Age of various language families

[Tore Janson]

----------------------------Original message----------------------------
Mikael Parkvall and Joanna Nichols both think that it would be interesting
to get an idea about the average rate of language splits within a
genetically defined family over a given time period. Several others have
pointed to the formidable problems of method and definition involved. For my
part, I also doubt that there is any way to find a reasonably reliable
procedure to find such a rate, or that the value of this average rate would
give us any meaningful information. In many ways the problem is similar to
the notorious one of finding the (average) rate for language change. We all
know what happened to the assumption by Swadesh that the rate is constant.
But I want to draw attention to another aspect of the question.  Parkvall
and Nichols look at the speech communities at a given time (now, in
practice) and try to count how the languages relate to attested or assumed
proto-languages. They then count the average number of languages coming from
each proto-language. Since all existing languages are assumed (with good
reason) to come from some proto-language, the average, with this method,
cannot go below 1, as Nichols sees. Several of the objections raised have to
do with the fact that languages that have disappeared completely, such as
Etruscan, are not accounted for at all. And that has to be done, at least if
one would like to get any kind of answer to Parkvall's question why there
are not "gazillions" of languages by now.
Therefore it would be better to count the number of languages at some time
in history, and the average number of  "daughters" to these at a later time.
In practice, we cannot do that, but suppose for a moment that we could, and
we will see something interesting.
Let us assume that at time A, there were three languages, called 1, 2, 3. At
a later time B, there may be for example the three languages 1a, 1b, and 1c,
meaning that language 1 has split into three, and 2 and 3 have disappeared.
There may also be the three languages 1a, 2a, and 3a, meaning that each
original language has exactly one daughter. If one counts from time B, as
Parkvall and Nichols, the average number of daughters is 3 in the first
case, and 1 in the second. But if one counts from time A, the average number
of daughters is 1 in both cases.
A moment of thought is enough to see that the later result will be true
regardless of the number of splits, as long as the number of languages is
the same at time A and time B. If there are 5000 languages at time A and at
time B, the average language at time A will have exactly 1 daughter at time
B. The splits that occur will be exactly balanced by the languages that
disappear.
On the other hand, if the number of languages rises from time A to time B,
the average number of daughters will rise too. (All this is true under a
large number of assumptions implicitly made by Parkvall and Nichols, among
others that languages are well-defined entities, that there are language
splits but not language amalgamations or languages without "mothers", and
that each language is spoken by a well-defined speech community of its own.)
If there are 200 languages at time A and 1000 languages at time B, the
average number of daughters will be 5. That is, the average number of
daughters is actually completely determined by the raise or fall in the
number of languages.
Now, a return to reality. The number of languages in the world at any given
time is dependent on the total number of people on earth and the average
number of people in each speech community with a language of its own. We
know, or can guess, something about this. An account may be found in a
recent book of mine: T. Janson (2002) Speak: A short history of languages.
See also, for example, D. Nettle (1999) Linguistic Diversity.
Very shortly, it is probable that for most of human history, up to around
10,000 years ago, the total population was very small, but speech
communities were also very small (perhaps a couple of thousand persons), so
that there may have been as many languages around as there are now for a
very long time. In such a situation, there are no more splits than
disappearances. In the last few thousand years, populations have raised
dramatically, but the size of speech communities seems to have risen even
faster. Thus, the total number of languages has probably gone down for quite
some time, and is certainly going down right now. As for splits, the number
has probably been high in some areas, and has been balanced by the fact that
many languages have disappeared.
 I think this example shows that it is important for historical linguists to
remember that languages are actually spoken by people, and that linguistic
changes do not happen within a theory or a model but have to do directly
with what happens to the language users.

Tore Janson