Home | Biodata | Biography | Photo Gallery | His Work | Tributes
[Back to Tamil Studies List]

Tamil Studies

Proceedings of the Second World Tamil Conference, Madras, 1968, pp. 314-318
Gift Siromoney


Tholkaappiyam is the earliest known grammatical work in Tamil literature and is assigned to the fourth century B.C.1 This early date assigned to the work is not unreasonable since the recent excavations at KaavirippuumpaTTinam have confirmed the existence of a flourishing Tamil civilization in 200 B.C. on the basis of carbon dating techniques. This takes the date of Tholkaapiyam to the Paninian or the pre-Paninian period.2 Staal has studied intensively certain types of grammatical rules used in Asht@a$dhya$yi of Pa$n@ini.3 These rules are of the same type as those used in modern generative grammars in the study of computer programming languages.

Chomsky, in his theory of generative grammars develops what he calls context-sensitive and context-free grammars.4 Context-free grammars consist only of context-free rules and they generate context-free languages. Context-free languages provide an excellent approximation to programming languages such as ALGOL used in modern computers. Context-sensitive languages are generated by context-sensitive grammars which consist of context-sensitive rules. Compared to context-free languages, context-sensitive languages provide a better approximation to natural languages.

We define a context-sensitive rule as follows:

ax b-->a y b

or x is replaced by y in the context a-b where a and b are strings and the symbol ---> is read as "replaced by". In Chomsky's Phrase Structure Grammar, x is a single letter and y is either a single letter or a string of letters and neither x nor y can be null. For example when the words maN and kuTam are combined we get maTkuTam.

maN+kuTam ---> maTkuTam

where the symbol "+" expresses concatenation. Here the letter N changes to T and this is called meypirithaathal in Tholkaappiyam. In a different context the letter N may not undergo any change. For example

maN + viiTu ---> maNviiTu

This is called iyalpuppuNarcci and we represent it symbolically as

For the purpose of describing grammatical rules in Tholkaappiyam we shall relax Chomsky's conditions and include cases when x or y is null. When x is the null element we get the rule a b---> ayb and this corresponds to mikuthal vikaaram. When y is null we get axb ---> a b which is called kuntal vikaaram in Tholkaappiyam.

In this discussion we have so far assumed a 30-letter alphabet consisting primarily of the 18 consonants and the 12 vowels. In this system we have shown that meypirithaathal or thirithal vikaaram corresponds to the Chomskian context-sensitive rule. Consider the rule 

This is an example of mikuthal since the letter y is introduced on the right side. Supposing we use an end-marker " # " to denote space or the end of a word, then our alphabet will have 31 letters including " # ". Then the example for mikuthal vikaaram is written as
kavi + #azhaku# --->kaviyazhaku#
where y replaces the space symbol # and this rule becomes context-sensitive in the Chomskian sense. Again if the common 247-letter alphabet is used, then
ka-vi +a-zha-ku---> ka-vi-ya-zha-ku,
where a is replaced by ya. Here again this is a context-sensitive rule in the Chomskian sense. When the null element is used, two rules such as axb--->a b and a b---> ayb are together equivalent to a single rule axb---> ayb.

The study of these grammatical rules in Tamil has a special importance today in this era of mechanical translation and generative grammars. The rules for the formation of words and those for the combination of words in a sentence have to be studied afresh. The study of Dravidian number names and their generative grammars have shown the existence and the importance of context-sensitive rules in Tamil.5 Such rules can be programmed in a computer using the COMIT "language" developed at the Massachusettes Institute of Technology.

Tholkaappiyam is divided into three sections and the first is on ezhuththu which literally means  'letter', the second on col (word) and the third on porulh (substance). Each section is divided into nine chapters and in the ezhuththu section, rules of sandhi called puNarcci are in six chapters out of nine. There are two distinct classes of rules of sandhi. The first is with reference to complete words in a sentence and the second is with reference to the formation of words. The formation of words includes the formation of number names, case-terminations of nouns, terminations of verbs and the conjugation of verbs with suffixes. Apart from general rules special rules which deal with particular words are given.

When two words (or morphemes) are combined according to a given rule the resultant will either exhibit no change (iyalpuppuNarcci) or some change (vikaarappuNarcci). According to Tholkaappiyar, the change can be of three types, viz., deletion of a letter, introduction of a new letter and substitution of one letter by another. In this paper we shall discuss a few important types of rules and no attempt will be made to enumerate all the rules.


The rule for the combination of  "vowel with vowel" is given in verse 141 as follows:

"To all words which end in vowels, if followed by vowels, the use of conjoining consonants is not forbidden."

For example kilhi+arithu--->kilhiarithu without the conjoining consonant (uTampaTumey) and alternatively kilhi+arithu--->kilhiyarithu with the conjoining consonant y. The first we shall denote by the rule [i--->i]a and the second by [i--->i+y]a which together with the first can be written in the form [i--->i+(y)]a where (y) shows that the use of y is optional. The use of the conjoining consonant is made obligatory in the twelfth century grammatical work called Nhannuul which treats the subject in greater detail.6 A rule [i--->i+(y)]a together with the rule [ii--->ii+(y)]a shall be written compactly as

In this notation the rules of the conjoining consonants are:

The following rule for ee has the two optional consonants y and v.

Exceptions to these rules are given in the seventh chapter which deals with all the words ending in vowels. In Tamil there are three demonstrative bases, called cuTTu, each of which is a pure vowel; a the remote, i the proximate, and u the medial demonstrative. The rules governing these demonstratives have to be collected from a number of rules, since some rules refer to earlier rules. They can be written as a single rule as follows:

All the general rules and the special rules are different from those of Panini. For example in Panini

tava + idam ---> tavedam  and dadhi + atra ---> dadhyatra.
If the rules of Tholkaappiyam were used
tava + idam ---> tava idam or tavayidam and dadhi + atra ---> dadhi atra or dadhiyatra.

In Tamil the short u is called kuttiyalukaram and the short i kuttiyalikaram and the rules governing these are discussed separately in the ninth chapter. In modern Tamil prose the short i has gone out of use and the short u cannot be distinguished from the standard u in speech even by sonographic analysis.


There are several special rules for certain specific words and number names. For example the rule for the combination of the word for six with a word for a unit of measure is given as follows:

aatu+uzhakku---> aatuzhakku or [u---> ]u.
Again verse 469 gives [u---> ]aa which yields
aatu+aayiram---> aataayiram.

However, some of the number names are derived in a highly artificial way and it is quite clear that those number names must have been well-established before the time of Tholkaappiyar. In his attempt to derive tholhlhaayiram from onpathu and nhuutu he uses a series of rules of deletion and introduction. Caldwell, commenting on this type of derivation, says " The plan of deriving any thing from anything was evidently not unknown to the ancient grammarians of the Tamil country"! 7


Tholkaappiyar divides sandhi into two classes called veettumai and alvazhi. The former deals with the six case-terminations of nouns where the appropriate suffix need not be present explicitly. All the other types of sandhi are classified as alvazhi. Separate sets of rules are given for these classes and there is no simple mechanical device for deciding to which class a given example belongs. The combination of the words depends upon the sense in which they are used. Further when a string of words is given there need not be a unique way of arriving at the original set of words. This problem of ambiguity and decipherability was recognized by Tholkaappiyar and is given in verse 143, which runs as follows: "Those are to be understood in context. They cannot be written in different characters."

In the computer programming languages, ambiguity must be avoided but in a language like Tamil it is sometimes deliberately cultivated for purposes of style. If a poem is composed in a certain style called cileeTai , then the poem can be interpreted in more than one way.

Tholkaappiyar did not treat his grammar as a static body of rules to be strictly kept by the users but did say that rules other than those found in his work could be obtained directly from the actual usage in Tamil (verse 406). It will be interesting to trace the changes in the rules of Tamil grammar from Tholkaappiyam to the present day and to compile a complete set of rules for modern Tamil prose on the basis of contemporary usage.


1 M. RAJAMANICKAM. " Date of Tholkaappiyam ", Annals of Oriental Research, Vol. 19, Part II (1964).
2 S. ILAKKUVANAAR, Tholkaappiyam in English, Madurai, 1963, p. 3.
3 J. F. STAAL, "Context-sensitive rules in Panini", Foundations of Language, Vol. 1 (1965), pp. 63-72.
4 N. CHOMSKY and G. A. MILLER, Handbook of Mathematical Psychology, New York, 1963, Vol. II, p. 294.
5 R. SIROMONEY, Grammar of Dravidian number names, Foundations of Language, Supplementary Series, (Forthcoming).
6 K. VELLAIVAARANAAR, Tholkaappiyam-Nhannuul, Ezhuththathikaaram, Annamalai University, 1962, p. 161.
7 R. CALDWELL, A comparative grammar of the Dravidian or South- Indian Family of Languages, University of Madras, Madras 1961, p. 350.

Go to the top of the page

Home | Biodata | Biography | Photo Gallery | Publications | Tributes