IE, dates, etc.

Thu Feb 26 11:32:05 UTC 1998

----------------------------Original message----------------------------

On Tue, 24 Feb 1998, Johanna Nichols wrote:

> ----------------------------Original message----------------------------
> In recent postings Alexander Vovin has asked me to justify the age of
> roughly 6000 years for IE and other old families, and Alexis Manaster Ramer
> says IE is no paragon and is a young family.

    Well,  I believe Alexis, who has an advantage of being 5 hours ahead
of me has already answered this (:-), but let me reiterate. *Even if* you
manage to demonstrate that IE is 6,000 years old, it still shows only one
thing: that IE is 6,000 years old. You still cannot conclude on this basis
alone that: (a) other language families are also no older than 6,000
years, and (b) since IE is 6,000 old, there is ceiling of 6,000 to the
comparative method.

    I would probably agree with you that IE split between 5,000 and 6,000,
but I still think that it is a mere *guess-work* supported by oblique
evidence, but not by an evidence from IE itself. Please see below.

>
> The dispersal of PIE is one of the best-dated ancient linguistic events on
> earth.  Evidence comes from several sources:
>
> (1)  Glottochronology.  This is actually reasonably reliable, provided you
> have (a) enough daughter languages to do several different pairings
> (glottochronology  uses a binary comparison),

    How many would be "enough"? And on what *linguistic* basis one would
decide what is "enough" and "not enough"?

 (b) an idea of the deepest
> branching structure, and (c) an idea of which daughter languages or
> branches are most divergent and which are most conservative.  (a) is an
> accident of fate and means that glottochronological dates are most reliable
> for larger families.  (b) and (c) come from standard comparative method.
> The median glottochronological age for the comparisons described in
> Tischler's 1973 monograph is around 5500 bp as I recall off the top of my
> head.
>

    Glottochronology was busted so many times that it became almost
tedious to go over it. But, well, let us do it again. The basic fallacy of
glottochronology lies in the fact that it a priori assumes that *all*
languages change *at the same rate all the time*. This is simply not true
not only regarding *different* languages but even one and the same
language. Examples are abundant in literature, but let me add few more
from the languages of East Asia I know best.

   (1) No matter how many pairings you do with main islands Japanese
dialects and Ryukyuan dialects (actually, they are languages)  they all
would point to the split between
Japanese and Ryukyuan dated by approximately 5 century C. E. Even someone
who is only superficially familiar with the history of these languages
would tell that this is complete nonsense.

   (2) If we did not know that Middle Korean was actually the language of
the 15th c., we had to assign much older date to it on the basis of
glottochronology: the language went berserk in the last 500 years and
replaced (including loans from Chinese) much more basic vocabulary than it
was supposed to do according to glottochronology.

   (3) I do not remember exactly off the top of my head, but it seems to
me that if you compare glottochronologically Old Chinese with Modern
Mandarin, you would get much shallower time than 7th c. B. C. E.

    Etc., etc. Thus, Swadesh just got lucky with his trials of European
languages. It does not necessarily work in the other parts of the world.
Now, if it is true, it does not matter how many pairings you'd get: they
all might have changed at different speeds.

> (2)  Linguistic paleontology, etc.  PIE has a set of native terms for
> wheeled transport -- 'wheel', 'axle', 'convey', etc.  Wheeled transport
> first appears in the archeological record c. 5300 bp, and the realia
> probably preceded the first archeological evidence by a few centuries.
> David Anthony has made the archeology-linguistics connection in detail
> (e.g. in *Antiquity* in 1995).
>

     This is the best piece of evidence you have, but still two points
here. First, the lack of archeological record does not
mean that wheeled transport *did not exist* before 5,300 bp. It might so
happen that earlier samples have not been yet discovered or that they
perished, as I believe, Miguel pointed out today. Second, what is your
basis for conclusion that "the realia probably preceded the first
archeological evidence by a few ceturies"? Guess-work?

> (3)  Closeness of earliest attested forms.  Vedic Sanskrit, Mycenaean
> Greek, and oldest Hittite give us a picture of the IE family something like
> 3000 years ago.  There is an obvious close family resemblance but no mutual
> intelligibility to speak of (I mean knowing one of these doesn't enable
> even a linguist to read another of them), so the IE family at ca. 3000 bp
> must have been a bit deeper than modern Romance or Slavic.
>

      This argument again stands on the same unproven hypothesis that
underlies glottochronology: all languages change at the same speed,
therefore, if old IE languages are as similar as modern Romance or Slavic,
(let me inter alia, disagree with that: being a native speaker of Russian
and a linguist I can read without any significant difficulty any Slavic
language except Czech), therefore they must be by default as old as
Romance or Slavic.

    I think that is as dangerous to estimate the age of families on the
basis of similarity as to establish genetic links on the same basis. I
believe that a couple of counterexamples will suffice.

    If I remember correctly, Arapaho is frequently cited as a language
that *looks very unlike* the rest of Algoquian. And yet, this is not a
basis for claiming it as a separate branch. There are also a number of
languages in Melanesia that underwent some very drastic changes making
them looking unlike their closest relatives, and yet, these changes appear
to be quite recent and have nothing to do with the chronology. On the
other hand, Evenki and Even, superficially look very similar, but there
are good grounds to believe that they belong to different subbranches
within Tungusic. Etc., etc.

> (4)  Absolute and relative chronology of branches.  Proto-Iranian (or
> pre-Proto-Iranian but probably not Proto-Indo-Iranian) contributes loans to
> Proto-Finno-Ugric, and a good archeological candidate in eastern Kazakhstan
> dates to about 2000 bp.  This is the incipient breakup of a major initial
> branch (Indo-Iranian) of PIE.
>

      I find it very unlikely that PFU was ever spoken in Eastern
Kazakhstan. PFU linguopalenthology obviously indicates a forest zone
with trees not found in the Eastern Kazakhstan. The contact should have
taken place more to the north.
But, anyway, I fail to see what relevance it might have to dating IE.

> All this is off the top of my head (these and other references can be found
> in my paper 'Modeling ancient population structures and movement in
> linguistics', Annual Rev. of Anthropology 26 (1997)), but the point is that
> several very different lines of inquiry converge on very similar dates: the
> PIE breakup took place around 5500 bp.
>
> Johanna Nichols
>
     All this is no more than oblique evidence that *may* or *may not*
have any relevance, and some of it obviously cannot work. There is no
*objective* way to assign dates to
protolanguages based on the direct language evidence alone. But, of
course, the absence of it today, does not mean that we won't figure it out
in the future (:-).

Alexander Vovin
Associate Professor of Japanese
Department of East Asian Languages and Literatures
382 Moore Hall
1890 East-West Road
University of Hawaii at Manoa
Honolulu, HI 96822
vovin at hawaii.edu
fax (808)956-9515 (o.)
t.(808)956-6881 (o.)