April 18, 2004

Professional Foolishness

A favorite activity here on Language Log is pointing out the deficiencies of what various non-specialists say and think about language and linguistics: journalists (here, here, and here), language pundits (here, here and here), geneticists, political scientists, actors, and sea captains. But as Mark Liberman pointed out in his post on Professionalism, foolishness and irresponsibility are by no means excluded by professional qualifications. A case in point recently came to my attention when I read Words and Rules by psycholinguist and best-selling popularizer Steven Pinker.

In Chapter 8 The Horrors of the German Language Pinker includes (p. 212) a family tree entitled The Ancestry of Modern English. English is shown within a partial family tree of Indo-European. Indo-European is shown as a daughter of Eurasiatic and a sister of Uralic and Altaic. No other subgroups of Eurasiatic are shown. Eurasiatic in turn is shown as a subgroup of Nostratic, with Dravidian and Afro-Asiatic as the other subgroups. Nostratic in turn is shown as a sister of Sino-Tibetan and New Guinea, with the parent labelled with a question mark.

This family tree doesn't look much like what will be found in standard references, such as the Ethnologue. It is flawed in several ways. To begin with, it is actually a fragment of a family tree. Several subgroups of Indo-European are omitted, as are several subgroups of Eurasiatic and one subgroup of Nostratic. A careful reading of the subsequent text will make the astute reader realize that the tree is incomplete, but this is never made explicit. The family tree has no legend or explanatory notes.

Secondly, it appears to indicate that Nostratic, Sino-Tibetan, and New Guinea form a language family. If so, this is misleading, because there is not the slightest reason to believe this to be the case. To my knowledge no such proposal appears in the linguistic literature. The question mark that labels the parent node may be intended to indicate that it is not known whether these three groups are related, but this is not clear. Such unclear relationships are usually depicted with broken lines. The question mark is more likely to be interpreted as indicating that the group suggested has no known name.

By far the most serious problem, though, is the fact that many of the relationships depicted are unproven. There is no language family known as New Guinea. Hundreds of languages are spoken in New Guinea. Some of them are clearly members of the Austronesian language family, the family that also includes such languages as Malay, Hawaiian, and Maori. The rest are grouped under the cover-term Papuan, but they have not been shown to be genetically related. Even the "lumpers" distinguish half-a-dozen language families in New Guinea; historical linguists who insist that relationships be established by the comparative method divide the non-Austronesian languages of New Guinea into 68 language families. There is a proposal by Joseph Greenberg for a language family known as Indo-Pacific, consisting of the Papuan languages, the languages of Tasmania, and the languages of the Andaman Islands, but it has never been supported by adequate evidence and is not taken seriously.

Eurasiatic is another proposal of Greenberg's. It consists of: Indo-European, Uralic-Yukaghir, Altaic, Japanese-Korean-Ainu, Gilyak, Chukotian, and Eskimo-Aleut. Like Greenberg's other proposals, that these languages are related has not been demonstrated by adequate evidence and is not generally accepted. Nostratic, as originally proposed by Vladislav Illich-Svitych, consists of: Indo-European, Uralic, Altaic, Afro-Asiatic, Kartvelian (also known as South Caucasian), and Dravidian. Eurasiatic and Nostratic in their original forms are competing proposals. The family tree shown by Pinker in which Eurasiatic is a subgroup of Nostratic is a compromise developed by some Nostraticists in recent years. Nostratic too is not generally accepted.

The highest level relationships shown, among Nostratic, Sino-Tibetan, and "New Guinea", are pure fantasy. There is no evidence for such relationships.

To be fair, in the text (p. 234) Pinker does mention that most linguists think that the traces of common ancestry are lost in the mists of time beyond groups at the level of Indo-European and Uralic, but he fails to give a real idea of how far out these proposals are. He characterizes the proponents of Nostratic as "a few dauntless linguists" (p. 236), as if the reason that most historical linguists consider such hypotheses to be unproven is that they are insufficiently bold.

I've outlined the problems with remote linguistic comparison in a previous post. Eurasiatic, Nostratic, and other such proposals have not been accepted because there is insufficient evidence that the similarities observed are not due to chance and because the possibility that they are due to borrowing rather than common descent has not been ruled out. It isn't a matter of boldness; it's a matter of evidence.

The impression that most readers will be left with is that Eurasiatic, Nostratic, and "New Guinea" are established language familes and that Nostratic, Sino-Tibetan, and "New Guinea" are known to be related. The careful reader will learn that most linguists are skeptical, but Pinker's disclaimer will have little force in the face of his characterization of the long-ranger fringe as "dauntless" coupled with his implicit endorsement.

What makes this especially annoying is that the inclusion of this family tree is completely gratuitous. The chapter is devoted to showing that rules are not restricted to English but are found in a variety of other languages. The approach that Pinker takes is to start with languages closely related to English and show how remoter and remoter languages also have rules. All that really matters is that his examples not be closely related. He could have made the same point just as well without any discussion of remote genetic relationships.

It's hard to say why Pinker chose to include this family tree. By his own statement, he knows that most linguists don't believe it, so it isn't a mistake made out of ignorance. It is highly unlikely that Pinker has any professional views of his own on historical linguistics. Training in psychology does not normally cover historical linguistics, nor can I find any record of his ever having published on or spoken about or otherwise manifested any interest in the subject. He doesn't seem to have any stake in it. Whatever his motivation may have been, the family tree that he presents is speculative and misleading, hardly appropriate for a work intended for the lay reader. He and his publisher should have known better.

Posted by Bill Poser at April 18, 2004 10:54 PM