Noun incorporation in Greenlandic: A case of syntactic word formation

NOUN INCORPORATION IN GREENLANDIC: A CASE OF SYNTACTIC WORD FORMATION Jerrold M. Sadock University of Chicago In Greenlandic Eskimo, a polysynthetic language, the word-building apparatus performs much of the work that is accomplished by the syntax of more familiar languages. In particular, numerous processes in this language create verbs from nouns. Evidence of a quite unusual sort shows that these noun-incorporation processes must follow certain ordinary syntactic rules, such as case assignment and modifier-noun agreement. The language thus falsifies pronouncements concerning the independence of syntax and word formation based on data from languages that are typologically very different from Greenlandic* Certain aspects of the grammar of Greenlandic Eskimo shed light on the question of whether, and to what degree, universal grammar must allow for the possibility of syntactic processing before the words of surface structure exist as such. Phrased differently, the question is whether any word-formation rules are interspersed among the ordinary rules of the syntactic component. The extreme relevance of Greenlandic to this issue was recognized by Rischel 1971, 1972, but little notice appears to have been taken of his important work. Thus I wish to present further data and arguments in support of what is essentially the conclusion drawn by Rischel. 1. Theory. The dispute between the advocates of prelexical syntax, as it is often called, and those who favor presyntactic lexical insertion has been somewhat obscured because of the failure on the part of both sides to make their claims precise. It is not immediately obvious to what degree the formatives of deep * I wish to thank Wolfgang Dressier, Eric Hamp, Alexis Manaster-Ramer, James McCawley, Pamela Munro, and Geoff Pullum for helpful advice and criticism of earlier drafts of this paper. I am especially grateful to Anthony Woodbury for his perceptive and painstaking critiques, and to Jörgen Rischel, whose insights form the basis of this work and who prevented me from committing an unpardonable scholarly sin. Naturally, I remain responsible for any mistakes still present. Most of the data in this paper have been taken from Bugge et al. 1960, Kleinschmidt 1851, Rasmussen 1971, Rischel 1971, 1972, and Schultz-Lorentzen 1927. Some pedestrian and presumably non-crucial examples were made up in order to simplify the Greenlandic vocabulary used. 300 NOUN INCORPORATION IN GREENLANDIC301 structure are required to resemble the words ofsurface structure in a theory without prelexical syntax, nor is it obvious how much divergence is allowed in a theory with such syntax. The most severe restriction that could be placed on universal grammar would be the requirement that all word-sized constituents of surface syntax be represented in deep structure by constituents containing all and only the morphemes of the surface word—in their surface order, and in a form suitable for processing by purely phonological rules. In the weakest possible model, there would be no restrictions whatever as to the point in syntactic derivations where the integrity of surface words would have to be respected, and no limits at all as to the degree of difference between the form of surface words and their deep-structure sources. Neither of these extreme alternatives is particularly attractive. The problem with the very strict construal of the ban on prelexical syntax is that it would more or less spell doom for syntax itself; e.g., it would make Affix Hopping and Z)o-Support rules impossible, and would remove rules of concord from the syntactic component. Since reflexive pronouns in English are clearly words, they would have to appear as such in deep structure. With no agreement rules or Reflexivization, numerous other rules of classical syntax (such as Passive) would also go by the board. And since Passive is the cornerstone of classical syntax, all the rules that depend on it for their justification would also be eliminated. In other words, there would be virtually no syntax in a theory that totally outlawed prelexical syntax. In fact, virtually syntax-free theories of grammar have become moderately popular lately; but they are motivated in part by the fact that they are consistent with a very strict adherence to the lexicalist doctrine—a doctrine which the data from Greenlandic seem to call into...