Postpositions.

Having made the claim that there is no clear cut distinction between case and postpositions in Tamil except for the criterion of bound vs. unbound morphology, we are forced to examine all the postpositions as possible candidates for membership in the system. Actually this is probably going too far in the other direction (and probably beyond the scope of this paper), since then almost any verb in the language can be advanced to candidacy as a postposition. There are the clear-cut bound morphemes (-ai, -ku, -il, -aal, etc .) and then there is the set of almost-bound morphemes that occur quite regularly with one or more of the pakka case morphemes. These are usually de-nominal forms with some semantically-locative meaning (meele, kiire, pakkattule, ulle, the points of the compass, hand directions) but there are also a number of de-verbal forms that, by virtue of their frequency of use, seem to have acquired independent status as postpositions. These are either infinitival or past-participial forms: -aaka, patti `about', tavira `besides, instead of', seendu/seettu `together, paattu `at, toward' (psychologically), poola `like', and a few others. In some cases, the verb from which these are derived is archaic or obsolete as a main verb, leaving the postposition `orphaned', at least in ST, with no verbal origin to act as its derivational source.

Other de-verbal forms, such as the so-called ablative irundu (which is affixed to the locative -le, kitte, or semantically-locative postpositions such as meele, kiire, or adverbs such as ange `there', have been considered from earliest times to fill the bill as case forms despite their clear formation on the locative, but because of the frequency of occurrence and unchallengeable independent status of the copula (iru) of which irundu is the past participle, a truly ambiguous situation results for which the grammarians have no comfortable explanation. Another good example of the questionable status of the ablative as a `true' case is shown by what happens when clitics oo and aavatu are affixed to ablative-marked adverbs such as enge. One would expect that clitics, which are always found word-finally in Tamil, would be suffixed to the ``case" marker irundu but in fact parallel to examples where this occurs, i.e. engerund-oo ``from somewhere or other" we also get engeyoo-irundu ``ibid." If the ablative were a true case it should not be possible to split it with something as word-final as a clitic.

We could also, of course, include other verbs not commonly used as postpositions, such as the past participle of poodu, `put' and the past participle of the Literary verb kol `hold' (Asher 1982:112) which can substitute for the instrumental, as in

#&&  # &a. &katti &poottu &vettunga &`Cut it with a knife' & & knife &hvg-put &cut &

or

#&&  # &b. &katti &kondu &vettunga &`Cut it with a knife' & & knife &hvg-held &cut &

instead of

#&&  # &c. &katti- & aale &vettunga &`Cut it with a knife' & & knife & instr. &cut &

It would be convenient if we possessed a linguistically-universal metric for evaluating case systems with which we could compare particular linguistic systems to determine what is or is not a valid system. I know of no model that can serve as a touchstone, although there are various attempts, ranging from the Sanskrit karaka system to Fillmore's The Case for Case. Languages with elaborate case systems such as Finnish and Hungarian cry out for comparison, but I suspect that underlying the case morphemes in these luxuriant case systems there are postpositions derived from some other form-class.