Age of various language families

Wed Oct 2 00:55:34 UTC 2002

----------------------------Original message----------------------------
Dear Mikael (et al.),

I've read the intervening comments but am replying to this original posting
because the names and numbers are below.

I'm sure you're right; of course there's a wide range of variation, but
other things being equal the greater the time depth of the node the more
its surviving daughters.  (NB:  *surviving* daughters.)  The factors that
make things unequal can often be inferred from such things as geography and
whether the splitting protolanguage had a state-like (or higher) vs.
smaller level of organization.  My impression is that the most important
factors (namely geography and political economy or something like that) are
so few and so great in their impact that you can work them into your
equation in a fairly rough form and get decent results.

Some years ago I did a survey of all the language families I could get
information for from the northern hemisphere, and counted the number of
surviving primary branches at great time depths -- about 5000 years or so.
This means that I counted (e.g.) Uralic as having two primary branches,
because it's older than 5000 years old and it has two initial branches;
Austronesian as hvaing 4 (3 Formosan branches plus Malayo-Polynesian, that
being one view of the tree), Indo-European as having 8, and so on; Basque,
Japanese, etc. as having one; and young families such as Muskogean,
Chumashan, etc. as having one (since the lack of any demonstrable kin means
that it's the sole survivor of its line back to the fade-out point of the
comparative method).  The average number of initial branches per stock was
about 1.5.  Of course the average number of branches at any lower level
would be greater, and the average number of languages even greater, because
in any count of surviving branches (or surviving languages) you're looking
only at the ones that do survive.  That is, some nodes branch further; none
disappear by definition.  So I believe it's a foregone conclusion that
you'll find your correlation, but it's worth doing because it would be very
good to know, even very approximately, just what the averages are at
different time depths and what kind of curve emerges from comparing
branching rates at different time depths.  I realize you're only looking at
individual languages, but that's an obvious place to start.

To comment on one of the earlier replies:  Yes, there's discrepancy from
linguist to linguist and family to family in what's called a language.  I
think the big problems come up with national languages, though, and it's
possible to find out approximately how many mutually unintelligible
varieties of, say, Italian or German or Arabic there are.  Or get minimum
and maximum figures, or something.  Again, you can get a big improvement
and results good enough for comparison.

My main reservations are:

1.  Language family ages aren't always very good, even if you do an
in-depth survey.  It may be safer to lump them into broad categories (e.g.
Romance-like, IE-like, etc.).

2.  Not everything listed in Ethnologue is a demonstrated genealogical
grouping.  Of those you list, beware of Na-Dene if it includes Haida, but
just Eyak-Athabaskan-Tlingit is a family with some work done on its age.
Nilo-Saharan is almost certainly not one family, and definitely not a
demonstrated one.  Niger-Congo too is quite likely to contain more than one
family.
        Suggestion:  use only families that are both demonstrated (or
evidently demonstrable) and reconstructed (or evidently reconstructible).
This definition  has more slack in it than is ideal, but it's much better
than using any group listed in Ethnologue.  You can find a list of the
families judged (by myself and Balthasar Bickel, after much consultation
with family experts and reading of literature) to be stocks (=demonstrable
and reconstructible) at

        www.uni-leipzig.de/~autotyp
        (go to Downloads and on-line access, the On-line tools)

Take nothing higher than the stocks and you'll probably be safe.

3.  Small quibble:  IE and Austronesian are two of the best-dated ancient
family breakups we have, and the figures you give aren't quite right:  use
5500 for IE and 6000 for Austronesian.

My paper is in Language 66:3 (1990).  For IE see e.g. David Anthony in
Antiquity 69 (1995) or J. P. Mallory, In Search of the Indo-Europeans
(1989); for Austronesian, various things by Peter Bellwood (archaeologist)
or Robert Blust, e.g. Blust in J. of World Prehistory 9:4 (1995).  For
Uto-Aztecan see now also Jane Hill in American Anthropologist 10:4 (2001).

Johanna Nichols

>----------------------------Original message----------------------------
>Dear fellow histlingers,
>
>As a result of reading some stuff by Bob "punctuated equilibrium" Dixon
>recently, I have been thinking about the relation between the age of a
>family and the number of languages it has split up into. Is there such a
>correlation in the first place, and if so, what does it look like?
>
>So, I just tried to plot these values in a diagram, using the Ethnologue's
>numbers for "number of languages in the family" and suggested family ages
>that I found here and there in various books. I feel, however, that I'd
>like to expand the number of families, so I'm wondering if some of you
>could supply me with more such data.
>
>Below are what I have found this far:
>(please note that I am well aware that some families, like Na-Dene, are
>unlikely -- for the time being, I just noted everything I came across)
>
>FAMILY   AGE   #LGS   SOURCE
>Algonquian   3000   38   Dixon (1997:2).
>Fennic   3500   29   Anttila (1972:301).
>Mixte-Zoque   3600   16   Suárez (1983:28)
>Mande   4000   68   Dalby (1988:448).
>Mayan   4100   69   Suárez (1983:28)
>Misumalpan   4300   4   Suárez (1983:28)
>Eskimo-Aleut   4600   11   (or more) Krauss (1973a:850).
>Fenno-Ugric   5000   32   Anttila (1972:301).
>Austronesian   5000   1262   Dixon (1997:29).
>Uto-Aztecan   5000   62   Mithun (1999:540)
>Uto-Aztecan   5100   62   Suárez (1983:28)
>Austronesian   6000   1262   Dalby (1998:47).
>Uralic   6000   38   Dixon (1997:2), Anttila (1972:301).
>Indo-European   6000   443   Dixon (1997:2).
>Na-Dene   9000   47   Swadesh in Krauss (1973b:952-3).
>Nilo-Saharan   10000   199   (or more) (Dalby 1988:453).
>Niger-Congo   12000   1489   (not explicit) Dalby (1988:348)
>
>
>Hopefully, some of you will be able to help me flesh this out a bit.
>
>
>/MP
>
>* * * * * * * * * * * * * * * * * * * * *
>Mikael Parkvall
>Institutionen för lingvistik
>Stockholms Universitet
>SE-10691 STOCKHOLM
>(rum 276)
>
>+46 (0)8 16 14 41, +46 (0)8 656 68 24 (hem)
>Fax: +46 (0)8 15 53 89
>
>parkvall at ling.su.se

* * * * * * * * * * * * * * * *
Johanna Nichols
Professor
Department of Slavic Languages #2979
University of California, Berkeley
Berkeley, CA 94720-2979, USA

Phone:  (1) (510) 642-1097 (direct)
        (1) (510) 642-2979 (messages)
Fax:  (1) (510) 642-6220 (departmental)
http://socrates.berkeley.edu/~jbn
http://ingush.berkeley.edu:7012/
http://socrates.berkeley.edu/~chechen
http://socrates.berkeley.edu/~autotyp