87At that time, the dominant dialect for poetry probably had, in its colloquial counterpart, a special phoneme which was an extremely short vowel and which became the source for the kuṟṟiyalukaram (see section 9). It must have been perceptible (as a svarabhakti) and could normally not stand as a syllable kernel,137 and this was the reason for the Tolkāppiyam (or the Ceyyuḷiyal) to include the two items called nērp^u and niraip^u, which later disappeared in medieval treatises (see section 10).

88The cīrs containing two such acais (such as E^u-E^u, E^u-I^u, etc.) must have had a certain special esthetic effect138 and this must have been the reason for the TP-Cey to call them “āciriyavuriccīr”, although they were far from being the most frequent cīrs in āciriyam.139 It was however recommended not to have too many of them inside a 4-cīr line. They had to be interspersed with the simpler cīrs called iyaṟcīrs “natural cīrs”.

18. The changing notion of taḷai

89At the early stage which I have just tried to evoke, in a speculative manner, taḷai was not one of the limbs of poetry, and it seems that the main goal of poetry (producing ōcai “pleasant sound” and memorable texts) was achieved thanks to the cīrs and their acais, and also thanks to the balancing (or weighing) act expressed by tūkku, which seems originally to have been the capacity to recognise how many cīr a specific aṭi should contain, as illustrated by the most well-known case, the centūkku, typical of akaval, but also seen in the coda of vañci poems, namely the succession of a penultimate 3-cīr line and of a final 4-cīr line.140

90The theoreticians, however, were also interested in counting eḻutt^us [i.e. the Level-1 units] and thought that this could be useful for classifying Tamil metres (as the same device had been used for classifying Vedic metres).141

91The word taḷai occurs nine times inside TP-Cey142 but it is not obvious to assign a fixed meaning to it, except perhaps as metaphorically referring to “constraint, fetters”, apt to produce a certain “gait”, if we try to remain within the realm of the metaphor of walking (alluded to in 11a) for poetical composition.

92The veṇpā and the kalippā metres, which are said by TP413i to be closely related143, may derive their existence from several experiments in constraining made by poets-cum-metricians.

93One of the experiments may have consisted in using only a subset of the set of all lines containing only iyaṟcīrs, selecting that subset by adding the constraint that “length (f_n) + length (b_n+1) = 3” [using the variables that stand in figure 1, inside section 3]. The consequence of that constraint is that if S ends with nēr, C must start with a nirai, and vice versa. The further consequence of that constraint is that the line length will be equal to the sum of number nine (= 3x3) and of the length of the first acai in the first cīr, to which is added the length of the last acai in the last cīr (see fig. 3).

94In the language of medieval metricians, such a line would be said to contain three occurrences of iyaṟcīr veṇṭaḷai. In the language of the Tolkāppiyam, it seems to be called an iyaṟcīr veḷḷaṭi (see TP368i).144 It should be noted that the TP-Cey also uses the expression “veḷḷaṭi” (inside TP456i) while defining kaliveṇpāṭṭu. If to the constraints already enumerated one adds another alliterative constraint called aṭi etukai, which is one among a group of possible features collectively called toṭai (see Chevillard 2011: 141), the consequence will be that b₁ and b₅ (on Fig. 3) will be quantitatively equivalent145 and that, if the constraint “length(fⁿ) + length(b_n+1) = 3” also holds between two metrical lines (i.e. for n = 4), we shall obtain that the total length of the iyaṟcīr veḷḷaṭi will be twelve.

95The general context of several other experiments may have consisted of using long cīrs, made of three acais instead of two. Those cīrs were called veṇcīrs (the expression is found four times inside TP-Cey)146 or veṇpāvuriccīrs (TP331i).147 They are very prominent in an important anthology such as the Kalittokai (see Charts in Appendix B) and also in a work such as the Kuṟaḷ (see Chart 8). This massive introduction of a new type of cīr could not be without consequences on the overall picture of Tamil metre because many new types of lines could be envisioned, some consisting of homogeneous elements (only iyaṟcīrs or only veṇcīrs),148 and some consisting of heterogeneous elements (i.e. a mixture of those two categories of cīrs). Regarding the latter case, it is very likely that the formulation of a shortcut concerning a method for dealing with them was the original intention of TP337i (veṇcī rīṟṟacai niraiyacai yiyaṟṟē) which can be translated as “the final acai of a veṇcīr behaves like the nirai [of an iyaṟcīr]”), if we follow Iḷampūraṇar’s interpretation, and do not follow Pērāciriyar’s extremely tortuous interpretation (under TP341p).149

96Inside that framework, an experiment for which the keyword is kalittaḷai, seems to have consisted in imposing a number of quite difficult constraints, such as, for instance, forbidding the use of the two cīr patterns designated by tēmā and puḷimā.150 This was certainly a very bold step because, as seen in Chart 9, those two types of cīr account for more than 60% of all cīrs in normal poetic usage. But it can be seen in the two Charts B1 and B2 (found inside Appendix B) that it is possible for a poet to accomplish that, as seen from the absence of those two cīr patterns in the ninety-six cīrs which constitute the taravu and the three tāḻicais of Kali 25. Formulating it otherwise, the “kalittaḷai” experiment seems to have been an experiment in “mannerism” (i.e. rising to the challenge of expressing meaning under rather artificial constraints).

97If we accept this possible background, we may also hypothesise that the veṇpā experiment may have been the “plain” counterpart151 of the kalippā mannerism experiment.152 The experiment in mannerism was continued in all the varieties of kali, as well as in the paripāṭal types. It is also seen in the topic of vaṇṇam. Therefore, whenever we see the word “taḷai” inside the Tolkāppiyam, we don’t have to think that it necessarily has the meaning which it acquired later, and for which the standard presentation will be now given, in the coming section.

19. The “standard doctrine” concerning taḷai

98That standard presentation is found inside the YA153 and the YK154 and can be summarised in a flow chart (see page 308, Figure 4), which details the various possibilities for the succession of two cīrs taken from a set of twelve possibilities (four cīrs having two acais and eight cīrs having three acais).

99I should hasten to say that such a presentation of the “seven taḷai” is not a traditional one. Nowhere in commentaries on metrical treatises do we find the instruction to compute “length (f) + length (b)”. What I have introduced inside Figure 4 under the condition “L = 3” is introduced by Kākkaipāṭiṉiyar,155 in a veṇpā, as a case of vikaṟpam and what I have introduced as the two conditions “L = 2” and “L = 4” is introduced, by the same, also in a veṇpā, as absence of vikaṟpam.156 The same data can be presented in chart form, in the following way:

100As we can see, there is a dissymetry in the treatment of the S (the “standing cīr” of Figure 1) and C (the “coming cīr” of Fig. 1), because C does not play any role in the specifying of the taḷai. There is however a more sophisticated version of the taḷai classification in which every taḷai (= T) is subdivided into two species: one ciṟappuṭai T (approx. “first-class T”) and one ciṟappil T (approx. “second-rate T”). The “first-class” (ciṟappuṭai T) is obtained only if C and S belong to the same category (i.e. are both iyaṟcīrs, or both veṇcīrs, or both vañciccīrs).157 The net result can be presented in chart form in the following way (see Chart 18B), with a plus sign as exponent “⁺” indicating the “first-class” items and a minus sign as exponent “^-” indicating the “second-rate” items.

101Having thus given the technical definition of taḷai, we must now briefly examine the use which is made of it, first from a descriptive point of view (which taḷai-s are found in existing literary works) and then from a prescriptive point of view (which are the instructions given by metrical treatises to would-be poets in which the taḷai is involved). Concerning the first point of view, the descriptive one, it seems to me, on the basis of the texts which I have examined, that the existings literary texts fall roughly into three groups, A, B and C, which can be presented in tabular fashion (see Chart 18C).

102Those three groups can be characterised in the following way:

• Group A: these are poems, or components inside poems, in which only taḷais 3 and 4 are used, which are collectively referred to as veṇṭaḷai. Under that category fall: the Kuṟaḷ, some Kalittokai poems [such as the kaliveṇpāṭṭu type poem (Kali 6) for which statistics is provided in Appendix B], many Tamil bhakti period poems, such as the three old antātis and many other works (see Chevillard, forthcoming). Those items can be said to illustrate the metre called veṇpā.
• Group B: these are poems or segments inside poems in which taḷais 6 and 7 are found, along with all the other taḷais. The presence of those taḷais is a mechanical consequence of the presence of the four vañcic cīrs (EEI, EII, IEI and III), which are normally avoided in other types of poems. Those items will be said to represent vañcippā.
• Group C: these are poems or components inside poems in which we have a mixture of the taḷais 1 to 5, with no clear prominence. In this group, a sub-group can be distinguished, containing poems in which taḷai 5 (i.e. kalittaḷai) is prominent. Many poems of this type are found in the anthology called Kalittokai [such as the ottāḻicaikkalippā type poem (Kali 25) for which statistics is provided in Appendix B], but that anthology also contains poems belonging to group A.

103On the basis of this classification, what is the situation of āciriyat taḷai, if we use that designation for collectively referring to taḷais 1 and 2? The choice of Yāpparuṅkalak Kārikai-22 for illustrating that type of linkage is the famous Kuṟuntokai 1 poem, which reads:

(KT1) ceṅkaḷam paṭakoṉ ṟauṇart tēytta // ceṅkō lampiṟ ceṅkōṭṭⁱ yāṉaik // kaḻaṟoṭic cēey kuṉṟaṅ // kurutip pūviṉ kulaikkān taṭṭē.

104In this poem, the first two lines are indeed an example of strict use of āciriyat taḷai.159 However, neither the third line, nor the fourth line are examples of strict āciriyat taḷai, because kaḻaṟoṭic is followed by cēey, while pūviṉ is followed by kulaikkān (which, in both cases, illustrates veṇṭaḷai). We would need of course a detailed statistical study of the Caṅkam corpus in order to make absolutely precise statements, but it is my perception that the connection between the “metre” called āciriyappā and the taḷai called āciriyattaḷai is a nominal connection, which was built on the polarity between a specific mode of recitation initially called akaval, which was associated by the TP-Cey to the name āciriyam and another mode of oral performance, initially unnamed. It is probably because veḷḷai “white [metre]” and āciriyam were perceived as polar opposites that āciriyattaḷai was defined (in TP362i) as something which can appear as the symmetrical of iyaṟcīrveṇṭaḷai.160 We have in the TP-Cey a succession of four cūttirams, which say:

(M1) akava leṉpa tāciri yammē “that which is called profération is [to be in] āciriyam”(TP386i);
(M2) atāaṉ ṟeṉpa veṇpā yāppē “they say that what is not [profération] is to be composed in white metre (veṇpā)” (TP387i);
(M3) tuḷḷa lōcai kaliyeṉa moḻipa “they say that the springing sound [is to be in] kali (TP388i);
(M4) tūṅka lōcai vañci yākum “the suspended (or balancing?) sound is in vañci” (TP389i).

105According to Iḷampūraṇar, these four cūttirams are part of a group [TP386i-TP392i] which explains the dimension of poetry called tūkk^u “weighing, balancing” (see beginning of section 17). The exact meaning of tūkk^u seems however to have later become partly blurred.161 The unnamed mode of oral performance corresponding to veṇpā was later called ceppal ōcai “telling sound”,162 but such a designation is unknown to the TP-Cey, although its commentators use it all the time.

20. Conclusion: appraising the three interpretations of TP357i

106I have certainly not yet provided enough information for the reader to have a clear idea of the totality of the long history of Tamil metrical theories. However, since this article has been built with a specific cūttiram, namely TP357i, as its focus, I shall now try briefly to re-examine the three interpretations proposed for that cūttiram by Iḷampūraṇar, by the YV commentator and by Pērāciriyar (closely followed by Nacciṉārkkiṉiyar) for the figures “seventy” and “six hundred and twenty-five”. As has been noted, the YV commentator replaces the expression eḻupatu vakaiyiṉ by eḻupatu taḷaiyiṉ and Pērāciriyar uses the reading eḻupatu vakaimaiyiṉ. Iḷampūraṇar explains (and demonstrates concretely) that the eḻupatu vakai are the “seventy [possible] subdivisions” which are obtained when one enumerates all the sequences that can be described by Fig. 1. He is thus in the same type of exhaustive enumeration as the creators of the descriptive “seven taḷai” grid presented in Fig. 4 and in Charts 18A and 18B. The difference however is that he has five types of cīr to deal with (instead of three), and therefore his list has “seventy items” and not seven. He does not call those items “taḷai” because for him taḷai is probably a prescriptive term, i.e. the instruction given to the poet to follow a set of constraints, rather than a descriptive term (i.e. one element inside the exhaustive set of all possible connections). Regarding his interpretation of the “six hundred and twenty-five [lines]”, which are the ultimate amplification (virittal) of the “five types of line” (aivakai aṭi), he is also in the same logic of exhaustive enumeration and it is unlikely that he would have considered it useful to really enumerate them.163

107The YV author seems to have considered it his duty to somehow create bridges between the modern, simplified, theories of metre (contained in the YA) and what is already for him a long history of metre description. In his time, many technical terms of the old system had acquired a new meaning but he still manages to preserve much of the ancient lore thanks to his citations of many treatises. For him, the Tolkāppiyam is only one among several authorities, although certainly a highly respected one. When it comes to his explanations of the “seventy” and the “six hundred and twenty-five”, I need not repeat them here, and they are not part of a complete commentary of the TP-Cey. They appear as an impressive attempt by a scholar at creating a new interpretation of TP357i (if we take Iḷampūraṇar’s interpretation as faithful to the original intention). The reason for this attempt might be the desire to emulate the complex mechanisms developed by Sanskrit metricians for classifying metrical patterns, in relation with the ṣaṭ pratyaya “problématique”. It is to be noted that an extensive Tamil version of the same topic (the āṟu teḷivukaḷ) is also included in the YV (see YV-1998: 535–46). Both those sections of the YV contain a number of embedded mnemonic veṇpās, which deserve to be studied for themselves.164

108Pērāciriyar is in a logic of “restauration” of the orthodoxy, as can be seen for instance in his discussion of the question whether veṇpā possesses ten loci or only eight. He is a believer in the existence of the First Caṅkam and in the fact that Akattiyar (Agastya) composed a grammar for it. He considers the author of the Tolkāppiyam as the disciple of Akattiyar. He is also a very astute scholar, capable of producing very subtle remarks (some of which seem very far-fetched). A real appraisal of his work would require, as an absolutely necessary preliminary, a scrutiny of all the occasions when he gives an interpretation for a cūttiram that is completely antagonistic to the interpretation provided by Iḷampūraṇar (see for instance n. 149). It would also require to be provided with a complete statistics of at least all the cīrs and all the taḷais used in the Eṭṭut Tokai and the Pattup Pāṭṭu. He has been criticised, along with Nacciṉārkkiṉiyar, for inventing an inexistent metrical category.165 Both of them lived in a time when the existing poetical corpus existed as a set of classics, which must of course be dealt with in a conservative way.166 They seem to demonstrate that one can be conservative in an innovative way. Their collection of 625 patterns called kaṭṭaḷaiyaṭi is mathematically beautiful but poetically imaginary in a great part (inside the attested corpus of Classical Tamil literature), as far as āciriyam is concerned. Pērāciriyar is really in need of the Caṅkam legend (and its two floods) for explaining that a great part of what he has described (and whichI have summarised in Chart 17) is not to be seen in the corpus found in the present in which he lives and has (mostly) not been used by the poets of the kaṭaic caṅkam.