Age of various language families

[Mikael Parkvall]

----------------------------Original message----------------------------
We're getting pretty far away from my original query now, but this is of 
course also interesting.

Jens Elmegaard Rasmussen wrote:

>There just is no such thing as a rule of language survival. Tribes and
>peoples influence each other by domination and genocide, some disappear by
>famine or floods. The oldest language group of all may have completely
>vanished, the most proliferate may be of quite recent making (as a
>split-off from something which has not remained or cannot be made out to
>be related). The whole expectation the rpompted the question is based on a
>monumental mistake. Sorry, but that's how clear it is to me.

I think there is one important flaw in this reasoning. First of all, let us 
limit ourselves to proto-languages which have descendants today -- 
otherwise, we'd simply get too speculative. Secondly, note that when I 
(somewhat sloppily perhaps) used the word "age", what I really meant is 
"time depth". In other words, the "age" of a given family is defined by the 
first known split-up.

Then let's discuss extinction. Any language risks disappearing, be it due 
to smallpox, assimilation, flooding or whatever. A "family" consisting of 
only one language would be wiped out if this sole language were to become 
extinct. Meanwhile, another family with hundreds of members could lose some 
of these, and yet continue to exist. In other words, a single natural 
disaster could indeed wipe out all speakers of Basque, but hardly all 
speakers of, say, Austronesian (lest it kills of humanity entirely).

There is no a priori reason why a meteor (or disease or assimilation) would 
target any specific language, nor is there any reason to suppose that 
splitting would be biased. Thus, we have to assume that all languages 
multiply and die regardless of their genetic relationship.

If we were to ask "which groups is more likely to have eight members 
tomorrow -- a) the one which currently has six, or b) the one which 
currently has two?", the answer must be a), since splitting is random, and 
there are three times as many potential mothers in one of the groups. Then, 
let's look at the same scenario the other way: "which group is most likely 
to have been one single language yesterday -- a) the one which currently 
has six members, or b) the one which currently has two?".

Note however, that we're dealing with _likelihood_ here -- reality may of 
course look different. Even unlikely things happen, of course.

In an expanding population, an individual high up in the family tree is 
likely to have more descendants than an individual further down. It is 
highly likely that my great-great-great-great-grandfather has more living 
descendants than you yourself have. It could be argued here that the number 
of languages on earth has remained constant, as opposed to the number of 
people, but if we limit ourselves to proto-languages with surviving 
daughters, the number of languages must certainly have increased over the 
past 10000 or whatever years.

In other words, it is certainly possible, as Jens suggests, that "the most 
proliferate [family] may be of quite recent  making", but, I maintain, it 
is statistically less likely. I suspect that if we looked closer at this, 
the correlation between family time-depth and number of attested daughters 
would be a weak one, but it would be more likely to be positive than 
negative. Also, we might possibly gain some insights by pondering upon the 
cases which are more aberrant than others.

My ability of making myself understood seems to have reached an 
all-time-low these past few days, but hopefully somebody understood 
something of what I was trying to say above.

In any case, while I enjoy discussing this, I would also love receiving the 
additional data I my original query concerned. That way we could at least 
get an empirical reusult (albeit a shaky one) to compare my guess with.


/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