Fortunatov's Law and tolkAppiyar's rules

Sudalaimuthu Palaniappan Palaniappa at AOL.COM
Mon Aug 3 00:26:09 UTC 1998

This posting is long because I am combining respones to two postings into one.

In a message dated 98-07-30 13:49:37 EDT, vidynath at MATH.OHIO-STATE.EDU writes:

<< >This is what Burrow says: "Intervocalically a single unvoiced consonant in
 >Sanskrit very often replaces a double consonant in Dravidian.

 Analogical reasoning would lead us to expect single intervocalic stops
 in Dravidian loans to be voiced in Sanskrit. But there is no apparent
 pattern here. Hamp restricts F's law to l[t|th|s]. This makes the
 problem worse.

 Even in case of doubled stops, there are examples with doubled stops in
 Sanskrit also: ku.t.t, hu.dukka etc. As Drav. shows variation between
 -CC- and -C- between languages, these are best taken as reflecting the
 sound of the source langugage.>>

I do not think this particular loan phonology issue has any relevance to the
operation of Fortunatov's law and Tolkappiyar's rules we are discussing. The
point we are interested in is if a "T" in IA could be a result of the l/L+t
similar to that discussed by Tolkappiyar's rules with respect to Tamil. The
answer is yes. As I had pointed out more than once, doubling is not mandatory.

 <<>This example also shows variations between alveolar, dental and

 All the examples of variation are based on _t/_t_t. We don't see t or
 .t changing. What we see here is change in Dravidian dialects where _t
 changes to something else, and the Sanskrit borrowings reflect what
 happened in the source language. This is very different from
 `Dravidians >mis<pronunciation of alveolars as retroflexes' theory
 which implies that >all< alveolars must have become retroflexes.>>

Examples relevant to the discussion have already been given in the past.

 <<>In many colloquial situations, L does not become T before consonants
 >like k, c, and p. Malayalam which has severed its connections with
 >centamiz offers evidence for the unique nature of t vs k, c, and p.

 The trouble is that these are attested at a much later date. The
 presence of words like `kalki' shows that -lk- is acceptable in Modern
 Tamil. But that tells us nothing about the situation in early First
 Millennium BCE.>>

I am amazed at the logical inconsistency here. Vidynath bases his view of n,
n2, N convergence on a situation in Pittsburgh much farther removed in time
and space than the South Asian Malayalam. See below the excerpt from his post
in the other thread "Re: Retroflexion in IA".(The Malayalam examples used have
been attested at least a few hundred years earlier than modern Pittsburgh
dialects.) If one can base one's conclusion about convergence in early First
Millennium BC in South Asia on a situation prevailing in Pittsburgh, cannot
one base one's conclusion about L+t >T on a situation in Kerala in South Asia,
that too a few hundred years earlier?

 <<Anyway, quoting Tolkappiyam (how old is that name?)
 for one part of the argument and rejecting it elsewhere leaves
 something to be desired.>>

There is no inconsistency here. Please see my posting related to diglossia.
Also see the next note.

 <<Analogical restoration is a known process. We need to be sure that it
 did not happen here, i.e. that -lk- etc were possible in proto-Dravidian,
 but that -lt- > _t_t and -Lt- > -.t.t-. The standard sources have nothing
 about morpho-phonemics of proto-Dravidian. If we rely on Tolkappiyam
 instead, the troulbe remains.>>

There is no trouble here. Please see Leavitt's paper for a discussion of all
the rules. Tolkappiyar himself gives a special treatment for the dentals.

<< We have been talking only about .t. But, there has been no attempt to
 explain how Dravidian with its n/_n vs .n contrast could lead to n/_n
 phoneme being split into n vs .n. I have not seen any reason to change
 my view that convergence should in fact lead to collapsing of all

Please see below the discussion of Vidynath's posting in the other thread.

<<>but in this case I fail to see why the merging of all n's would be the
>necessary outcome of convergence.

I based this on something I read about dialect interactions. My laziness
in keeping a notebook has come back to haunt me: I cannot locate the
book where I read it. [This was specifically about Pittsburg dialect.]
The basic argument is as follows: Dialect A distinguishes two phonemes P
and Q. Dialect B does not. Speakers of B rely on context etc. to tell
the words apart. This works even if P and Q are pronounced differently.
But speakers of A rely only on the sound and have trouble understanding
B-speakers. With repeated interaction, A-speakers start relying on context
to tell words apart. So the difference between P and Q becomes irrelevant.

Given the distributional peculiarities in Dravidian, the same argument
should work for n. In Dravidian, n occurs word-initially and before t;
_n everywhere else. Sanskrit anu would come out a_nu. If Sanskrit speakers
hear it as, then the anu vs distinction is lost. The same
works in reverse. [I don't know PDr lexion, so I will use Tamil.] Tamil
pa_ni and will come out the same from Sanskrit speaker's mouth. In
both cases, the speakers must rely on context to distinguish the words,
leading to the irrelevance of the phonemic distinctions.>>

There are several problems with these assertions.
First of all, the phonology of Tamil n given by Vidynath is wrong. n can occur
medially -intervocalically as well as doubled. It can also occur in the word-
final position as in "verin".

Secondly, if this theory were true, IA in India should not have maintained any
phonemic distinction of voiced and unvoiced stops. (PDr did not have such
contrasts. But apparently IE did.)

 <<And nobody has even touched .s, the most commonly retroflexed sound in
 Sanskrit. I see no way of getting there from s-hacek. And why would
 Dravidian, with prominent lack of sibilants in the earliest stages
 lead to a split in sibilants? I am yet see any one pick up this

Kuiper (Aryans in the Rig Veda, p. 26) says "On the basis of class. kaluSa-
'turbid' : Tamil kalUz- 'to become turbid' it has been suggested that class. S
can sometimes stand for the Dravidian retroflex /.r/, for which in Roman
script the symbols _l, L, ..l, .r, ..r and .z are in use." Dravidianists also
posit a dialectal alternation of *-c- and *-z- in PDr itself. The similarity
between Dravidian z and Sanskrit S has been noted by Bh. Krishnamurti. BhK
says, "This may suggest that a dialectal (less probably positional) variant of
*c-, having a phonetic value similar to Skt. S was replaced by z with a later
dialect mixture which produced two series of correspondences in the same
language." (Telugu Verbal Bases, p.48) Thus at least in some Dravidian
dialects, there was a sound similar to S.

 <<In sum, it is quite premature to assume that Dravidian origin of
 retroflexes is established fact. If you have not explained .s, you have
 explained nothing about the origin of retroflexes in IA.
 <FONT COLOR="#000000" BACK="#fffffe" SIZE=3> >>
I leave it to IE scholars if this statement is justified. Further, even if
internal development did occur, what was so unique about South Asia that an
internal development was able to sustain itself at such a level as in South
Asia? Madhav Deshpande noted in his article in "The Indo-Aryans of Ancient
South Asia" that " The contacts with non-Aryans outside India did not lead to
retroflexion either in the Indo-Iranian dialects, or in the pre-Indo-Iranian
(r-and-l) dialects. The retroflexing influence was manifest in India." To me
biology gives an analogy. Genetic mutations can occur spontaneously all the
time. But only certain mutations which give some advantage towards survival of
the species in an environment get passed on in the long term. In the case of
retroflexion, whatever independent or spontaneous developments there might
have been in IA in common with IE, they were able to become permanent features
in South Asia because of the presence of prior retroflexion prevailing in
South Asia. This is my view.

In any case, I do not have any emotional attachment to the origin of IA
retroflexion one way or other. I have presented some evidence which has not
been noticed till now and which can be used by scholars with the objective of
pursuit of truth. As I cannot devote more time to this discussion, I wish
others good luck.

S. Palaniappan

More information about the INDOLOGY mailing list