NEGATION IN TAMIL: SEMANTIC AND SYNTACTIC ASPECTS Harold Schiffman University of Pennsylvania Handout for LING 319/519, SARS 319/519 GRAMMATICALIZATION

Negation in modern Tamil is a problematical area for scholars and grammarians interested in syntax and semantics. While there is a simple negative marker -le that is affixed to infinitives to mark negation, there are many more complex forms in the language that cannot be treated as a variation of this simple form, and even if negation were shown to be a simple morphological process, many semantic complexities and anomalies of the system would be obscured. Most western grammars¹ of Tamil have not adequately dealt with Tamil negation. Often the point of departure has been a European, sometimes even a Latin, model against which the Tamil system is contrasted. Since European languages usually form negatives in a fairly straightforward manner by the addition of some negative particle to even otherwise complicated verb phrases, little thought is given to how negation works in Tamil as a system.

In recent years a number of studies have appeared that attempt to handle the complexities of Tamil negation within various theoretical frameworks, with varying degrees of success (Subramoniam 1959, Agesthialingom 1967, Schubert 1970, Rangan 1971, 1972). What would be useful would be a systematization of the complexities of the surface-structure negative forms that remains simple, avoids constructing complex paradigms or elaborate semantic constructs, but accounts for the data within a known framework. Of these analyses, only Rangan (1971, 1972) begins to meet these criteria, but even his analyses ignore some of the complex negative forms. What complicates the issue, of course, is that the Tamil verbal phrase, as is the case in most Dravidian languages, is itself very complex, involving sometimes several main verbs, verbal participles, aspectual verbs, " dais, or auxiliaries combined in such a way that the boundary between morphology and syntax is often indeterminable. When negation of any of these is introduced, the semantics of the whole phrase can become extremely intricate. If each of these complex negative constructions were to be treated as part of a morphological paradigm, a complete catalogue of them would consist of a list of all the sentences in the language. This is obviously undesirable.

Another aspect of negation in Tamil is that, unlike modern Indo-European languages, where negation is often accomplished by the affixation of negative `particles' such as English not, Russian ne, German nicht, Spanish no, to basic positive verbs, Tamil exhibits several negative forms that often cannot be directly derived from the positive forms, or at least not from a single positive form. I will call these forms syntaotic negatives to distinguish them from the morphological negative, which are older and simpler.

Historically, older forms of Tamil possessed a simple tense-aspect system, and separate negative paradigms for verbs existed alongside positive paradigms. Apparently the two sets of paradigms were equivalent except for the presence of the notion Negative in the non-positive forms, and furthermore negatives were tenseless; thus the negative `morpheme' replaced the tense morpheme, but all other notions remained the same.

The Old Tamil negative kaaNoom survives now in Tamil as a kind of frozen idiomatic form, used to expressed the notion `I don't see (anyone); where is everybody? where'd everybody go.'

In the subsequent historical development of the language from Old Tamil to Modern Tamil, some portions of the negative paradigms have been lost, while some new negative forms have developed that have re-used or reinterpreted (re- lexicalized) older material to produce new negatives that are not only syntactically " re complex than the older ones, but are also more complex syntactically and semantically than the `corresponding' positive sentences. I say `corresponding' because it is in many cases difficult to ascertain which negative sentence corresponds to which positive one.

What seems to be happening is that neutralization of some semantic material occurs in the positive, furthermore, some of the new negatives (syntactic negatives) are semantically ambiguous, where their positive counterparts are not, since by their very complexity it is possible to interpret them in varying ways while their positive counterparts, being simpler, have no such ambiguity. We shall see however, that this notion of the positive being unambiguous is illusory, since by the addition of time adverbs to the sentence, we observe homonymy in the positive as well as in the negative.

Typical grammars (Schiffman 1999) give the following paradigms for positive and negative Tamil sentences (non-Brahman spoken dialect).

Ex. 2a

Agesthialingom (1967: 215) states that varalle (come-not) can mean either `did not come,' `do/does not come,' or `will not come', and gives as an example the following paradigm:

Ex. 2b

In the `morphological' negative the distinction of tense is typically neutral ized, so that the negative form with -lle can be either past, present, or future, as is obvious if we include time expressions such as neettu `yesterday' iNNekki `today' and naaLekki `tomorrow'. The `alternate' future negative form, pooha-maaTT-een however, can only mean future, i.e., can only co-occur with time adverbs that imply time subsequent to the present, e.g., iNNekki `to day' or naaLekki `tomorrow.'

The future tense illustrates the `skewed' nature of the system admirably. The unambiguously future positive form consists of verb stem plus future marker plus PNG:

Ex. 3

poo-v-een `I will go'

iru-pp-een `I will be'

Depending on the time adverbs that can co-occur with this `future' form, we see that it has a number of separate readings, one being `distant or indefinite future', another `habitual'.

Ex. 4a-b

• (4a) avan aDutta vaaram uur-ukku poo-v-aan
  "He will go to town next week" [Distant future]

• (4b) avan aDikkaDi sinimaa-v-ukku poo-v-aan
  "He often goes to the movies" [habitual]

Depending on the time adverbs, the simple future has two distinct meanings, at least when translated into other languages, and when negated, two different forms appear. The negative of 4a would be

Example 5a

whereas the negative for 4b i6 in form quite different:

Example 5b

If we are to try to establish a `global' notion for the simple future in Tamil (i.e., the forms in v and pp ), combining the notions `distant future' and `habitual', we find that simple negation of either component of this `global' semantic unit gives us separate forms that are not interchangeable. We have thus a kind of homonymy in the positive such that the notions `distant future' and `habitual' are expressed by the same grammatical form, but distinct forms appear in the negative of both categories. Further" re, there is homonymy in the negative as well, which further skews the system. The form pooradulle, which represents the meaning [habitual negative], also expresses the notion [immediate future negative ], (not [distant future negative]) and its positive analog, e.g., pooreen, is ambiguous, meaning either [immediate future] or just plain [present]. We have in each case two forms with four meanings, but the two negative forms do not match neatly with the two positive forms:

Example 6

In the negative, the form pooha-lle, which represents present negative, as we have already seen, also is the standard past negative, while the positive past has a different form, poo-n-een. For the five categories listed, therefore, we have three forms each in the positive and negative, but they do not match up one-to-one.

The attempt to establish `global' semantic categories which cover the territory adequately enough to handle both positive and negative forms of verbs is thus fraught with tremendous difficulties because of the overlap of one form having two meanings or more, but not the same two meanings for both positive and negative.

Obviously, the question of how to form negative sentences in Tamil, or how to establish a set of positive sentences and a set of negative sentences and to relate them semantically, could be answered in a number of different ways.

1. One might be to consider that there are no correspondences between negatives and positives in Tamil, and to let it go at that. This is unacceptable from a number of points of view, one being logically untenable, another semantically untenable, and from a universal point of view, somewhat strange.

2. Another solution might be to look to a Praguian sort of analysis, claiming that negatives are marked and positives are unmarked, and marked sentences contain semantic information that is semantically neutralized in the positive. This is essentially what is done with Slavic perfectives and imperfectives, the perfective being the marked form and therefore having redundant features not found in the imperfective.

3. This point of view unfortunately tends to toss in to the wastebasket the question of why the negative system in Tamil should be skewed, and how it happens that there is so much polysemy and syncretism in the system. (But review what Schwegler and others have said (below) about the tendency for negative expression to be overdeveloped...)

4. A third analysis, utilizing Fillmore negative-transporation rule (Fillmore 1963) has been posited to account for the fact that the position of not in an English sentence can sometimes vary without changing the meaning, while other times it changes the meaning:

Ex. 7a

and

are seen as more or less semantically identical, so the position of not is free, as contrasted with:

Ex. 8a

versus

where not can not be "transported" from the embedded sentence to a higher sentence without changing the meaning. Another way to view the subject, using the concept of scope of negation is illustrated by some syntactic differences in French negative sentences: In the first example (9a), the partitive de is marked 'singular' when it occurs with the negative ne (which is the general 'rule' of French grammar for 'negative partitive' constructions.)

Ex. 9a

But in Ex. 9b the partitive is marked 'plural', and the meaning is different:

The difference between 9a and 9b is that in the first sentence, what is negated is the proposition [X have bottles], whereas in 9b what is negated is simply [E bottles]. This is indicated in French by a difference, not in the position of the negative particle (as in English), but in the form of the `partitive' preposition de versus des

This difference in the scope of the negative is handled in transformational grammar by bracketing in different ways the portion of the tree structure that is negated, as shown in Figure lO.

Examples 10a and 10b

       S                                 S
      / \                               / \
     /   \                             /   \
    NP    VP                          /     \
    |    / \                         /      VP
    |   /   \                       NP     /  \
    |  Neg  VP                      |     /    NP
    |   |   / \                     |     V    / \
    |   |  /   \                    |     |   Neg N
    |   |  V   NP                   |     |   |    \
    Je NEG avoir bouteilles         Je  avoir NEG  bouteilles

By later rules, the surface structures of the two sentences are derived, 10a with de and 10b with des as the surface structure marker of the differing scope of the negative.

In Tamil, also, it seems that the notion of scope of the negative is explanatory, especially with regard to the modal system, where again there can be a number of different negative forms paralleling simple/single positive forms:

Ex. 11a-c

• (11a) naan pooha-Num `I must (need/want/should) go'
• (11b) naan pooha-veeNDaam `I need not/don't want to go'
• (11c) naan pooha-kuuDaadu `I mustn't/shouldn't go'

The two modal forms veeNDaam `need not, don't want' and kuuDaadu `should not, must/should not', can both be related to (lla), particularly as answers to the question (lld) naan poohaNum- aa? `Must/should I go?' Why we should get two possible answers to the question (lld), and which answer is most appropriate, depends on the scope of the negative:

Tree-structure (bracketing) of 11 a-c:

             S                                  S
            / \                                / \
           /   \                              /   \
          /     \                            /     \
         NP      VP                         /       \
         |       / \                       NP        VP
         |      /   \                      |        /  \
         N     VP    VP                    |       VP   \
         |     / \    \                    N       /\    \
         |     V  AUX  Neg                 |      V Neg  AUX
         |     |    |    |                 |      |  |    |
       naan   pooha Num  NEG              naan pooha NEG Num
                    veeNDaam                      kuuDaadu

        I      go must/need Not            I   go  Not must/should
         `I needn't go'                   `I shouldn't/must not go'

Because of portmanteau morph-ism, here, the sequence `modal [Num] + [Neg] is realized lexically as veeNDaam , while the sequence [Neg] + modal [Num]' is realized lexically as kuuDaadu.

Other examples of different scopes of negatives are illustrated in the following durative negative forms.

Ex 12a

that is, the bracketing is [I be Neg [eat] durative] i.e. "I was doing somethlng duratively, but it wasn't eating."

versus Ex 12b

i.e., the bracketing is [I be Neg [eat durative]] `I continued to not eat'.

Here the difference between the bracketing of NEG results in two forms that are surface-structurally different, but semantically involve simply a different placement of constituents, i.e., the scope of the negative is different. In almost every case of complex negative forms such as 12a and 12b, the difference can be explained as a difference in scope of the negative, and this explanation can reduce the paradigmatic complexity of the Tamil negative system immensely--we simply eliminate certain paradigms, because they properly are part of the syntax rather than of the core grammar. That doesn't mean that the morphology is totally simple, but it is perhaps simpler than Rangan and/or Rajaram claim. What has complicated matters is that some of these syntactic bracketings have become grammaticalized in different ways, with forms like kuuDaadu having the apparent form of a root kuuD- (which shows up in LT in the positive as well, e.g. kuuDum ) plus a negative allomorph -aadu (used with many other verbs as a negative future impersonal). But now the morpheme kuuD- must be treated as a cranberry morph, or as fused with -aadu since it never appears elsewhere in ST.

We mentioned earlier the problem of homonymy, which we have yet to propose a solution for. It seems evident, however, that if there are more negative forms than positive ones, and some negative forms are polysemous, then the positive forms are not overtly expressing some semantic information that the negative forms are. Two solutions spring to mind here:

One would be to posit neo-Bloomfieldian zeroes a la Trager (1953) wherever there is a semantic category expressed in the negative that is not expressed in the positive, and then have rules that supply no lexical matter for the positive, but do so for the negative. That is, except for the notion [Negative], the semantic (deep remote), structure for positive and negative sentences would be the same, but since no overt surface-structure morphemes are available to represent certain notions in the positive sentences, those nodes are simply not filled. In the negative, of course, they are sometimes filled with the proper lexical item, and sometimes two nodes are represented by one lexical item, the so-called portmanteau effect. A second solution is look at how grammaticalization of negative forms has come about, and see if that approach can tell us something about why there are sometimes more negative forms of verbs than there are positive ones, and why the portmanteau effect is different in the two. The first "explanation" is of course what H&T have referred to as the propensity or "psychological proclivity" (Schwegler 1988) for the development of negative emphasizers and that they have their point of origin in contexts of contradiction i.e. emotionally loaded contexts. A proliferation of negative forms is perhaps of more use in discourse, in arguments, in seeking agreement (or not), and verbs have been recruited from lexical items and grammaticalized in various ways (the way the French particles pas etc. have). This reminds us again of the importance of 'speaker centered-ness' in grammaticalization. (See also recent study by by Schrauf. We must also look at the historical morphology of older Tamil, where tense marking was not so common, especially for stative verbs. With stative verbs, with dative subjects, we have an emphasis on marking habitual states and the Old Tamil morphology for this has lingered on in the modal verbs, which in large part deal with states: -probability, ability, knowledge, liking, etc. In the future habitual and distant future, where the positive has one form -um for the two exhibited by the negative, we have similar underlying structures, but no lexical material being available for the positive, none is supplied: The notion [habitual] which for semantic reasons is in some sense more of an aspectual type of notion (Schiffman 1969) needs to be handled here as well. The two underlying structures are identical except for the absence of a node for negation in the positive, naturally. The node [habitual aspect] is realized lexically as the future marker v- in the positive, while in the negative it takes the lexical form -radu, identical to the verbal noun, before the negative - lle. For a more recent treatment of this topic, go to this link.

If such a treatment seems to throw no light on why there is so much homynymy and portmanteau mophology in the Tamil negative system, I can only say that I do not know of any explanation other than an historical one, i.e. the grammaticalization of modals and negation has worked at a different level than the positive and tense-marked verbal system. (Again the proclivity to develop more emphatic negatives has resulted in recruiting various lexical verbs to serve as negatives, especially negative modal verbs , which have come equipped with the archaic negative morphology in existence when they were just lexical verbs.)

This brings up the subject of the existence of an archaic negative system, which has left remnants in various portions of the morphology, generally in reflection of the previous `tense-less' nature of the negative system. Parallel to this has been developed a grammaticalized system (as opposed to the older morphological one) which has been created to handle negation where aspect and tense is involved, since apparently tense and particularly aspect were not well-developed in archaic Tamil at all. The language has changed, and new forms of negatives have evolved to handle this. I have not even mentioned some other kinds of negatives, e.g., the `obstinate' negative, which is not even found in modern Literary Tamil, but occurs only in the spoken dialects (Schiffman 1999).4

NOTES

1. Here and throughout, examples are given from modern Spoken Tamil (ST) rather than from Literary Tamil. This is because ST has an even more complex negative system than LT, and because most recent studies have been of ST rather than of LT. Tamil material here is enclosed in slashes (italic) or is italicized; semantic notions are enclosed in square brackets (C]) or italicized and underlined.

2. For example, Rhenius 1853, Lazarus 1878, Pope 1905, Arden 1942.

3. Rhenius (1853:177) for example explains the various modal forms by giving examples of sentences in question form, with possible answtrs to these questions, both positive and negative.

4. It seems clear to me in the light of the above that more work needs to be done on the semantics of existential quantifiers and particularly the modal and stative verbs before definitive statements can be made. What is not needed is a proliferation of paradigms, a la classical grammar, listing all the negative forms that are possible. Paradigms lack explanatory power, whereas syntactic analysis indicates that at least some of the paradigms have constituent structure, and by taking into account the scope of the negative, some clarity is achieved. We also need an evaluation procedure to help ascertain when two sentences are exactly equivalent except for one being positive and the other negative. While it often seems clear in English and other western languages when a sentence is the negative of another, this kind of judgement is not so readily available in Tamil; perhaps this kind of question is an artifact of western education and/or Aristotelian logic, which has no place in traditional Indian grammatical tradition.

1. Agesthialingom, S. (1967) A Generative Grammar of Tamil. Annamalai University, Annamalainagar.

2. Arden, A. H. (1942) A Progressive Grammar of the Tamil Language. Madras, Christian Literature Society.

3. Chomsky, Noam. (1957) Syntactic Structures. Mouton, The Hague.

4. Fillmore, Charles. (1963) `The Position of Embedding Transformations in a Grammar.' WORD 19:208-31.

5. Lazarus, J. (1878) A Tamil Grammar. Madras, Addison and Company.

6. Pope, G. U. (1905) A Catechism of Tamil Grammar. Oxford, The Clarendon Press.

7. Ramanuj an, A. K. and E. Annamalai. (1968) Preliminary Studies for a Reference Grammar of Tamil. Chicago, mimeographed.

8. Rangan, K. (1971) `Negation in Tamil' in Biligiri, H. S. (Ed.) Papers and Talks, Mysore.

9. __________(1972) `Some Aspects of Negation in South Dravidian Literary Languages' in S. Agesthialingom and S. V. Shanmugam, (Eds.) Third Seminar on Dravidian Linguistics. Annamalai University, Annamalainagar.

10. Rhenius, C. T. E. (1853) A Grammar of the Tamil Language. Madras, American Mission Press.

11. Schiffman, Harold. (1969) A Transformational Grammar of the Tamil Aspectual System. University of Washington, Studies in linguistics and Language Learning, Vol. VII. Seattle.

12. _________(1999) A Reference Grammar of Spoken Tamil. Cambridge: Cambridge University Press.

13. Schubert, James. (1970) `A Semantic Classification of Tamil Negation.' Berkeley conference on Universals in Indian Grammar.