centum and satem vidal 2013
TRANSCRIPT
-
7/23/2019 Centum and Satem Vidal 2013
1/4
Centumand satem
Miguel Carrasquer Vidal, August 2013
Traditioally, PIE is reostruted ith three series of dorsal stops:palatovelar * * *h(plain) velar *k *g *g
h
labiovelar *kw
*gw *g
hw
In what follows, I will use the voiceless reconstruct in every series (*,*k, *kw) as shorthand for thewhole series, i.e. unless otherwise noted, everything said about *also goes for *and *h, etc.
Some scholars, starting with Meillet 1894, have questioned the validity of this three-way
reconstruction, and prefer to reconstruct just two series:
(plain) velar *k *g *gh
labiovelar *kw
*gw
*ghw
The basic arguments in favour of just two velar series are1:
1.
The plain velar series is statistically rarer than the other two and is almost entirely absent
from morphology;
2. The plain velars, to the exclusion of the palatovelars, appear most often in certain
phonological environments (before or after /u/, after /s/, and before /r/ or /a/; also before
/m/ and /n/ in some Baltic dialects);
3.
Alternations between plain velars and palatovelars are common in a number of roots across
different satemlanguages, where the same root appears with a palatal in some languages
but a plain velar in others (most commonly Baltic or Slavic; occasionally Armenian, but rarely
or never the Indo-Iranian languages). This is consistent with the analogical generalization ofone or another consonant in an originally alternating paradigm, but difficult to explain
otherwise;
4.
The traditional explanation of a three-way dorsal split requires that all centumlanguages
share a common innovation that eliminated the palatovelar series. Unlike for the satem
languages, however, there is no evidence of any areal connection among the centum
languages, and in fact there is evidence against such a connectionthe centumlanguagesare geographically non-contiguous. Furthermore, if such an areal innovation happened, we
would expect to see some dialect differences in its implementation (cf. the above differences
between Balto-Slavic and Indo-Iranian), and residual evidence of a distinct palatalized series
(such evidence for a distinct labiovelar series does exist in the satemlanguages). In fact,
however, neither type of evidence exists, suggesting that there was never a palatovelarseries in the centumlanguages.
These are all valid points, but they cannot obscure the fact that even if we take away all the cases in
the phonological environments mentioned above, and we eliminate all forms where variations occur
within the satemlanguages (sometimes even within the same language or language family), there
still remains a solid number of etymologies where we find *kin the satem languages and *kin the
centum languages. Some of these are even important suffixal morphemes, such as the diminutive
marker *-iko-.
1Text slightly edited from the excellent introduction to the topic in the Wikipedia article onProto-Indo-
European phonology.
http://en.wikipedia.org/wiki/Proto-Indo-European_phonologyhttp://en.wikipedia.org/wiki/Proto-Indo-European_phonologyhttp://en.wikipedia.org/wiki/Proto-Indo-European_phonologyhttp://en.wikipedia.org/wiki/Proto-Indo-European_phonologyhttp://en.wikipedia.org/wiki/Proto-Indo-European_phonologyhttp://en.wikipedia.org/wiki/Proto-Indo-European_phonology -
7/23/2019 Centum and Satem Vidal 2013
2/4
The answer to the first point above has already been arrived at, independently, by several scholars
subscribing to the traditional theory (including myself).
It consists in reassigning the phonetic quality of the three series as follows:
traditional notation alternative notation
velar * * *
h
*k *g *g
h
uvular
2 *k *g *g
h*q *G *G
h
labiovelar *kw
*gw *g
hw*k
w *g
w *g
hw
This goes a long way to answering the other objections as well. Uvular consonants are indeed
universally less frequent than velar ones, and less likely to occur in morphological material. Also, the
opposition between uvular and velar may have been neutralized already in the proto-language, as
well as at later dialectal stages, in a number of environments (especially in direct contact with
another occlusive consonant). Conversely, the frequency of *kin the neighbourhood of *ais due to
the uvular consonant colouring adjacent (short) vowels, much in the same manner as the uvular
fricative *h2.The environment is here created by the back velar itself. Note that reductionist scholars
who also do not accept PIE *amay find this aspect agreeable.
In this reinterpretation, the centumand satemphenomena are reduced to phonologically trivial
frontings uvular > velar and velar > palatal3, which may have occurred independently at several
times, as also witnessed by the satemcharacter of Luwian (see Melchert 1987, 2012), which is
certainly independent of the main satemphenomenon affecting Indo-Iranian, Armenian, Balto-Slavic
and Albanian.
As I hypothesized i y artile o Brugas la, PIE *emay reflect the merger of (at least) Pre-PIE**a, **i, **u. Such mergers are not uncommon, and we find examples of them in Tocharian and the
North Caucasian languages. Typically, the merger is accompanied by palatalizations (reflecting former
*i) and labializations (reflecting former *u) of the surrounding consonants. In PIE, the existence of
labialized consonants is indisputable (*kw, *g
w, *g
hw, *h3, and as I tried to demonstrate in my paper
on Armenian -k`, also *sw). For the dorsals, this means that the labiovelar series arose where velars
occurred in the neighbourhood of Pre-PIE **u(and **u). Strictly speaking, we should also expectuvulars in the same environment to have their own labialized uvular reflexes. Evidence for this can be
gathered from the admittedly rare etymologies where PIE *kwoccurs in the neighbourhood of *a
(e.g. *(s)kwalos- hale. o, rather tha reduig the uer of dorsal series from three to two, I
would suggest to increase it to four4:
velar *k *g *gh
uvular *q *G *Gh
labiovelar *kw *gw *ghw
labiouvular
*qw *G
w *G
hw
On the other hand, these four series reduce to two original ones (velar and uvular).
But did the loss of pre-PIE *ialso produce palatalized versions of the velar and uvular series?
Palatalizations did occur in Pre-PIE dentals, which I will discuss in a forthcoming paper. But what
about the dorsals?
2oe prefer to refer to these souds as ak elar istead of uular.
3The often asserted implausibility of a backing palatal > velar is contradicted by the Egyptian Arabic backing of
Arabic *to g. It is true, however, that such cases are rarer than the opposite process.4One may wonder whether this also extended to the fricatives, where we might expect velar *x, uvular *
(with labialized variants *xw
and *w
). We have evidence for only two: *(=*h2) and *w
(=*h3).
-
7/23/2019 Centum and Satem Vidal 2013
3/4
Perhaps a case can be made, at least for the inherently unstable palatalized uvular series. Note that
we stated above that *kcoloured an adjacent *eto *a. While this undoubtedly happened in a
considerable number of cases, there is still uncontroversial evidence for *kin the neighbourhood of
*e, even if we factor out cases of Ausgleich due to Ablaut variations *o~ *e ~ . We cannot allow for
arbitrariness in the outcome of PIE *qe: it must either give *kaor it must give *ke, but not both at
random. One possible way out of this is to explain the variants as deriving from two different Pre-PIEprototypes: *kacomes from **qa, while *kecomes from **qi
5(and of course *k
wacomes from
**qu). This solution allows for an elegant re-formulation of the centum~ satemsplit.
Diagrammatically:
ka ki ku qa qi qu
centum ke ke kw
e ka ke kwa
satem ce ce k(w)e ka ke k(
w)a
All Indo-European languages eliminated the uvular stops. The centumlanguages are those languages
where **qi> *kemerged with **ka/**ki. The satemlanguages are the languages where the
development **qi> *kepushed **ka/**kito become palatovelars (at least in some environments,cf. Luwian). Both developments are quite natural, and could have occurred independently in
different branches of the IE family.
5Cf. for instance, Melcherts undisputable example of *k in a fronting environment: CLuw. ii- to o