1. Introduction

The Puṟam, or historical, poems of Old Tamil Caṅkam poetry deal in the main with the Pāṇṭiya, Cōḻa and Cēra dynasties. On the basis of the absence of the Pallavas in these poems, the poetic tradition has been dated before the appearance of that dynasty in the northern part of Tamilnadu in approximately the fourth century A.D. However, as I have recently tried to show, we are dealing with historical poetry evoking scenes from the past of Tamilnadu before the arrival of the Pallavas (Tieken 2001). In their cultural politics the Pallavas were focused on North India and its Sanskrit tradition. By contrast, Caṅkam poetry aims to depict a society which was still more or less free from North Indian influences and Sanskrit culture. Furthermore, by presenting bards from an earlier period praising kings in songs in Tamil, Caṅkam poetry aimed to provide that language with an ancient literary tradition. As historical literature Caṅkam poetry should not be dated in but after the period it describes.

1The majority of the approximately 2364 Caṅkam poems is formed by Akam, or love poetry. The Akam poems depict life in small, dreary villages. It is not, as has been generally assumed, a poetry of the village but about it. The point of reference is a leisure class living in cosmopolitan towns and cities, which could afford to make fun of the poor villagers. An investigation of the various Akam texts in the Caṅkam corpus has shown that we are mainly dealing with imitations of specific genres of Kāvya. The main conclusion of my earlier study was that Caṅkam poetry was a written literature which belonged to the Kāvya tradition and which closely adhered to the conventions of that tradition. This is also the case in the language of the poetry. Thus, it was found that the Akam texts correspond typically to texts in Prākrit. When Tamil, a local, spoken language, came to be used as a literary language, it was apparently assigned the role of a Prākrit. Conformingly, the historical Puṟam poems typically present local history in contrast to epic mythology, which was the domain of Sanskrit. This use of Tamil corresponds to what we see in, for instance, the Velvikudi and Dalavaypuram inscriptions of the Pāṇṭiyas, with their praśastis in Sanskrit, presenting epic mythology, followed by praśastis in Tamil providing detailed local history. On the basis of this evidence it was concluded that Caṅkam poetry was a creation of the same Pāṇṭiyas who had been responsible for these two inscriptions and therefore cannot be earlier than the end of the eighth or the beginning of the ninth century.

2According to this new dating Caṅkam poetry is almost contemporary with Bhakti poetry, if not later than that, as some scholars date the beginning of Bhakti poetry as early as the fourth century when the first great temple centres appeared in Tamilnadu. However, in this connection I argued that the date of Bhakti poetry is to be kept strictly apart from the date of the first great temple centres in Tamilnadu or that of Bhakti as a movement. In fact, going by the references in inscriptions to the Śaiva canon we would have to conclude that this devotional literature is not much older than the ninth or tenth century. At the same time, one of the conclusions of my study was that the chronological relationship between Caṅkam poetry and Bhakti poetry is difficult to establish with certainty. In fact, Caṅkam poetry might well be later than Bhakti poetry. In any case, Bhakti poetry seems to subsume a previous, "classical", literary tradition in Tamil, and Caṅkam poetry, as a poetry supposedly from the past, might have been composed to serve that very purpose.

3Among the texts of the Caṅkam corpus generally a distinction has been made between early, truly bardic poetry (in, e.g., Kuṟuntokai) and late, classical poetry (in, e.g., Kalittokai). With the late dating of Caṅkam poetry in the eighth or ninth century the basis of this distinction, which was not very obvious in any case, has disappeared completely, in the sense that the corpus does not include any really bardic poetry at all. This is not to say that there could not be earlier or later texts. For instance, a possibly later addition to the corpus is the Patiṟṟuppattu. In this connection I pointed out the curious format of the text, which resembles that of Bhakti poetry. Furthermore, while, for instance, the Puṟanāṉūṟu deals with the Pāṇṭiyas, Cōḻas and Cēras, the Patiṟṟuppattu is exclusively devoted to the Cēras. Accordingly, when all is said and done, it could be that the version of the Caṅkam corpus which we now have was compiled under the auspices of the Cēras, which is precisely what the traditional legend of the origin of Caṅkam poetry wants us to believe (Tieken 2001: 202). As such, the Patiṟṟuppattu is the product of the adaptation by the Cēras of the culture and poetic tradition of their eastern neighbours, the Pāṇṭiyas. This acculturation is otherwise recorded in the Cilappatikāram, which text is thus to be dated in or after the period in which that feat took place, and not, as has been assumed so far, in either the second or fifth century A.D. (Tieken 2001: 202-208).

4The conclusion regarding the late date of Caṅkam poetry goes completely against the current view of the language of the Caṅkam poems, which is generally held to represent a stage in the development of the language much earlier than that of the eighth or ninth century. In order to distinguish Caṅkam Tamil from the language of the Bhakti poems (supposedly fourth century) and the inscriptions (from the seventh century onwards) it has been labelled Old Tamil, or sometimes even Early Old Tamil.

5Compared to Bhakti poetry or the inscriptions, Caṅkam Tamil indeed shows a number of unique and presumably archaic linguistic features. But what is unique or archaic is not necessarily old. However, given their occurrence in Caṅkam poetry the need to prove this has generally not been felt. As a result, other possible explanations for the linguistic peculiarities were simply not considered, let alone explored. This is precisely what I intend to undertake here for at least a number of such peculiarities, namely the various non-past tense suffixes. However, before doing so I will have a closer look at the linguistic evidence in general which has been advanced in favour of an early date for Caṅkam poetry.

2. The linguistic evidence for an early date of Caṇkam poetry

The following attempt to review the linguistic evidence used in the early dating of Caṅkam poetry is considerably frustrated by the fact that as far as I know there is no systematic treatment of this evidence. For instance, apart from mentioning the rareness of loanwords from Sanskrit, Zvelebil provides a list of no more than five items (Zvelebil 1992: 116, n. 70. See also Zvelebil 1973: 36). Characteristically of the careless way in which the matter is treated, the list ends in an "et cetera" and would represent only "a few diagnostic examples" (Zvelebil 1973: 36).

6One of the more obvious archaic feature of the language of the Caṅkam poems, however, is indeed the rareness of loanwords from Sanskrit (see, e.g., Zvelebil 1975: 89). As we know from inscriptions, this indeed does not agree with the linguistic situation in the eighth and ninth centuries. It should be added, however, that the rareness of loanwords from Sanskrit does not argue against the late dating of Caṅkam poetry either, for the characters who are made to speak in this poetry all belong to a period well before the large-scale introduction of North Indian Sanskrit culture into Tamilnadu under the Pallavas; or else they belong to small villages as yet untouched by Sanskrit culture. Thus, it could be argued that the poets avoided Sanskrit words on purpose, namely as part of their attempt to create a convincing picture of a local Tamil culture. The point is that the interpretation of the absence of Sanskrit loanwords in Caṅkam poetry itself depends on the interpretation of the nature of the poetry. As such the rareness of Sanskrit loanwords in Caṅkam poetry does not constitute independent evidence in the dating of that poetry.

7According to Zvelebil, the Caṅkam poems would contain "many forms which are obviously more archaic, earlier in development, than analogical forms found in the devotional hymns" (Zvelebil 1992: 116, n. 70 see also Zvelebil 1973: 36). In this connection Zvelebil refers to Old Tamil yāṉ "I" against "its undoubtedly later form nāṉ", the use of the deictic prefix i-  "this" instead of the later deictic adjective inta, and the absence of the double plural suffix –ar-kaḷ and the present tense morph -k(k)i(ṉ)ṟ-, which are both common in Bhakti poetry.1 However, on closer inspection the evidence is of an ambiguous nature. Take, for instance, the first person pronoun nāṉ. I do not wish to deny that the form yāṉ is more archaic. The point is, however, that in Dravidian the forms with initial n- are generally attested only relatively late. Thus, the Kannada inscriptions in the seventh century have only ān. It is only from the tenth century onwards that beside this ān we find instances of nān (Zvelebil 1977: 40). Something similar is seen in Telugu: beside a large number of occurrences of ēnu Nannaya´s Mahābhāratam (12th century) counts only four instances of nēnu (Zvelebil 1977: 46). Given this situation we should reckon with the possibility that nāṉ has become common only fairly late in Tamil as well. This means that its absence in Caṅkam poetry, though representing an archaic situation, is not necessarily such an early feature as it has been assumed to be so far.

8Another supposedly archaic feature of the language of Caṅkam poetry is the absence of the double plural suffix -ar-kaḷ. However, the problem is not the absence of this combination of the personal plural suffix –ar with the neuter plural suffix –k(k)aḷ, but the rareness of the latter suffix in Caṅkam poetry on its own,2 which is curious as the suffix –k(k)aḷ is otherwise widely attested in Dravidian (e.g. Zvelebil 1977: 12-15) and therefore must almost certainly have been present in the language of the poets. The situation could be summed up by arguing that for some reason the authors of Caṅkam poetry avoided to mark the category of the neuter plural, a phenomenon to which I will return below. The point I would like to make here is that the absence of the double plural suffix –ar-kaḷ does not suggest an early date perse.

9Yet another so-called archaic feature mentioned is the deictic prefix i- in, e.g. i-vvūr "this village". We are indeed dealing with a curious construction here. That is to say, while the prefix itself is still found in Modern Tamil in, e.g. i-ppoḻutu as well as in a relatively late Pāṇṭiya inscription (SII XIV, no. 190, line 11: i-vvaṉaivōm), what is exceptional in Old Tamil is its occurrence before any noun. The Modern Tamil counterpart of i-vvūr is inta ūr. However, there seems to be no information on how long the "free" use of i- as in i-vvūr has persisted in Tamil. Secondly, no information is available so far regarding the nature of the construction, or whether it is found in the other Dravidian languages as well and, if so, in what kind of sources (e.g. literary sources, spoken language). For all we know, the construction might have been restricted to Caṅkam poetry. Therefore, given this state of affairs I do not see how one can draw any conclusion concerning the dating of Caṅkam poetry from the occurrence in it of formations like i-vvūr.

10To the absence of the present tense suffix – (k) ki (ṉ) ṟ- in Caṅkam poetry I will return below.

11As said, the items enumerated by Zvelebil represent only "a few diagnostic examples", the list ending in an "et cetera" (Zvelebil 1973: 36). Other possibly relevant material has thus to be gleaned from this and other scholars´ publications. As far as I see, the list may be enlarged by at least the following items, which will be discussed below: the non-past tense marker –t(t)-, the use of verbal nouns functioning as predicates, and certain types of periphrastic constructions.

12To begin with the non-past tense marker –t(t)- (e.g. aṟiti Puṟ. 36: 2, āṟṟutir Puṟ. 58: 20, cāṟṟutu [m] Puṟ. 104: 1, kiṭatti Puṟ. 272: 5), P.S. Subrahmanyam characterizes it as "a very old tense type in Tamil… appearing in the most ancient extent texts" (Subrahmanyam 1971: 243; see also 313-316). Below I will return to this suffix in more detail. What may be noted at this stage is that the instances of this tense suffix are not restricted to "early" Caṅkam poetry, as some occasional instances of it are also found in "later" Bhakti poetry (see Tēvāram VII (Cuntarar) 2,1 añcutum and 2,6 pāṭutir).3

13As for the use of verbal nouns as predicates, I am concerned in particular with verbal nouns in –ku (e.g. nōkō yāṉē Puṟ. 116: 9, celkam Puṟ. 207: 1) and – (v) al (e.g. celval yāṉē Puṟ. 162: 7). It is generally assumed that we have to do with a genuine Dravidian feature which has somehow survived in Caṅkam poetry, thus proving the archaic nature of this poetry. In this connection I would like to quote Bloch:" [i] l semble donc finalement que le système flexionnel de type pronominal se soit développé secondairement. Il succède à l´usage de noms verbaux capables de sujet pronominal au nominatif. On en trouve d´assez nombreux exemples dans la vieille poésie tamoule" (Bloch 1946: 45). Bloch has been quoted with approval by Zvelebil, who adds" [t] rue to this historical and comparative observation, he [Bloch] states, that the use of verbal nouns in predicative function had been once predominating in the ancient stages of evolution of all Dravidian languages" (1957: 653). It is not clear, though, on what this latter conclusion is based. In fact, it appears as though the archaic nature of the construction has been based solely on the early date of Caṅkam poetry, which, as I have argued elsewhere, is something which has still to be proved. What is clear, though, is that celval and kāṇku are unique to Caṅkam poetry, a fact already observed by Bloch (see above) and reiterated by Zvelebil (" [t] his use has been found especially in Old Tamil" Zvelebil 1957: 653).4 For all we are in a position to know, celval and kāṇku in the predicate function may be innovations in the poetic language of Caṅkam. This is a point to which I will return below.

14A third and final supposedly archaic feature is formed by phrases like pāṭukam vammiṉō "let us go to sing his praise" (literally "we (who) will sing his praise, let us go") (Puṟ. 32: 6), niṟ kāṇku vanticiṉē "I have come to see you" (lit. "I (who) will see you (I) have come") (Puṟ. 125: 4), uṟṟaṉai pōṟi "you look as if you have experienced" (Aiṅk 58: 3), paṟṟiṉaṉ parivāṉ "having grabbed, he will pull" (Kali 79: 12), and añcēṉ peyarkkuveṉ "not being afraid, I will hide myself" (Naṟṟ 362: 9). Phrases of this type include many instances of negative periphrastic constructions like Puṟ. 31: 11 celvēm allēm "we will not go" ("we will go, we are not such persons"). To these examples may also be added those of the following type, which involve the use of a participial noun, e.g., tarumār "he who will bring or collect "5, as a kind of infinitive. See, e.g., Kuṟ. 216 avarē … viḻupporuḷ tarumār… kāṭiṟantōr "he crossed the jungle in order to bring back richess".

15Phrases of this type play an important role in Steever´s reconstruction of so-called Serial Verb Forms (SVF) in Dravidian (Steever 1988: 42ff). The instances from Caṅkam poetry are quoted by him as evidence of the archaic nature of the type of construction. If I understand Steever correctly, he interprets these phrases in Caṅkam poetry as archaic forms directly continuing a Dravidian type of construction, which, however, was somehow lost in the medieval period.

16It should be added, however, that similar types of sentences are apparently still found in texts "later" than Caṅkam, namely in Bhakti poetry and in Kampaṉ´s Rāmāyaṇam. Thus, Steever (1988: 42-3) quotes ceṉṟēṉ allēṉ from Tēvāram and kēḷēm allēm kēṭṭaṉam from Kampaṉ´s Rāmāyaṇam. Furthermore, if we turn to the literatures of the surrounding languages, we may find instances of the negative periphrastic construction in the tenth-century Kannaḍa text, the Vikramārjunavijaya (e.g. mīruvar allar "they will not transgress "; see Steever 1988: 55). Clearly, if the construction dates back to before the fourth century, which is generally believed to be the latest date of Caṅkam poetry, it was still in use in the tenth century.

17It is possible that the instances of the formation in question in Bhakti poetry or the Rāmāyaṇa constitute deliberate archaisms. However, if so, the same might apply to the instances in Caṅkam poetry as well. For, as already noted by Steever himself (Steever 1988: 44 ff), in Caṅkam poetry for most of the above constructions there are alternatives. For instance, beside celvēm allēm "we will not come" there is cellalam "we do not go" (Puṟ. 101: 1), with negative al suffixed directly to the verb stem, or uḷḷār "he did not think" (Kuṟ. 16: 1), with the negative suffix –ā; and beside kāṇku vanticiṉē "I have come to look" there is uraikka vantatu "it came to declare" (Puṟ. 28: 7), with the infinitive uraikka. In this connection it is also interesting to note that, for instance, the periphrastic negative construction in both the Tamil (celvēm allēm) and Kannaḍa texts (mīruvar allar) is restricted to literary texts composed in the Kāvya tradition. Given this distribution of the instances the question may arise if we could not be dealing with cases of learned paraphrasis.

18When all is said and done, the linguistic evidence for the archaic nature of Caṅkam poetry is very meagre. The forms and features which have been quoted in this connection are no doubt peculiar but are not necessarily old.6 The conclusion that the language of Caṅkam poetry is typical of the beginning of our era, is, to say the least, premature.

19In what follows I intend to have a closer look at the non-past tense forms found in Caṅkam poetry. On closer consideration some of these formations are not old at all but are still used in the present-day non-literary languages of Tamilnadu. Some other forms appear to be restricted to Caṅkam poetry only, which does not make them old either. Rather, some of the latter forms might well be artificial constructions. As I will argue, the implications of the rareness of the present tense morph –k (k) i (ṉ) ṟ- in Caṅkam poetry might be a matter of sociolinguistics. When all is said and done, the main problem posed by the various non-past tense formations is their great number and variety. However, as I will try to show, the Caṅkam poems are not unique in that, but have parallels in late Apabhraṁśa texts. The same applies to the occurrence of artificial formations. However, before turning to the language of the Caṅkam poems I will briefly sidestep into the languages of the inscriptions and Bhakti poetry, which furnish our main sources for a Tamil which, at least according to my reckoning, is more or less contemporary to that of Caṅkam poetry.

3. The language of the inscriptions

One of our main sources of evidence for the Tamil of the period under investigation (eighth and ninth century) are inscriptions. In this connection I would like to leave aside the so-called Tamil Brāhmī inscriptions from the beginning of our era. These inscriptions are generally too brief to allow a fruitful comparison with the language of Caṅkam poetry.7 Tamil made its appearance in inscriptions again, that is after a break of several centuries, with the Pallavas. Before the seventh century the inscriptions of this dynasty were entirely in Prākrit and Sanskrit. The first inscription in which Tamil appeared was the Paḷḷaṅkōyil inscription of Siṁhavarman III from the seventh century, in which, however, its use was restricted to the description of the technical details of the donation. For the praśasti Sanskrit was used. The first praśastis in Tamil are found in the Velvikudi and Dalavaypuram inscriptions of the Pāṇṭiyas in the early part of the ninth century. They occur side by side with a praśasti in Sanskrit. However, during the Cōḻa period Tamil had become the sole language of the inscriptions.8

20Compared to Modern (Written) Tamil the language of the Pallava inscriptions is strikingly modern.9 Thus, if we concentrate on the verbal tense system, the Pallava inscriptions have a past tense with -t (t)/nt/i-, a future with -v/p (p)-, and a present with –k (k) ki (ṉ) ṟ- as well as -āniṉṟ-. Of the three tenses the past tense is clearly the most frequent one and needs no comment. By contrast the present tense in the inscriptions is rare. This, however, may be explained with reference to the function of inscriptions, which simply do not provide many occasions for the use of the present tense but mostly mention past decisions and promises concerning future behaviour. In what follows I would like to discuss some peculiarities of the non-past tense formations found in the inscriptions.

21To begin with the composite present tense morph -āniṉṟ-, it is found in the Mahabalipuram inscription of Nandivarman II (Mahalingam 1988: no. 92, ll. 3- 4): celāniṉṟatu aṟupattu a (i) ntāvatu, "the current 65th year" and the Kanchipuram inscription of that same king (ibid. no. 93, section K, l. 110): rājyañ cevyāniṟka, "while (they) were ruling". The same inscription, which provides short descriptions of a series of historical reliefs, contains two interesting instances of the present tense verbal noun. Thus, in section H we find: pala kiri-nati-vanakahana (m) ka (ḷai) kkaḻintu varukiṉṟamai kkē(ṭṭu), "hearing him coming after he had crossed many mountains, rivers, woods and thickets", and in section I: avar āṉai varukiṉ (ṟa) mai kēṭṭu, "hearing his elephant´s coming". All other instances of the present tense which I have been able to trace concern participles, e.g. varukiṉṟa in ibid. no. 132, ll. 18-19, niṉṟaruḷukiṉṟa in no. 136, l. 7, viṟkiṉṟa in nos. 171, ll. 26- 27) and 224, l. 2, āḷkiṉṟa in no. 171, l. 30, and irakṣikkiṉṟa in no. 246, l. 17.

22Especially the two instances of the verbal noun from the Kanchipuram inscriptions show that the present tense refers to a simultaneous action or situation: "hearing that (at that very moment) he/it was coming". In this connection an interesting passage may be found in an inscription from the twelfth year of king Parakēsarivarman (SII III, 3-4, no. 99), in which (iv) vāṇṭu erivārikañ ceykiṉṟa erivārikapperumakkaḷōm, "we the great people of the tank-committee who are taking care of the tank this year" in line 2 contrasted with a (v) vavvāṇṭu erivārikañ ceyyum erivārikapperumakkaḷ, "the great people of the tank-committee who will take care of the tank every other year (after us)" in line 5.11

23Both morphemes, -āniṉṟ- and –k (k) i (ṉ) ṟ-, are found in the Pāṇṭiya and Cōḻa inscriptions as well. For the Pāṇṭiya inscriptions, see, e.g. EI XVII, no. 16, ll. 103- 104: maṟṟavaṉṟaṉ rājyavatsalam mūṉṟāvatu celāniṟpa, SII XIV, no. 37, ll. 15-16: maṉṟāṭukiṉṟatu, ll. 51-52: ceykiṉṟārum, no. 206, ll. 14 and 19: varukiṟa, and EI XX, no. 3C, ll. 7, 9 and 10: viṟkkiṉṟa. For the Cōḻa inscriptions, see Agesthialingom and Shanmugam (1970: 62-3).

24The majority of the instances of the present tense morphs –k (k) i (ṉ) ṟ- and - āniṉṟ- seems to consist of participles and participial nouns. Predicates appear to be rare, which should, however, as said, be considered in the light of the function of inscriptions.

25As to the future tense, the situation in the inscriptions is peculiar. That is to say, "simple" futures like paṭuvār "they (will) fall" (Mahalingam 1988: no. 226, ll. 8-9) are rare. What we commonly find instead are periphrastic constructions consisting of a future participial noun followed by the past tense of the verb ā: e.g. erippōm āṉōm (ibid. no. 137, ll. 21-22) and erippōm ā (y) iṉōm (SII XIV no. 12, ll. 10-11). As indicated, "simple" futures are rare. Cases in point, beside paṭuvār quoted above, are arcippōm (Mahalingam 1988: no. 139, l. 12), ratṣippār (ibid. no. 98, l. 2. See also irakṣippār in SII XIV no. 16, ll. 21-22), peṟuvār and paṭuvār (Mahalingam 1988: no. 110, ll. 44-45 and 51-52), and erippōm (SII XIV no. 190, l. 8). However, in the latter case we may be dealing with an instance of omission by anticipation, erippōm in line 8 being followed by kuṭuppōm āṉōm in line 9.

26This type of phrase seems to be typical of the future. Instances involving other tenses are rare. The few instances I have been able to trace involve periphrastic negatives like pēṟātōm āṉōm (Mahalingam 1988: nos. 214, l. 3 and 241, l. 13) and kāṭṭātōm āṉōm (ibid. no. 246, l. 8) and a rare instance of a periphrastic past tense, as in kuṭuvittōm āṉōm (ibid. no. 206, l. 5).12 Moreover, a vast majority of the instances of the periphrastic construction involve the first person plural, which is also the most common form in the inscriptions ("we the people of the sabhā of X will do (or: promise to do) this or that"). The instances showing a simple future without the verb "to be" seem to involve in particular the third person plural. For the Cōḻa inscriptions, see, for instance, iṟaippār, paṟippār and tirumeḻukiṭuvār in SII XIII no. 14, l. 5.13

27In trying to understand the periphrastic future (first person) it is important to note that it is not the only type found but that it occurs side by side with the simple future (third person). A second point is that in almost all cases of the periphrastic future we are dealing with what may be characterized as a solemn promise. It is tempting to argue that the periphrastic formation was introduced because the simple future, which in the contemporary spoken languages was presumably also used for the habitual, was somehow felt to be insufficiently marked to express a promise. However, the next question would then be why the periphrastic construction was considered to be better suited for the latter function. For this, I think, we may turn to Sanskrit, in which solemn promises are likewise expressed through a periphrastic construction, namely dātāsmi, i.e. dātā + asmi. In this connection, we should bear in mind that the authors who introduced the periphrastic future in the Pallava inscriptions must have been heavily indebted to Sanskrit. The reason for this is that inscriptions were originally the domain of Sanskrit (or Prākrit). Tamil was fitted in only later, replacing or substituting Sanskrit or Prākrit, which must have involved a contrastive study of the grammars of the respective languages. This might have led to the observation that in Tamil the future overlapped with the habitual and, as a next step, to the creation in Tamil of the periphrastic future after Sanskrit dātāsmi.

28As to the instances of the third person, which mostly involve the simple future, it should be noted that in Sanskrit the third person is likewise simply dātā, that is, without following asti.

29At the same time, however, we may come across occasional periphrastic formations of the third person (see note 13), which would not fit into this scenario. Likewise, we may find occasional instances of the negative periphrastic construction (pēṟātōm āṉōm) or of the past tense (kuṭuvittōm āṉōm). However, as to these instances, including those of the third person, the question may be raised if we could not be dealing with the results of the accidental extension of the periphrastic construction in promises to other contexts due to the great frequency of that construction.

30Whether or not erippōm āṉōm is a loan translation from Sanskrit, it is peculiar to inscriptional Tamil. As far as I know, it is not found in, for instance, contemporary Bhakti poems. Apparently we are dealing with a kind of Tamil developed especially for the use in inscriptions. The fact that this language, which goes back to the Pallavas of the seventh century, remained in use among the later Pāṇṭiyas and Cōḻas shows that in the inscriptions we are dealing with a learned language.

4. The language of the Bhakti poems

In the absence of overall studies of the language of the Bhakti poems our view of it remains of necessity impressionistic. As far as I can see it now, however, the tense suffixes are the same as those in the inscriptions and in Modern Tamil: - t (t)-, -nt-, and -i (ṉ)- for the past, -p (p)- and –v- (and -um) for the future, and – k (k) i (ṉ) ṟ- for the present. In contrast to the inscriptions the Bhakti poems abound in finite verbs in the present tense.

5. The finite verb tense system in Caṇkam poetry

Looking at the finite verb tense system in the Caṅkam poems, what immediately strikes the eye is the multiplicity of forms. That is to say, side by side with the three tense paradigms of past (iruntēn Puṟ. 399: 18 and ōcciṉaṉ Puṟ. 308: 9), present (cērkiṉṟa Pari. 22: 35) and future (pirivār Kuṟ. 174: 5), found in Modern Tamil, the inscriptions and Bhakti poetry, we have in particular for the non-past a number of alternative paradigms, namely:

1. The (present) tense marker –t (t)-, of which only forms for the first person plural (paṭartum Puṟ. 399: 13) and the second person singular (ceyti Puṟ. 12: 5) and plural (āṟṟutir Puṟ. 58: 20) are available.
2. -pa as in epicene plural eṉpa (Kuṟ. 12: 4).
3. The tense marker – (u) m, which is restricted to the third person singular. E.g. tiriyum (Puṟ. 183: 4) and kākkum (Puṟ. 191: 5). While in these examples its status as a tense marker is not clear, forms with – (u) m- have served as the basis for personal finite verbs (uṟaikkuntu Puṟ. 384: 7, eṉmaṉār Kuṟ. 234: 3) and participial nouns (niṉaiyumōr Naṟṟ. 104: 12).
4. Non-past forms ending in –v/p (p) al found only for the first person singular. E.g. uraippal (Akaṉ. 28: 2) and ceppuval (Akaṉ. 217: 15).
5. Forms ending in –k (k) u and –k (k) am for the first person singular (e.g. amaiku Kuṟ. 132: 3) and the first person plural (celkam "we/I (will) go away", Kuṟ. 114: 3) respectively.
6. A complete paradigm for all persons and genders consisting of the root plus – (k) kuv- followed by the personal endings, e.g. kēṭku + v + aṉ (Puṟ. 225: 9).
7. Some rare forms ending in – (k) kiṟp. E.g. paṭarkiṟpīr Kali. 39: 38, tarukiṟpāy Kali. 144: 49, and irukkiṟpōr Akaṉ. 387: 20.

These various constructions have all been placed into the broad category of the non-past tense, which, besides the habitual, includes the present and future tenses. As far as I know, no study has been carried out so far to find out if there are differences between the various forms.14 A study of this type is sorely needed. It is, for instance, striking that a great number of the instances of the formations like kāṇku/pāṭukam are found in so-called infinitival phrases like kāṇku vanticiṉ (Puṟ. 17: 33) and pāṭukam vammiṉō (Puṟ. 32: 6). As far as I know, for instance, the non-past suffix –t (t)- is not found in such contexts. In the absence of such a study the forms under consideration will of necessity have to be taken as having more or less identical functions.

31Most of the seven forms listed here are typical of Caṅkam poetry. An exception is the ending –um, which has been common in Tamil throughout its history. It should be noted, however, that the formations based on this ending (e.g. uṟaikkun-tu) are again peculiar to the poetic language. To the latter type of formation I will return below. First I will deal specifically with, first, -k (k) i (ṉ) ṟ-, secondly, with ceyti (1) and kēṭkuvaṉ (6), and, thirdly, with ceppuval (4), amaiku (5), and tarukiṟpāy (7). From a closer consideration of the origins of these formations a more or less consistent picture emerges of poets fashioning a language suitable to the scene depicted in the poems. As I will try to show, some of the processes involved in the creation of the poetic language have parallels in the North-Indian Kāvya, to which tradition Caṅkam poetry belongs as well.

32I must immediately add, however, that one of the forms which escapes definition in this context is ākupa (2). Evidently, we are dealing with an exceptional construction here, which becomes clear if we compare ākupa "they (living beings) are or become" (Kuṟ. 17: 4) with ākuva "they (things) are or become" (Naṟṟ. 317: 7). Thus, while the ending –a is that of the neuter plural, ākupa is used for the human plural, and the suffix is always –pa, also after so-called weak stems. The latter peculiarity is also seen in participle nouns such as koḷpavar (Naṟṟ. 292: 3) as against koḷvōr (Aiṅk. 187: 4), in which case koḷpavar is generally explained as due to dissimilation from *koḷvavar. However, the same peculiarity is also seen in the case of verbal participles in –pu, which suffix is always -pu, also in the case of weak verbs. See, for instance, koḷpu (Akaṉ. 25: 8) beside koḷvēṉ (Kali 65: 18).15 With all this, the origin of the formation ākupa, which seems to be restricted to Caṅkam poetry, remains unfortunately a mystery.

6. The rareness of the present tense in –k (k) i (ṉ) ṟ- in Caṇkam poetry

In the attempt to explain the rareness of the present tense suffix –k (k) i (ṉ) ṟ- in Caṅkam poetry I think we should start from the nature of the language of that poetry compared to that of the languages found in inscriptions and Bhakti poetry, in which the morph is found in abundance. As said, Caṅkam is a poetry about the past as well as about the village. Thus, the speakers in the historical Puṟam poems belong to the past, or to be more particular, to a period lying well before the wholesale introduction of North-Indian (Sanskrit) culture in Tamilnadu which took place under the Pallavas. The poems are meant to evoke a pure, undiluted Tamil society. A same type of society can be seen in the Akam love poems, which present unlettered people living in small villages far removed from the cosmopolitan culture of the towns. In this connection it should be noted that the language of the village poems of Akam does not differ from that of the historical poems in Puṟam. Apparently, the two varieties of language overlapped. And, indeed, a clear example of this is the absence of loanwords from Sanskrit, which as far as the ninth-century situation in Tamilnadu is concerned may be characterized both as an archaic feature and as a feature typical of a village dialect.

33By contrast, in the inscriptions we are dealing with a learned, administrative language. As to Bhakti poetry, whatever else it may be, it is also the outcome of the growing influence of the North-Indian Sanskrit culture in the South. Given this situation, it might be argued that in accordance with the poetic setting of Caṅkam poetry its poets had decided to avoid the use of – k (k) i (ṉ) ṟ- as an element otherwise typical of a highly Sanskritized (Bhakti), formal written language (inscriptions) in the same way as these poets avoided loanwords from Sanskrit.

34To this picture some more detail may be added. In this connection I would like to consider the present-day distribution of –k(k)i(ṉ)ṟ-. This present tense suffix, which is common to Tamil and Malayalam, is hemmed in between Telugu and Kannaḍa in the North and the Tinnevelly dialect in the South, all being languages and dialects which have a suffix –t (t)- instead. This distribution suggests that we are dealing with an innovation which had its origin in a specific centre somewhere in the Tamil or Malayalam speaking area and from there spread to the surrounding areas through a process of adoption. It is interesting to note then that this "new" suffix –k (k) i (ṉ) ṟ-, which made its appearance in the seventh century in the inscriptions of the Pallavas from the northern parts of Tamilnadu, was absent (or rare) in Caṅkam poetry but common in Bhakti poetry. This agrees with the fact that, while Caṅkam poetry implies a rejection of anything Pallava, in particular the Sanskrit culture introduced into Tamilnadu by that dynasty, Bhakti poetry with its North Indian gods and mythology marks the success of the cultural politics of these same Pallavas. As can be seen on the maps of the sacred places in Tamilnadu mentioned in the Bhakti poems provided by Hardy (1983: 256 ff.), the centre of gravity of Bhakti poetry was lying in the northern and western parts of Tamilnadu. It comprised the realm of the Pallavas and their successsors, the Cōḻas, and the realm of the Cēras. By contrast the realm of the Pāṇṭiyas plays only a relatively minor role in Bhakti poetry. The situation may be summed up as follows. The suffix –k (k) i (ṉ) ṟ- seems to have had its origin in the north-eastern part of Tamilnadu and had been adopted by the Pallavas in the formal, written language of their inscriptions.16 As shown by Bhakti poetry, its spread in Tamilnadu and into Kerala seems to coincide with the diffusion of the North Indian Sanskrit culture of the Pallavas. The absence, or rather, the non-use of – k (k) i (ṉ) ṟ- in Caṅkam poetry would basically have been the result of the attempt on the part of the Pāṇṭiya poets to revive a pure Tamil culture, which involved avoiding anything associated with the Pallavas, including their language.

35If the absence of –k (k) i (ṉ) ṟ was indeed a matter of avoiding an existing suffix, the occasional instances in Caṅkam poetry of the present tense participle with –k (k) iṉṟ- should be interpreted as lapses on the part of the poets.

7. ceyti and kēṫkuvaṉ

While formations like celval, kāṇku and irukkiṟpōr, which will be discussed in the next paragraph, are not only restricted to Caṅkam poetry but are also without any basis in the other varieties of Tamil or, for that matter, in the other Dravidian languages, the situation of the forms kēṭkuvaṉ and ceyti is different. To begin with ceyti, I have already noted above that instances of this suffix are not restricted to Caṅkam poetry but are also found in "later" Bhakti poetry. Furthermore, it appears that the present tense marker –t (t)-, apart from being common to the language to the North of the Tamil speaking area, is found in the present-day dialects in the southern parts of Tamilnadu as well (Andronov 1969: §§ 81, 96, 104, 109 and 126; see also Kamatchinathan 1969: 66-67). Given this situation the instances of this tense suffix in Caṅkam poetry do not necessarily point to an early date of Caṅkam poetry. For all we know, the suffix has always been there in the Tamil speaking area but has never made it into the standard written language.

36The same explanation may be applied to that other form, namely kēṭkuvaṉ, reflexes of which are found in the language of the tribal Kotas in the Nilgiri mountains (Tamil koṭu-kkuv-ēṉ, Kota koṛ-kv-ē "I will give "; see Subrahmanyam 1971: 308). In both cases, in ceyti and kēṭkuvaṉ, we seem to be dealing with formations from areas which are situated at the fringes of the earlier Pallava-Cōḻa realm, which at the time formed the homeland of the formal inscriptional Tamil. As said, we are dealing with formations which have not found their way into this formal Tamil. They are typical of Caṅkam poetry, which is to represent, among other things, the language of villagers living outside the pale of the learned textual tradition. We may thus be dealing with features borrowed from the spoken varieties of Tamil to set the poetic language off from the formal language and give it a "primitive" appearance. As I will try to show below this very process has parallels in the North-Indian Kāvya tradition.

37All this does not explain why the non-past suffix –t (t)- is used only for the first and second persons singular and plural and why only for –k (k) uv- we have a complete paradigm. However, in connection with the question as to why for – k (k) uv- we have a complete paradigm, I would like to point to the morph – k (k) i (ṉ) ṟ-. The evidence of the inscriptions and Bhakti poetry suggests that the – k (k) i (ṉ) ṟ- was the present tense suffix of the contemporary written – literary - language of the time. Whatever the origin of –k (k) i (ṉ) ṟ-, both –k (k) i (ṉ) ṟ- and – k (k) uv- have in common the element k (k) (u). It might then be argued that by giving preference to the k (k) u present above the one with t (t) the language of Caṅkam poetry was kept from becoming incomprehensible next to the other literary language, which had a non-past suffix –k (k) i (ṉ) ṟ-.

8. Old Tamil celval, kāñku and irukkiṟpōr

The range of forms for the non-past tense found in Caṅkam poetry includes the verbal nouns celval and kāṇku, which are used as finite verbs for the first person singular. As argued above, contrary to what has been maintained so far there appear to be no direct predecessors of these two forms in Proto-Dravidian. In fact, the reconstruction for Proto-Dravidian of the use of verbal nouns as finite verbs has been based solely on forms such as celval and kāṇku, which are otherwise typical of Caṅkam poetry. All we can say about the forms in question is that they are curious. This alters the focus of the problem they pose, for if we are not dealing with old or archaic forms, with what are we dealing? Zvelebil´s discussion of the two forms is unfortunately not very helpful in trying to answer this question. For, while he compares celval and kāṇku with modern verbal nouns used as predicates, such as Modern Tamil eḻutiyatu and pōkiṟatu (Old Tamil kaṇṇiyatu, e.g. Akaṉ. 5: 6),17 strangely enough he leaves unmentioned verbal nouns like ceyvatu. In Modern Tamil the latter type of verbal nouns is commonly used as a predicate: kumār eṉṉa ceyvatu "What shall Kumar do?" and kumār colvatu tāṉē "Kumar should have told it (or: should tell it), shouldn´t he" (Lehmann 1993: 77-78). For the earlier period, see, for instance, the Triplicane inscription of Dantivarman (Mahalingam 1988: no. 105, ll. 17-18): īḻaviḷakkum vaṭṭillōhappaṉaiyum raṭṣippatu, "we will protect the īḻaviḷakku, the vaṭṭil and the metal pot". When all is said and done, the difference between Modern Tamil ceyvatu and colvatu, and inscriptional raṭṣippatu, on the one hand, and Old Tamil celval, on the other, consists only in the respective verbal noun suffixes -atu and – al. Or, to put it differently, celval is celvatu in a different guise. As to kāṇku, once a predicate celval is accepted, it is only one step further to the predicative use of the verbal noun kāṇku.

38As I see it, we should seriously reckon with the possibility that celval and kāṇku, the occurrence of which, it should be emphasized once more, is restricted to Caṅkam poetry only, are artificial forms. But why would the poets have taken recourse to creating artificial forms? This question may be considered in the light of the attempt to fabricate an archaic language, or rather, a language which sounded archaic or primitive.

39Below I will try to show that in creating artificial forms Caṅkam poetry would not stand on its own, at least within the Kāvya tradition. Another point, however, is that the explanation of celval and kāṇku as artificial forms fashioned after ceyvatu does not account for the restriction of the use of these two verbal nouns to the first person singular. In modern Tamil verbal nouns like ceyvatu are used for the third person (singular) as well (see above). In the absence of studies of the inscriptional language which include findings of this type, it is at present not possible to say anything with regard to the use of the verbal noun in inscriptions. All we could say is that the poets of Caṅkam poetry used forms like celval and kāṇku in particular for the first person.18 As to the use of the verbal nouns in –ku in Caṅkam poetry it is to be noted that they are relatively frequently found in interrogative sentences (yāṉ evaṉ ceykō Kuṟ. 25: 2; compare kumār eṉṉa ceyvatu quoted above). In addition, they are invariably followed by the particle ō, that is, also when the verbs are found in sentences which are not necessarily interrogative (nōkō yāṉē Kuṟ. 212: 5). The particle ō is absent, however, when the verbal noun is used as a kind of infinitive as, e.g., in niṟ kāṇku vanticiṉē (Puṟ. 125: 4). Furthermore, side by side with a form for the first person singular we find one for the first person plural, which seems to consist of the suffix -ku to which the personal ending -am for the first person plural has been added: e.g. viṭukam (Kuṟ. 106: 6).19 This form has to be distinguished from, e.g., eṉkuvem (Kuṟ. 191: 7), which latter is part of a complete paradigm see above).

40On the other hand, the very occurrence of artificial forms in Caṅkam poetry opens new possibilities for the explanation of several other isolated forms. A case in point might be the instances of participle nouns of the type irukkiṟpōr. It is very tempting to interpret these extremely rare forms as learned creations based on the regular non-past paradigm uṇkuvam (Puṟ. 136: 27), after the latter form had come to be analysed as containing a combination of two tense suffixes, namely "present" –ku- and "future" –v-. The element –kiṟ- would then be a kind of abbreviated form of the present tense suffix -k(k)i(ṉ)ṟ-20and thus, incidentally, testify to the poets´ familiarity with the latter suffix.

41In the same way it might be asked if formations like tarum-ār, eṉm-aṉar and ākun-tu (Puṟ 380: 6), pūkkun-tu (Puṟ. 396: 2 with a variant reading pūṅkun-atu, for which, see Subramoniam 1962: 474) are not invented constructions as well. In any case, these forms are completely isolated. In none of the other languages in which the suffix – (u) m occurs it is followed by personal endings as is the case here. The only exceptions are Old Telugu and Parji (Subrahmanyam 1971: 317 ff.). Other evidence of the manipulation of the language may be found in the periphrastic constructions referred to above, e.g. celvēm allēm, kāṇku vanticiṉē and tarumār… iṟantōr or in the use of the deictic vowel i of ippoḻutu in ivvūr.

9. Caṇkam Tamil as an artificial language

Above, I have argued that the absence of loanwords from Sanskrit in Caṅkam poetry was a matter of avoiding such words as they would not fit into the specific poetic scenes. The same explanation has been suggested for the absence, or rather the extreme rareness of the present tense suffix –k (k) i (ṉ) ṟ- in this same poetry, as this suffix belonged to the formal, written language developed by the Pallavas. Some other peculiarities of the Old Tamil language, i.e. uraippal, ceyku and tarukiṟpāy, have been explained as artificial forms, that is, as the result of what may have been an attempt to create a language which sounded archaic and primitive. Finally, I have suggested that the non-past forms ceyti and kēṭkuvaṉ, as formations belonging to the spoken languages of different regions, have been adopted to give the poetic language a primitive appearance.

42At this point I would like to note that the linguistic situation met with in Caṅkam poetry is not unique. In connection with the mixture of forms drawn from different dialects, I would like to refer to the language of the Nāsaketa rī Kathā, a work in so-called Middle Mārwāṛī. According to Smith the language of this text which was in use for about 250 years for all kinds of literature, "does not answer to any single geographically definable form of speech, but is rather a compilation of features drawn from several distinct dialect-areas" (Smith 1975: 436). Its basis is a western form of speech, mixed with forms from Dhūḍhāṛī and from even further east. A similar mixture is found in the Vīsaḷadevarāsa. For instance, in this text we find the first person singular present tense –auṁ, which corresponds to the Modern Mārwāṛī forms found in the regions of Jodhpur and Jaisaḷmer, side by side with the ending -ū, the latter corresponding to the present-day forms from further north and east (Smith 1975: 436, 1976: 41). In the same dialect one may also come across an occasional instance of a Sanskrit ending, as in the third person plural jhaḷakaṁti (Smith 1976: 41).

43It is interesting to note that this plural ending –aṁti (from Sanskrit –anti) has also occasionally come to be used for the third person singular (e.g. karaṁti; see Smith 1976: 41). Most likely we have to do here with an extension of its use analogical with the "regular" ending –ai, which, at least from the point of view of Sanskrit, is singular, but which in the Vīsaḷadevarāsa is used for both the singular and the plural. As suggested above, something exactly like this might have been at work in the construction of the predicatively used verbal nouns uraippal and kāṇku in Old Tamil poetry. As suggested, in uraippal we might be dealing with a form artificially created after uraippatu by replacing -atu by a functional equivalent suffix -al. kāṇku might have been created after forms like uraippal, the verbal noun ending -ku substituting - (pp) al. -kiṟ-p-āy in tarukiṟpāy would have been created according to the same principle after -ku-v-aṉ in kēṭkuvaṉ.

44In fact, the principle accounting for the use of plural karaṁti for the singular because the singular karai was used for both the singular and the plural does not seem to stand on its own. In this connection I would like to refer to the use in Apabhraṁśa of the particle kiri (Skt kila) as a particle of comparison (Tieken: forthcoming). See, for instance, the following instance from the vidyāvilāsapavāḍau (lines 67-70), in which kiri is found side by side with such regular particles of comparison as jāṇe, jisyā and jima:

adhara suraṁga jisyā paravālī sarala sukomala bāha
pīṇa payohara atihiṁ maṇohara jāṇe amiyapavāha
ūrayugala kiri kadalīthaṁbhā caraṇakamala sukumāla
mayagala jima mālhaṁtī cālai bolai vayaṇa rasāla,

Her lips are red (lit. of a good colour) like (jisyā) coral, her arms straight and tender, her thick round breasts are extremely lovely like (jāṇe) a stream of nectar, her pair of thighs are [like] (kiri) the stems of the plaintain plant, her lotos-feet are tender, she swings her hips like (jima) an elephant, she speaks words full of love.

The use of kiri in this and some other passages can be explained as an extension of its original function, which is to indicate that the speaker is only passing on a message or is pretending to do so. In the Apabhraṁśa passage quoted above the poet is passing on a stereotypical poetic fancy. An "intermediate" instance of kiri is found in Virāṭaparva, lines 337-8, in which it is found side by side with the particle of comparison jāne:

etalai śuśarmā dali ḍhola vājaiṁ
jāṇe asāḍhū kiri meha gājai,

In the meantime in Śuśarmā´s army the drums were beaten: it sounded like the thundering clouds in the month of Āṣādha.

Finally, I would like to come back to the avoidance in Old Tamil poetry of loanwords of Sanskrit and the present tense suffix –k (k) i (ṉ) ṟ- as features which do not fit into the poetic scenes. This phenomenon may be compared to the situation in Kāvya, that is, the very literary tradition to which, as I have tried to show elsewhere (Tieken 2001), Caṅkam poetry belongs. In Kāvya we see a constant switching from one dialect to another, each of which has its own domain and its own register. In this connection I would like to refer in the first place to the Sanskrit drama. A telling example is that, reported in the Nāṭyaśāstra (XVII 33), of the king, who normally speaks Sanskrit but when intoxicated by his own power (aiśvaryeṇa pramatta) switches to Prākrit. A case like this is found in the beginning of the third act of Harṣa´s Nāgānanda in which the drunken viṭa, who normally speaks Sanskrit is made to speak Prākrit.

45In the same way persons who normally speak Prākrit might under certain circumstances switch to Sanskrit. Thus, in drama the leading female roles and their companions do not speak Sanskrit but Śaurasenī Prākrit. However, the Nāṭyaśāstra points out that under certain circumstances the queen (or princess, rājñī), the courtesan (gaṇikā), and the female artisan (śilpakārī) may switch from Prākrit (i.e. Śaurasenī) to Sanskrit.21 If we turn to the list of so-called circumstances (NŚ XVII 37-41) we notice that for the queen they contain mainly topics of conversation: matters related to war and peace (sandhivigrahasaṁbandhaṁ), the course of the stars and planets (grahanakṣatracaritaṁ), and omina (consisting of the cries of birds, khagānāṁ rutam). The lists of the gaṇikā and śilpakārī are a mixture of topics and pragmatics. The gaṇikā (v. 40) may switch to Sanskrit to amuse sarvaloka (her clients, or perhaps the audience, krīḍārthaṁ sarvalokasya); the use of Sanskrit (or: during a performance, prayoge) provides pleasure (sukhāśrayam); moreover, Sanskrit is part of her practice in the kalās (kalābhyāsāśrayaṁ). The śilpakārī (v. 41) switches to Sanskrit to show off her knowledge of her art and her dedication to it (kalopacārajñānārthaṁ), and also when she wishes to please the king (krīḍārtham pārthivasya).

46Something similar can be seen in that other branch of Kāvya literature, namely inscriptions. The inscriptions of the early Pallavas in Tamilnadu are in Prākrit. According to Lévi (1902: 112-113) the scribes of these inscriptions were well acquainted with Sanskrit but felt bound to avoid the use of this sacred language for such pedestrian topics as a grant of land. In these same inscriptions Sanskrit occasionally does make an appearance, but, typically, in some sacred formulae at the end of the inscriptions (e.g. svasti gobrāhmaṇavācakaśrotṛbhya iti).

47It appears that in Kāvya the use of either Sanskrit or Śaurasenī is determined by function. Sanskrit, for instance, is used for discussing topics of learning. It is the language of a specific body of knowledge, which included the knowledge of the affairs of the state and omina. Śaurasenī is used for ordinary conversation. This distinction is matched by the nature of the respective languages. Of the two, Sanskrit is obviously the more complicated language, and if Śaurasenī indeed represents the spoken language, learning Sanskrit must have required a special effort. Sanskrit is not only the language of learning but also a learned language. And indeed the mere use of Sanskrit suffices to mark the speaker as a scholar (the viṭa) or else as a person (the king) who is prepared to submit his actions to the advice of learned persons (his brahmin advisers). However, when this same learned person is to be presented as drunk or otherwise intoxicated he is made to lapse into Prākrit. Sanskrit, with its complex grammar and its association with a learned textual tradition does not fit with this type of behaviour.

48The avoidance of Sanskrit in the portrayal of a drunken character may be compared to the avoidance of Sanskrit loanwords and of the present tense suffix –k (k) i (ṉ) ṟ- in Caṅkam poetry, assuming that these latter features did indeed belong to a register incompatible with the scenes in Caṅkam poetry.

49In this connection I would like to return to the rareness of the plural suffix – k (k) aḷ in Caṅkam poetry. As already argued above, most probably the suffix was known to the authors of the poems or was even available in their own language. If so, its absence from the poems must be due to an attempt to avoid the suffix. The question which then arises is why they avoided the suffix. In this connection I would like to refer to the situation in present-day Tamil. While in Modern Written Tamil the plural suffix seems to be obligatory, in the spoken language the plural of non-personal nouns may under certain circumstances remain unexpressed, for instance if the plural is already indicated by a number. On closer consideration the situation met with in Caṅkam poetry thus appears to be one step further removed from the one in written Tamil than in the spoken variety. Caṅkam Tamil, in which the plural suffix –k (k) aḷ has been removed completely and systematically, might be characterized as an imitation of the spoken language, exaggerating the peculiarities of the latter. Of course, we are dealing with an hypothesis only here, which moreover tries to explain a ninth-century phenomenon with reference to the situation in the present time. However, as in the case of the avoidance of Sanskrit loanwords and the present tense formation with –k (k) i (ṉ) ṟ-, here, too, a parallel may be found in the North-Indian Kāvya tradition. In this connection I would like to refer to the elision of all intervocalic plosives (except for the retroflex ones) in Māhārāṣṭrī Prākrit in Hāla´s Sattasaī. The latter text offers short monologue verses, which are put into the mouth of rustic types living in small villages. Elsewhere (Tieken 2001) I have argued that we do not have to do with a poetry of these villagers but a poetry about them. The text itself has its origin in the very same cosmopolitan milieu which produced the Kāmasūtra. This implies that the language is not the dialect spoken by the villagers but an imitation in Sanskrit of that dialect. In accordance with the rustic, unlettered people featuring in the poems, the dialect, which is marked by the complete elision of the intervocalic plosives (except for the retroflexes), is, at least as far as its phonology is concerned, as far removed from Sanskrit as is possible. At the same time we are almost certainly dealing with an exaggeration here. In this connection I need only refer to the amount of ambiguity arising from the indiscriminate elision of the intervocalic plosives, which is almost unthinkable in a "normal" language. Take, for instance, the word maa, which could stand for Sanskrit mata "opinion", mada "intoxication", mṛta "dead" and mṛga "dear, wild animal".

50In this connection I would like to mention another peculiarity met with in the Sattasaī. To add local colour to the speeches of the villagers so-called deśī, or regional, words have been inserted (e.g. tuppa "red colour" and sippī "conch"). We are to believe that these words belong specifically to Maharashtra, which is where the scenes are set. The effect intended by the insertion of Deśī words is on final analysis the same as the one aimed at by avoiding Sanskrit words. Unfortunately, the question to what extent something like this has been at work in Caṅkam poetry as well remains largely unclear. Leaving aside the rareness of Sanskrit words, we are unable to evaluate the nature of the vocabulary of Caṅkam poetry and the extent to which it comprises rustic or outdated words.

51What I have tried to show is that the peculiarities of Caṅkam poetry which have been observed above actually belong to the "tricks" of Kāvya. These peculiarities seem to be part of the attempt to create a language appropriate for poor wandering bards from the past and typical of unlettered people living in small villages in the countryside. Parallel developments are indeed otherwise found specifically in the literary Prākrits, including Apabhraṁśa, which within the Kāvya tradition likewise function as non-written languages. It should be noted that, at least at first sight, this attempt at creating a primitive language is almost completely annuled by the style of the Caṅkam poems. For, the style of this poetry is also marked by extremely long sentences showing a sheer endless chain of sentences embedded in other sentences. Other characteristics are the use of long compounds. With these features we seem to be far removed from the language of ordinary conversation. It should, however, be added that a similar development is seen in Prākrit texts. Thus, as already noted by Bühler, long sentences are typical of the Prākrit inscriptions of the Sātavāhanas (Bühler 1890: 59), and long compounds abound in Prākrit Kāvya texts like the Setubandha and Gauḍavaho. Even the short gāthās of the Sattasaī include several verses which combine long compounds with a highly convoluted imagery.22 As I see it these latter features only show that in both the Prākrit texts and the Tamil poems we are ultimately dealing with a highly learned poetry. Apart from the supposedly simple language the speakers in the Sattasaī use, the poems are veritable riddles, the solution of which demands an agile mind on the part of the reader and, among other things, a thorough knowledge of the Kāmasūtra (see Tieken 2001). The same applies to Caṅkam poetry, which requires careful puzzling. On closer consideration the introduction of Prākrit or of a pure Tamil without Sanskrit loanwords was just one of the literary tricks of the Kāvya, and a highly sophisticated one at that.

52In this connection it should be noted that Kāvya is literature as an art form. It is characterized by the exploitation of everything which is theoretically possible, linguistically as well as stylistically. A case in point of exploring the linguistic possibilities may be found in the second canto of Aśvaghoṣa´s Saundarananda, which up to verse 45 is a showcase of all (theoretically) possible aorist forms, from verse 40 onwards of the passive of the aorist. Kāvya also abounds in learned paraphrases. A simple example is the use of pratispardhin "vying with" in comparisons ("the girl´s face is vying with, that is, resembles, the moon") (see Bühler 1890: 52). Furthermore, the poets favoured complicated circumscriptions. A case in point, noted by Renou (1959: 3-4), is Kirātārjunīya 11.62:

sa pumān arthavajjanmā yasya nāmni puraḥsthite
nānyām aṅgulim abhyeti saṁkhyāyām udyatāṅguliḥ

[C]elui-là est un homme à la naissance conforme (à son caractère), celui au nom duquel, (nom) mis en tête (d´une liste), lorsqu´on fait un dénombrement, le doigt levé (du dénombreur) ne passe pas à un autre doigt (de la main ; mais, bien au contraire, demeure fixé sur ce nom)"

10. Concluding remarks

Often, in support of the early dating of Caṅkam poetry around the beginning of our era linguistic evidence has been called in. The language of the poems would show many archaic features. At the beginning of this article I have tried to show that the archaic nature of several of the features concerned is not as well established as one would wish. Apart from that, if a feature is archaic it need not automatically be old or early as well. In this respect the argument is often completely circular, recourse being taken to the supposed early date of Caṅkam poetry itself. Whatever is exactly the case, all this does not alter the fact that Caṅkam poetry abounds in linguistic peculiarities, the occurrence of which is in many cases typically restricted to this poetry only. In the present study I have tried to explore possible alternative explanations for some of the peculiarities of the verbal tense system. The explanations offered are grounded in the study of constructed and artificial literary languages. The approach from this angle is justified by the rareness of loanwords from Sanskrit in the Caṅkam poems. That is to say, if Caṅkam poetry has indeed originated only in the eighth or ninth century, the rareness of loanwords from Sanskrit suggests that we are dealing with a language which was consciously kept free from such words, in short, with an artificially fabricated language.

53The present study is no more than a first exploration of a question which has arisen only very recently. Also, it has been based only on a small number of features. Whatever thoughts one may have of its results, it should have become clear that the assumption that the language of the Caṅkam poems represents an early stage of Tamil and as such cannot possibly belong to a period as late as the eighth or ninth century still awaits proof. This proof should avoid the circular argument and not itself depend on the early date of the poetry. In the meantime I venture one final remark, namely concerning the terminology used in literary as well as linguistic studies. Caṅkam poetry is often called Old Tamil poetry. One of the conclusions of the present study is that the language of Caṅkam poetry is not Old Tamil. As far as I can see, the latter term should be reserved for the language of the Tamil Brāhmī inscriptions from the beginning of our era, which constitute our earliest source for Tamil. There simply is no Old Tamil literature for there is no literature from the period of these inscriptions. The earliest literature from Tamilnadu is in Sanskrit, namely the Mattavilāsaprahasana by the Pallava king Mahendravarman I (610-630) (see Tieken 1993). In this respect the situation in Tamilnadu does not differ from that in medieval Europe, in which the earliest attempts at literature were mostly in Latin, the vernacular languages taking over the role of literary language only later.

54To call the language of Caṅkam poetry eighth or ninth-century Tamil would not be correct either. While the inscriptions from that period cannot be expected to give a faithful picture of the contemporary (spoken) language, it is almost certain that the language of the poems does not present that language either. The contemporary language of the learned and literary circles in the towns must have contained many more loanwords from Sanskrit than the few found in the poems. Apart from that, the occurrence side by side of different formations for the non-past tense would show that we are not dealing with one monolithic, geographically definable language but with a composite language pieced together with elements from different dialects from different areas. As I have tried to show we are indeed dealing with a language fabricated specifically for Caṅkam poetry. Therefore for labelling the language of that poetry I would like to revive a traditional term from Āṇṭāḷ´s Tiruppāvai (XXX 5), namely "Caṅkam Tamil" (caṅkattamiḻ).