Tamil has a number of verbs, sometimes referred to as `aspectual verbs' (or aspect markers, aspectual auxiliaries, verbal extensions, post-verbs, intensive verbs …) that are added to a main or lexical verb to provide semantic distin ctions such as duration, completion, habituality, regularity, continuity, simultaneity, definiteness, expectation of result, remainder of result, current relevance, benefaction, antipathy, and certain other notions.

Researchers have generally found these aspect markers difficult to describe in a categorical way, and in only a few studies (Annamalai 1985) has any attempt been made to treat aspect in Tamil (or for that matter, any Dravidian language) as a variabl e component of the grammar, or as a system that is in a state of dynamic evolution (Steever 1982). In this paper I will attempt to summarize what is known about the facts of aspect-marking in Tamil and to place it in a larger framework, recognizing th at a number of different analytic approaches to this difficult topic are necessary. Aspect is a variable category in Tamil, but this variability is a product of the process of grammaticalization (Hopper and Traugott 1993), rather than purely sociolinguistic.

The focus of this paper: The focus of this paper will be on those elements of the meanings of lexical verbs recruited to serve as aspectual verbs that have to do with spatial relations, in particular relations perceive d by the speaker to express his or her personal space, and/or relationship to his/her body. That is, verbs that express motion away from or towards the speaker, verbs that express static proximity to or distance from the speaker, verbs that express benefi t-faction to the speaker (e.g. as the result of obtaining some benefit, containing some action or benefit, or something pertaining to the speaker) seem to be those that are recruited as aspect markers. Furthermore, any action that is antithetical or malef active to/for the speaker (speaker perceives the motion, the action, the pertainment to be threatening, annoying, disgusting, invasive, etc.) may also become an aspectual verb, but with this clear element of antipathy or malefaction overtly present, unlik e the benefactive verbs. In other words, malefaction is probably more marked than benefaction, so benefaction is the default unless a clearly malefactive/antipathetic action is involved.

To be a candidate for successful aspectual grammaticalization, a lexical verb must contain at least one of the four elements of motion (or perhaps better, deixis), stasis, con/per/ob-tainment, and malefaction/antipathy, with the added complexity that t hese can combine in various ways to yield different surface aspectual verbs, as well as complexities of meaning for individual aspectual markers that are a challenge for my analysis. Just to illustrate briefly, the location/static verb iru 'be (located)' when combined with the containment verb koL are both recruited to form, in combination, a durative/continuative aspectual marker kiTTiru. The metaphoric 'holding' of koL (its lexical meaning is 'hold, contain') plus the metaphoric static 'be ing' of iru yield an aspectual marker expressing duration: 'holding the being, or continuing the holding.' Just as in English, the verb 'hold' has been extended metaphorically, e.g. in telephone usage, from its original meaning expressing 'holding' the telephone receiver (i.e. not hanging up) while waiting for a line or a connection. 'Will you hold?' thus means, not "Will you hold the receiver and not hang it up?" but "Will you wait for a connection?" It is then further extended to express the kind of waiting at airports when planes are in 'holding' pattern: waiting for a runway. The literal meaning of holding is no longer present (nobody's holding anything in their hand) but waiting, in particular waiting in a pattern that gives priority to those airplanes that have been there the longest, is the metaphoric meaning. It is no accident that the verb koL 'hold, contain' involves originally holding something in the hand, in proximity to the body (i.e. no further than arm's length). This defines the original perimeter of what is one's personal space, and things entering or leaving this space, or the action of leaving s.t., or remaining in proximity to it, and the verbs that express this, are the crucial ones for recruitment as aspectual markers.

Aspectual verbs in Tamil are on a kind of continuum from completely grammaticalized to only-beginning-to-be-grammaticalized. This is evident from the fact that some aspectual verbs (those I call completely grammaticalized) have complete freedom of occu rrence with other verbs---they can occur with any verb, transitive or intransitive, with all persons, and with all tenses. They can be used with modals, in the negative, in any context. The most completely grammaticalized aspectual verb is viDu, wh ich is based on the lexical verb viDu 'leave, let.' Furthermore, such c.a.v.'s also exhibit phonological peculiarities, which is also a marker (Hopper and Traugott) of grammaticalization. In this case, the initial [v] of the lexical verb is lost in ST, something that never happens at word boundaries in Tamil, but is quite common word-internally.

Less advanced on the scale of grammaticalization are aspectual verbs that can only be used in certain contexts, e.g. only with transitive verbs, or only with 3rd person subjects, or only in the past tense, or perhaps have restrictions on use with modal or negative verbs. The less grammaticalized, the fewer instances of phonological peculiarity we also witness, so that the AM vayyi 'future utility' (based on the lexical verb vayyi 'put (for safekeeping)' does not exhibit v-deletion (as doe s viDu) and can only occur with transitive verbs. Fully-grammaticalized AM's can even occur with their lexical analogs, e.g. naan ade viTTu-TTeen 'I completely left it; I finished it and got it over with', whereas less completely grammatical ized AM's do not occur so freely with their lexical analogs.

Verbs that are the least grammaticalized are those that are only used in very limited contexts, or are used only by certain speakers andor dialects, or are only used in a metaphoric sense some of the time. The verb koDu 'give' is an example of t his limitation; some speakers can use it with other verbs freely, but many can or do not. There is a phrase sollikoDu meaning 'teach' ('say and give; give by saying') that expresses this metaphoric usage well, and is used by many speakers. But beyo nd this collocation many do not use it, or would consider this a lexicalization, a 'phrasal verb', but not an aspectual usage, so I relegate it to 'early candidacy' as an aspect marker. If some argue that it is an aspectual marker, it is so only in some d ialects, or for some speakers.

Note of course that the verb koDu 'give' in its basic meaning involves the use of the hands, and involves motion of an object from the physical space of one person (the giver) into the physical space of the recipient, and perhaps into hisher han ds. Certainly ritually or ceremonially the hands are involved; drop-kicking a book into your neighbor's office, or parachuting it from the roof of WMS hall would not be considered 'giving'.

The hands are also involved in the verb taLLu 'push, shove' which is the source of the AM taLLu 'riddance; distributive'. The latter meaning indicates that motion is 'out of one's hands' and into the hands or possession of 'unspecified re cipients.'

Aspect and Commentary. Tamil aspect markers provide information and commentary about the manner in which an action occurred, especially how it began or ended, whether it was intentional or unintentional, whether it had an effect on the sp eaker or on someone else, whether it preceded another action or was synchronous with it, and so on. Some of these notions are what have been considered aspectual in other languages (having to do with the completion or non-completion, the continuity or dur ation, the manner of inception or completion) but some have little or no relation semantically to classical notions of aspect, by which I mean aspect as seen, e.g. in the Slavic languages. These `extended' uses of aspect markers sometimes therefore involv e value judgments by the speaker about the actions of others, i.e. they indicate what the speaker's attitudes or expectations about the verbal action in question are.

Most aspect markers are derived historically from some lexical verb that is still in use in Tamil but has its own lexical meaning. The `meaning' of aspect markers is primarily grammatical or syntactic and can usually only vestigially be related to the lexical meaning of the verb from which it is derived. It is here that the role of metaphor comes into play, since it is by metaphoric extension of the lexical meaning (especially the spatial meanings) that the grammatical meaning is arrived at.

Syntactically, aspect markers are added to the adverbial participle (AVP) of the lexical (`main') verb. Aspect markers then are marked for tense and person-number-gender (PNG), since the AVP preceding them cannot be so marked. Morphologically (but not phonologically) they then act identically to the lexical verb from which they are derived, i.e. take the tense markers etc. of the class of lexical verb they are identical to. In most analyses of aspect in Tamil, researchers have focussed on Literary Tami l, and have tried to show that aspect can be considered a

syntactic process, since the aspect markers appear to function independently of morphology. I believe this claim is based on the artifact of modern writing, where Tamil aspect markers are (or can be) written with spaces between them and the AVP of the lexical verb. In Spoken Tamil, as I have tried to show, aspect has become (or is becoming) grammaticalized, i.e., aspectual markers now function (in many cases) as part of the morphology of the language, as evidenced by the phonological processes that app ly to them that usually only apply word-internally (Schiffman 1993). In this paper I will attempt to show that Tamil aspect is a category that is variably grammaticalized. My focus will be on the role of metaphor in the evolution of this system, such that the lexical meanings of the source verb being aspectualized gradually yieldexpand to metaphoric extensions of those meanings, and as th e lexical meanings are leached out, grammaticalization of the verb as an aspectual marker takes place. These aspect markers are thus no longer lexical verbs, nor are they, when completely grammaticalized, independent verbs at all---they lose their syntactic independence and become morphological suffixes affixed to main or lexical verbs, as evidenced by their lack of syntactic boundedness, whi ch we see from the phonological rules that apply to them but not their lexical analogs.

What is crucial in the role of metaphor is the types of lexical meanings that the source verbs have or had; in almost all cases, we can identify the following semantic elements:

• Deixis, having to do with motion, spatial relation, or proximity to(ward) or away from the speaker. This varies, of course, and as we shall see, the 'meanings' of different directions with regard to speaker are valued differently (e.g. 'up is good, down is bad' etc.)
• Stasis, having to do with continuity, duration, lack of boundedness, habituality, etc. of action or state.
• Antipathy: the value of the action or state is negative; (speaker does not like some action or state; the negative or pejorative notions are what I refer to below as attitudinal .) (Remember how speaker-centered-ness involved in grammaticalization often involves expressing negative thoughts.)
• Containment, in particular abrupt closure, interruption, boundedness, or finiteness of an action, but also continuity or duration.

In many cases, more than one of these semantic elements is part of the lexical meaning of the source verb, and the metaphoric extension/abstraction of any or all of these meanings moves the verb along the continuum to grammaticalized aspect in differen t ways. Aspect as a grammatical category is thus derived from the semantic elements of deixis, stasis, antipathy, andor containment. (Parallels with the evolution of aspect in other languages may suggest themselves, since many languages that have aspectua l systems seem to have involved the grammaticalization of semantic elements similar to one or more of these meanings.)

What are the aspect markers of Tamil? The verbs that can be treated as aspectual are actually on a gradient scale of grammaticalization; those that are more completely grammaticalized (primarily aspectual and minimally attitudinal) are viDu `com pletive', koNDiru `durative', vayyi `future utility', aaku `finality, expected result', vaa `iterative', poo `change of state', koo `self-benefactive', iru<sup>1</sup> `perfect', iru² `res ult remains', and iru<sup>3</sup> `epestemic.'

The lexical analogs of these aspectual markers are, respectively, viDu `leave, let', koNDiru (no lexical analog, but made up of elements of the lexical verb koLLu and iru `be located'), vayyi `put away', aaku `become', vaa ` come', poo `go', koL `hold, contain', and iru `be located.'

Attitudinal markers The aspect markers that are less completely grammaticalized (i.e., are primarily attitudinal but nonetheless involve some aspectual notion) are taLLu `distributive', tole `riddance', pooDu `malicious inte nt', and some others that vary from dialect to dialect, such as koDu which has a `benefactive' meaning in some dialects. The lexical analogs (or `source verbs') of these aspectual markers are, respectively, taLLu `push, shove', tole ` (go to) ruin', pooDu `drop, plunk; put on (clothes).' (The lexical analog of koDu is, not surprisingly, `give').

The notion that attitudes or value judgements might be semantically related to aspect may seem at first problematical, but as Johnson and Lakoff show (1980), notions that are originally spatial or deictic, such as the prepositions `up' and `down' are u sed metaphorically in many languages for positive and negative meanings: things that are `up' are good, and things that are `down' are bad; but we also see that these same prepositions have evolved (probably also via metaphor) into aspectual notions in En glish, so that `up' as a verbal extender has the meaning `completive' as in `eat up, use up, tie up, burn up' while `down' used with the same or similar verbs has another meaning, perhaps not clearly aspectual: `tie down, shut down, pin down, burn down', etc. Similarly in Russian, the preposition u meaning `in proximity to; in the possession of' (u m'en'a est' `I have' (`near-me is') is also used as an aspect marker of completion or inchoativeness: znat' `to know' vs. uznat' `c ome to know, realize'; snut' `to sleep' vs. usnut' `to fall asleep.'

An attempt to schematize these four elements as they semantically characterize the lexical verbs in question is shown in Tables 1 and 2:

(v)iDu 'completive'. This aspectual verb contributes the semantic notion that an action was, is, or will be complete or definite. It is similar to aspectual verbs in other languages (Russian, Hindi, etc.) that impart the notion of 'perfective' (not perfect). Its lexical correlate is viDu 'leave, let.'

Examples:

2. naan vand-iDreen
I come-compl-pres-png

'I am definitely coming; I'll come for sure.'

1. avane anuppuccuDu
   him send-caus-compl-imp

1. ade saappiTTuTTeen
   it-acc eat-compl-past-png

2.1.2. vayyi 'future utility'. 'put away, put somewhere for safekeeping'.

The aspectual verb vayyi has a lexical analog vayyi 'take, put s.t. somewhere for safekeeping'. It is usually used with transitive main verbs only (since the MV vayyi is definitely transitive,) but may occur with some intransitive verbs, such as siri 'laugh' (see example below). Other aspectual verbs (e.g. (v)iDu ) may follow vayyi , but when present vayyi always follows immediately after the AVP of the main verb. The aspectual metaphor conveyed by va yyi is the notion that some action is performed because it will have future consequences, use or benefit; it is often translatable as 'in reserve' or 'up', e.g. 'stock up (on)', 'read up (on) something, 'study up on something', 'lay in (or up) a stock of (something)', and implies that an action is done with an eye to future consequences, or preemptively. In the examples below, the glossed portion within square brackets is not literally present in the Tamil sentence, but is given as one or more of the consequences that the use of vayyi implies.

1. taNNiire kuDiccu veppoom
   water-acc drink fututil-fut-1pl

1. ammaa piLLengaLukku doose suTTu-veccaa
   mother children-dat pancake heat-fututil-past-3sgfm

1. pooliiskiTTe edeyaavadu oLari- vekkaadee
   police-to something-or-other babble-fututil-neg-imp

1. naan naaye kaTTi-vekkalle
   I dog-acc bind- fututil-neg

1. sundaram tan makaLukku nalla eDattle kalyaaNam senju-veccaar
   Sundaram his daughter-dat nice place-loc marriage did-fututil-past-3sg-epic

1. talevar kuuTTatte taLLi-veccaar
   head-person assemblance-acc push-fututil-past-3sg-epic

1. DairekTar oru jook sonnaar; naan siriccu vecceen
   Director a joke said; I laugh fututil-past-1sg

1. kalyaaNa viiTTle tummi kimmi vecciDaadee
   marriage house-loc sneeze echoredup fututil-compl-neg-imp.

This aspectual verb has the lexical analog aaku 'become.' It is usually found only in the neuter past, i.e. aaccu. Suffixed to a main verb it expresses the notion that the action was expected, or occurred after a long wait, or as a regularly expected occurrence.

1. poosT vand-aaccu
   mail came-xpreslt

'The mail has come [as it usually does by this time of day]'.

1. inda kaNakkukaL-ellaam paatt-aaccu
   this bills- all seen-xpreslt

1. saappiTT-aaccaa?
   eat- xpreslt-q

The aspectual verb vaa has a lexical analog vaa which means 'come'. The notion conveyed by aspectual vaa is that an occurrence is or was of long-standing duration, but more as a series of connected events (or waves of occurrences) rather than as uninterrupted continuity. vaa may often express a kind of 'narrative' or 'historical' ('episodic', 'customary', or perhaps even 'mythological') past, describing an action that was common practice in a past time. Since it is only used in LT, the example(s) below are in LT rather than ST.

The aspectual verb poo resembles the lexical verb poo 'go' in its morphology. It is used to express the notion that a change of state has definitely taken place (or will definitely occur). As such it is aspectually comple tive but the main verbs to which it is attached always themselves have some semantic notion of change; the addition of poo shows that the change is complete. Usually the net result is also judged to be unfortunate or undesirable, so that it must be (as an aspect marker) marked "+antipathy".

The aspectual verbs that express, in addition to various aspectual notions, notions about the speaker's attitude toward actions or other speakers, are, as mentioned taLLu 'distributive', tole 'riddance', pooDu < /I>'malicious intent', and others that we will mention only in passing.

1. neettu varakkuuDaadu-NNu sonneenee. aanaa, neettu paattu vandu-pooTTaanga.
   yesterday come-neg-necess-qt said-emph. but, yesterday deliberately came-malice -tense-png

   'I told them not to come yesterday, but they deliberately came anyway [the jerks!]'

   23. tiruDanga en naaye koNNupooTTaanga
   thieves my dog-acc kill-malice-tense-png

   '[Those dirty rotten] thieves [deliberately and in cold blood] went and killed my dog.'

   24. koRande taaLe kiRiccu pooTTadu
   child paper-acc tear malice-tense-png

   'The child [carelessly] tore the paper.'

   25. avan ajaakkradeyaa kadave tirandu pooTTuTTu pooyirukkaan
   he carelessly door-acc opened malice go-perf-png

   'He has gone out, carelessly leaving the door ajar.'

Durative or continuous action similar to the 'progressive' construction verb+ing in English, is expressed in Tamil by combining koo in its AVP form kiTTu with the 'stative' aspectual verb iru, i.e. kiTTiru, and affixing this to the AVP of a main verb: vandu+kiTT-irundeen 'I was com-ing.' Thus the 'containment' semanteme combined with the 'statis' semanteme results in 'continuous, durative action,' similar to English 'holding' as a metaphor for continu ing to wait (on the telephone, above congested airports, etc.)

43. ellaarum peesi- kiTTiru-ndaanga
all speak-durative-were

44. raaman saappiTTu-kiTTiru-kkaaru
Raman eat- durative-perf-png

45. kamalaa vanda poodu, naan paDiccu-kiTTiru-ndeen
Kamala came when, I read-durative-tense-png

46. koRande eeRu maNikkuLLee tuungi-kiTTiru-kkum
child seven o’clock-within sleep-durative-will-be

This summary of some of what is known about Tamil aspect is not meant to be exhaustive; it may never be possible to encapsulate all that can be said about this issue. But we do seem to be able to make the following generalizations:

Another variable feature of Tamil aspectual verbs is that there are pragmatic considerations that are involved in the choice of whether to mark aspect or not. Since aspect is not an obligatory category, it may or not be present. Howeve r, there is a tendency not to mark aspectual distinctions in certain constructions, even if they might be technically grammatical. The reasons for this are pragmatic ones, having to do with politeness, shared perceptions, the nature of truth propositions, etc. There is also a tendency to use aspectual marking to add speaker commentary to the sentence, even with the 'purely' aspectual markers, but especially with the attitudinal ones.

I suggest that it is in the growing body of literature on grammaticalization (e.g. Harper and Traugott 1993) that we can begin to put our finger on how the verbal category of aspect can be dealt with. Most analyses of VAM have dealt wi th aspectual verbs as if they were simply a special kind of verb with special syntactic properties, rather than a morphological marking of an obligatory or optional category.

This approach has characterized most previous analyses (including my own). In a recent study (Schiffman 1993)) of certain phonological processes that seem to apply only word-internally in Tamil, however, I have had to conclude that in spoken Tamil, ver bal aspect marking, once explainable as a syntactic process, has become morphological; since VAM is clearly not inflectional, it seems rather to be derivational. The fact that the evidence for this is phonological is problematical, but it is the very abse nce of the phonological process in LT that has meant that verbal aspect marking can be handled as a syntactic phenomenon with impunity in LT. And in fact, most syntactic studies of VAM (e.g. Steever 1983, Annamalai 1978, Dale 1975) have relied on data fro m LT, rather than ST. Had these studies used the spoken language for data, they would, I believe, have been forced to draw other conclusions. A lot can happen in seven centuries in the history of a language, and as far as VAM is concerned, the difference seems to be that the process of grammaticalization of this category has progressed since the time LT was last codified, i.e. in the 13th century.

What does the evidence from studies of grammaticalization in general offer this analysis in particular? Let us review some of the characteristics of grammaticalization, the process by which lexical items become grammatical morphemes in various languages:

1. Happens mainly with verbal items; becoming modal, auxiliary, or aspectual in some way.

2. May be more widespread in the spoken language than in the written. Newly grammaticalized modals in English like `gonna’, 'sposeta' and 'hafta' (derived, of course from supposed to and have to, respectively, are found with those pronunciations in the spoken language, but tend to be replaced by other items, e.g. `must, should, going to’ in literary dialect.

3. The process is gradual, lasting perhaps centuries, and retaining traces of the lexical past of the auxiliary even when fully grammaticalized.

4. Phonologically, newly-grammaticalized material operates differently than the old lexical item: [haefta] instead of have to, [sposta] instead of supposed to, going to is replaced by [gonna] except when 'motion toward' is meant, not futurity:

• He's gonna go to the store.

  but not

• * He's gonna the store [= He's going to the store.]

Grammars of languages in which grammaticalization is ongoing and not yet codified tend to ignore the process and state only the older norm; the existence of newer developments may be noted humorously (as in cartoons, jokes, etc.), but are not taken ser iously. As we can see, the Tamil aspectual system exhibits many of the signs of grammaticalization:

[1] Annamalai, E. 1985 The Dynamics of Verbal Extension in Tamil. Trivandrum: Dravidian Linguistics Association.

[2] Arden, A. H. 1942. a Progressive Grammar of Tamil. Madras: Christian Literature Society.

[3] Dale, Ian. 1975. Tamil Auxiliary Verbs. Unpublished Ph.D. Dissertation, School of Oriental and African Studies, University of London.

[4] Emeneau, M. B. 1956. "India as a Linguistic Area." Language.

[5] Hopper, Paul and Elizabeth Traugott. 1993. Grammaticalization. Cambridge: CUP.

[6] Lakoff, George and Mark Johnson. 1980. Metaphors We Live By. Chicago: University of Chicago Press.

[7] Schiffman, Harold F. 1969. A Transformational Grammar of the Tamil Aspectual System studies in Linguistics and Language Learning, Vol. 8, Dept. of Linguistics, U. of Washington, Seattle.

[8]. __________________. 1993: ``Intervocalic V-deletion in Tamil: Its Domains and its Constraints." Journal of the American Oriental Society 113(4)513-528.

[9] ___________________1972 "The aspectual marker iru." Journal of Tamil Studies 6,2:31-43.

[10] Steever, Sanford B. 1983. A Study in Auxiliation: the Grammar of the Indicative Auxiliary Verb System of Tamil. Unpublished University of Chicago Ph.D. Dissertation.