Sleuth who rose up family tree

Joseph Greenberg died in 2001, aged 85, working almost to the last on his final, posthumously published book. His prodigious capacity for languages (first manifested as he taught himself Hebrew and Ancient Greek in high school) continued until nearly the end, though, as William Croft's obituary recounts, Greenberg did notice a certain slowing down in his later years, complaining that "every time I learn the name of a new student, a fact about Nilo-Saharan flies out of my head".

Greenberg not only contributed to 20th-century linguistics, he helped to shape it. His main contributions were typology and language classification. Yet the universal welcome and almost reverence we find for his pioneering research on typology can be contrasted with a heated and somewhat acrimonious debate over his methods and results in classification.

Linguists today would take it virtually for granted that theories of language should be based on cross-linguistic data. Greenberg's initial surveys of 30 or so languages would, from today's perspective of computer databases and interactive maps, seem laughably inadequate. But in the 1960s seeking patterns in comparative data revolutionised the field.

Croft, once Greenberg's student and now a well-known typologist and theorist of grammar, might have been expected to produce a collection of Greenberg's widely applauded writings on typology and language universals, but has instead taken the brave and extremely valuable step of editing some of the highly controversial writings on classification and language families.

The collection is arranged into four central sections: part one on "Classification, grouping and subgrouping"; part two on "Classification, sound correspondences and reconstruction"; part three on "Indo-Pacific, Amerind, Eurasiatic"; and part four on "Genetic linguistics and human history". These are flanked by an editorial introduction and two comprehensive bibliographies, one on Greenberg's contributions to genetic linguistics, and the other listing reviews, reactions and discussions of those contributions. These make this book a particularly valuable advanced teaching and research resource.

What, though, is genetic linguistics, and why is Greenberg's work in this area so controversial?

A family such as Indo-European is beyond contention: but students of Indo-European are spoiled because we are dealing with extensively studied languages, often with written records stretching back centuries, which we can use to check our educated guesses and test our methods. Indo-European is a large family, but might family groupings elsewhere be even bigger and, indeed, might Indo-European itself fit into larger groups? Are all the languages in the world today ultimately related? What are the implications of such megafamilies for human population history, or for evidence from archaeology, genetics or anthropology?

As these questions get bigger, they become more intuitively appealing but much more controversial. Greenberg's first forays into comparative linguistics involved his reclassification of the African languages from the five Semitic, Hamitic, Sudanic, Bantu and Bushman groups proposed in the 1940s, into 16 and subsequently 12 families. The earlier classification had been based on a mixture of purely superficial typological features that could easily reflect chance or borrowing rather than relatedness, and traits such as race, which are simply not linguistically relevant.

Greenberg accepted only linguistic evidence in arguments for language classification, and prioritised similarities between form and meaning, not general structural facts; he also argued that exactly the same methods could and should be used for both written and unwritten languages. These principles are both sound and generally accepted in historical linguistics today.

Greenberg's next step, however, was to compare more and more languages simultaneously. This method of mass comparison starts to oppose itself to the standard comparative method, which involves the painstaking establishment of relatedness for small, low-level groupings, and repeated comparison and reconstruction for each subsequent grouping on the way up the family tree. Using mass comparison, Greenberg resolved his 12 African families into four, moving on to propose the megafamilies Indo-Pacific, Amerind and Eurasiatic (in which the mighty Indo-European is but a minor branch).

With the publication of Greenberg's book on Amerind in 1987, the controversy really heated up. Greenberg's previous publications, as Croft's collection demonstrates, had gone into very little detail on his methodology, leading many to believe that he had used the comparative method. In his book on the Americas, however, Greenberg discussed mass comparison in sufficient detail for other comparativists to realise that this was a very different proposition. The resulting criticism (on the grounds that the method is too lenient in what it allows to count as a match between languages and cannot distinguish real signals of relatedness from chance or borrowing) can be traced through Greenberg's reviews and rebuttals in parts two and three of the collection.

For many historical linguists, these defences of mass comparison remain unconvincing: Greenberg's undoubted erudition and eye for linguistic data cannot compensate for an unreliable method. Moreover, in his later years, Greenberg's focus shifted even further up the family tree, with Croft's obituary suggesting that he "had no intention of stopping short of a complete genetic classification of the world's languages". Yet today Indo-Pacific is largely ignored, Eurasiatic is the focus of considerable scepticism, and Amerind is rejected by many scholars of New World languages.

Given this background, one can understand why Croft has opted not to include essays that focus specifically on the reconstruction of "Proto-World", or "global etymologies", which most historical linguists would reject out of hand. He resists the temptation to eulogise and to downplay the controversies. Naturally, he does provide an outline of Greenberg's method and reactions to it, and for the most part a helpfully even-handed one, in the introduction. Here, we find a damning but poignant recognition that aspects of mass comparison are difficult to defend.

Observing that Greenberg stressed its affinities with biological cladistics, Croft notes: "However, Rankin... argues that Greenberg's use of similarity of form-meaning pairings to identify groups is more like biological phenetics than cladistics; Greenberg did not reply directly to Rankin."

This collection certainly constitutes a service to the historical linguistics community, and will be - and should be - read by advanced students and researchers interested in language families and in the interaction of historical linguistic evidence and human history. Even when we disagree with methods and classifications it is vital to read the case for them first hand, and that has not been easy for some of Greenberg's work because some important papers are out of print. Croft is to be congratulated in allowing us some additional insight into the development of Greenberg's ideas and his reactions to his critics.

April McMahon is Forbes professor of linguistics and English language, Edinburgh University.