September 18, 2006

No, Really, Oldest Writing Isn't Oldest Known Language

I've been hearing from Language Log readers who make good points and make me realize that I neglected to dot a few crucial i's and cross a couple of vital t's in my recent post Oldest Writing System vs. Oldest Language.

Here's Rob Perez:

While we can presume that there are Mesoamerican languages older than the oldest known written language, is there actual evidence of such a language such that it could be said to be known? The reporter didn't make the claim that it was the oldest language (according to your quote) but that it was the oldest known language. Seems to me that those could be different.
And then Robin Shannon:

The reporter only claimed that it could yield the key to "the oldest known language in the Americas". Known language rather than actual language. Surely this is a fair enough call since we can not know of any languages without any physical evidence from the language. Of course there were languages before this but they are not known languages.

Fair enough, indeed. But I'd still disagree. That argument goes through only if it's reasonable to link "age of language" with "age of earliest attestation [written record] of language". But that's the whole problem: it isn't. There are lots of known languages in the Americas that have to be older than 3000 years -- admittedly, by the roughest of estimates, but still, these are not wild guesses. The Salishan language family of the Pacific Northwest provides one example: the estimated time depth for the whole 24-language family is about 4000 years, which means that the parent language of the family was last spoken ca. 4000 years ago. The parent language (a.k.a. proto-language) is not attested; few parent languages are. Latin, the parent of the Romance languages, is an exception; Ancient Greek, the parent of Modern Greek, is another. So, probably, is whatever language the newly-discovered Olmec inscriptions are written in -- if it has one or more descendants, it's the parent of the family. It's certain that any language recorded 3000 years ago is not still spoken; languages change much too fast for that to be possible. Another example: the time depth of the Algonquian language family of North America (28 languages) is ca. 2500-3000 years; and it has two distant relatives in California, which pushes the date back at least another 2000 years. So the total time depth is at least 5000 years, which means that the parent language of that family is both known (though not attested) and considerably older than 3000 years. Those are just two of numerous examples.

I guess I should say why I claim that these proto-languages are known: historical linguists have a powerful comparative method that permits, with impressive reliability, the reconstruction of sizable chunks of a proto-language's vocabulary, sound system, and word structure. The method involves systematic comparison of the structures of the daughter languages, which are attested and in most cases still spoken, and it's one of the major success stories in the historical sciences. The point is that the existence of the proto-language of a well-established language family is not a matter of guesswork: it is the only viable hypothesis that can account for the systematic correspondences found throughout the vocabulary and structure of all of the proto-language's daughter languages. [Please note: I am not responsible for the sexist terminology of historical linguistics. Languages have sisters, proto-languages have daughters, and there's also the popular term Mother Tongue; no brothers, no sons, no fathers. Makes sense, I guess, since sister languages have just one parent.]

Posted by Sally Thomason at September 18, 2006 08:31 AM