Language

The Amazonian languages covers language families which have a majority of their languages within the Amazon-Orinoco basin. Dixon and Aikhenvald define this region as extending ‘from the north coast of South America, east to the mouth of the Amazon, west to the Andes, and [End Page 594] south to the southernmost headwaters of the Amazon tributaries’ (4). Given that there are some 300 languages from approximately 20 families plus more than a dozen genetic isolates in this region, it is no mean feat to cover them in less than 450 pages. The volume fits nicely within the stated goals of the Cambridge language survey series: It gives a typological overview of languages of the Amazon basin, reflects current scholarship on South American genetic relationships, and is a useful starting point for curious linguists (advanced undergraduate students and beyond) who know little about the South American language picture.

This book has sparked something of a firestorm within certain South American linguistic circles. In my opinion, the controversy centers on about two pages in the ‘Introduction’ (1–21), where D&A present their own brief evaluative commentary on the history of indigenous language work in South America and the state of scholarly cooperation between, and production by, ‘missionary linguists’ and ‘scholars from the local universities’. Amidst such controversy, any review is itself bound to meet with criticism for one or another perceived sin of omission, commission, or interpretation. But aside from noting the existence of this sociopolitical controversy, my intention in this review is to focus on the scholarly substance found in the remaining 444 pages of the work.

There are fifteen chapters (five by one or both editors) plus an orientation section with terminological and typological conventions.¹ One welcome convention throughout the volume is to consistently distinguish transitive (A) and intransitive (S) subjects, as numerous South American languages display complex ‘splits’ involving these syntactic roles. Aside from the ‘Introduction’ and two chapters on language convergence, chapters survey features of one or more families. About half the chapters primarily synthesize information found in greater detail in other published works. Thus, the references provide useful starting points for further study (though chapters are uneven in how well they reference what has been done in the past decade). Additionally, scholars are advised to consult the primary works to assure themselves of the accuracy and interpretation of data; I spot several misleading or incomplete statements about languages with which I am familiar.

The ‘Introduction’ lists fifteen features which D&A assert make the Amazon region, as defined, a linguistic area in the technical sense. D&A suggest these features ‘are shared by all (or most)’ languages in the area, and not by Andean Quechuan and Aymaran families (though D&A note that some features, such as agglutinative word structure, are shared). I find this hypothesis immediately open to question. Some languages within the geographically-defined region lack a number of the proposed features while certain excluded languages, to the north and to the south, have as many of the listed traits as do some within the region. For example, Guaykuruan (Vidal 2001) and Mapudungu to the south, and Chocoan languages to the northwest, fit the proposed criteria as well as do a number of included languages. In particular, Pilagá (Guaykuruan) easily meets ten of the fifteen features—as Yagua, within the region, also does. Epena Pedee (Chocoan) likewise satisfies ten to twelve of the indicated features (Harms 1994).

One feature that should most certainly be eliminated from the list for establishing a language area is D&A’s observation that bound pronominal prefixes ‘typically appear further from the root than prefixes that mark valence-changing derivations’ (9). It is a very strong statistical universal that inflectional (or syntactic) affixes occur farther from the root than do derivational affixes; thus, the existence of this feature in any set of languages is meaningless for languagearea arguments (cf. also Bybee 1985). In contrast, features which D&A...