June 10, 2004


The claim that John McWhorter referred to, that Kusunda, a poorly known language of Nepal, is related to "Indo-Pacific", has been making the rounds of the historical linguistic grapevine for a while now. I haven't yet had time to study carefully the paper that has now appeared in the Proceedings of the National Academy of Science, but a few comments are possible off the cuff.

First, there is no such language family as "Indo-Pacific". "Indo-Pacific" is the name that Joseph Greenberg gave to a putative language family that included all of the non-Austronesian languages of New Guinea, the languages of the Andaman Islands, and the languages of Tasmania but not the other languages of Australia. As usual with Greenberg's work, the proposal was supported by virtually no evidence of the sort considered probative by historical linguists, and the proposal has not been accepted by most historical linguists.

Here is the evaluation given by George van Driem in volume I of his 2001 book Languages of the Himalayas:

Racial notions have continued to be uncritically applied to language groupings. As late as 1971, Joseph Greenberg resurrected the old idea that "the bulk of non-Austronesian languages of Oceania from the Andaman Islands on the west of the Bay of Bengal to Tasmania in the Southeast form a single group of genetically related languages for which the name Indo-Pacific is proposed." This hypothesis is identical to Finck's 1909 family of "Sprachen der ozeanischen Neger", a group for which indeed the name "Indo-Pacific" had already been in use, with its roots in the "Pan-Negrito Theory" of physical anthropologists (cf. Skeat and Blagden 1906: 25-28). Appropriately, Roger Blench has described the Indo-Pacific hypothesis as "essentially a crinkly hair hypothesis". (pp. 139-140)
The linguistic evidence which Greenberg adduced for Indo-Pacific is unconvincing, and lexical look-alikes and superficial typological similarities in languages cannot convincingly demonstrate a theory of linguistic relationships conceived solely on the basis of the physical attributes of the speakers. (p. 141)

The other important point that can be made is that it is possible for pronouns to be borrowed. Here are a few examples:

  • The entire pronominal system of Pirahã, a language of the Brazilian Amazon, was borrowed from the Lingoa Geral, the trade language once widely used in the area.
  • English they and them are loans from Scandinavian.
  • Young Thai speakers currently use /mi/ and /ju/, borrowed from English me and you, much of the time. This provides an escape from the very complicated native Thai honorific system. These loans have not completely replaced the traditional pronouns, but they have come into wide use, and this situation provides an example of how and why pronoun borrowing resulting in complete replacement might come about.
  • Japanese has quite a few words roughly equivalent to English I/me. One of them is 僕 [boku], which is used by men in casual circumstances. This is a loan from Chinese. It's original meaning is "manservant", a meaning that it retains in a few fairly obscure compounds in Japanese.
For further information, check out this paper on Pronoun Borrowing [PDF document] by fellow Language Logger Sally Thomason and Dan Everett.

The idea that resemblances in pronouns are good evidence of a genetic relationship because they can't be borrowed has been around for quite a while and continues to be promoted by people like Ruhlen, but it just isn't true. There are some other things to say about resemblances in pronouns, which many linguists now suspect involve sound symbolism, but that's another topic.

Let me close by disagreeing in one detail with John's statement that:

Almost to a man, historical linguists assume that the attempts by Ruhlen and the late Joseph Greenberg proposing that languages of great antiquity retain traces of their origins in single ancestral languages are irresponsible.
It isn't a matter of assuming. The reason that most historical linguists reject the claims of Merritt Ruhlen and his ilk is that when they are investigated they turn out to be badly flawed. The "evidence" presented isn't probative. It consists of lists of words that kind of, sort of, look similar and mean similar things. There's no reason to think that the similarities they present are not due to chance. A large percentage of the data turn out to be wrong. The statements that they make about the history of historical linguistics and the lessons that we should draw from it are false. It's true that at this point when we hear about the latest claim by these people we think "Oh, another crank proposal", but that reaction is based on their lousy track record, and we don't say anything until we've actually looked at it.

Posted by Bill Poser at June 10, 2004 01:12 PM