r and s in Turkic

[Lyle Campbell]

----------------------------Original message----------------------------
In view of the several messages on the problem of Turkic sound
correspondences involving r (and liquids) and s (and sibilants), I'll throw
into the mix a small bit about it taken from a recent paper of mine
evaluating Dolgopolsky's Nostratic.  Since this bit starts with one of
Dolgopolsky's specific proposed Nostratic forms, the content relevant to
the discussion of the Turkic problem becomes clear only towards the end of
it all.  (I claim no first-hand knowledge of the topic, and attribute most
of the significant content to Juha Janhunen, except for any mistakes I may
have made -- sorry for any diacritics which do not come across).
        Lyle Campbell
 
        [48] *p'oK'ü 'wild cattle, pack' (Indo-European *pek'u / *pek'we-
'cattle'; Altaic *p'ok'ür'- 'bovine animal, bull').  This set clearly
involves borrowing.  The Altaic *p'ok'ür'-, represented only by Turkic, is
a clear example of a documented loan, involving one of the strong points
among the arguments of those who oppose the Altaic hypothesis.
Proto-Turkic *s split into s and z, and *S ("sh") into S and Z ("zh") in
specific environments (involving roots of two syllables and with long
vowels), and then in the highly influential Bulgaric (Chuvash) branch of
the family z > r, and Z > l.  As a result, words in Mongolian (and
Tungusic) which have an  r or l corresponding to s, z, S, or Z in other
Turkic languages can only be borrowings from this branch of Turkic, not
true cognates to other Altaic languages (or they are accidental
similarities).  There is a sizeable number of these in the Dolgopolsky's
putative Altaic lexical comparisons.  In this case, in set [48], the word
involved is Proto-Turkic *pöküs 'bovine', borrowed from Bulgaric into
Mongolian and from there on into Tungusic (Janhunen 1996a:240-1, 255).
This set would be questionable in any case, given the important role of
cattle in the prehistoric cultures from the area of the Proto-Indo-European
homeland and in the territory of the various so-called Altaic languages.
        I should add here that Starostin and Dolgopolsky (in discussion in
the symposium) disputed this interpretation of the Turkic facts, preferring
reconstructions of Proto-Turkic which reflect the liquids rather than the
sibilants and in this way they deny that borrowing is a problem for these
"Altaic" forms.  This interpretation would require assuming that the
liquids (l/r) were original and changed to sibilants in certain of the
Turkic languages, a kind of sound change seldom seen in the world's
languages, though changes in the other direction are common (as in
rhotacism).  There is considerable literature precisely on this topic.
Among Turkologists, those who believe in the Altaic hypothesis (as well as
Doerfer, who opposes Altaic, though he holds Mongolian forms in these
comparisons to be Turkic loans) postulate original liquids (which then
would make the sibilants of other Turkic languages the results of later
sound changes); those who oppose the Altaic hypothesis (with the exception
of Doerfer) hold the sibilants to be original (which makes the liquids the
results of later sound changes).  Put differently, those with a grasp of
phonological systems and phonetic plausibility all postulate a change of *s
> z > r  and  *S > Z > l, where the steps in the change are seen as
incremental, intimately interrelated, and natural.  No one with a sense of
phonology postulates the reverse, the unnatural and implausible changes of
*r > z/s  and  *l > Z/S, which are almost unknown in languages elsewhere
(Shherbak 1986b, Janhunen personal communication).
 
[[From:  Campbell, Lyle. In Press. Nostratic and Linguistic Palaeontology
in Methodological Perspective. In:  Nostratic: Evaluating a Linguistic
Macrofamily, ed. by Colin Renfrew and Daniel Nettle.  Cambridge: The
McDonald Institute for Archaeological Research.]]