On the nature of noun incorporation

ON THE NATURE OF NOUN INCORPORATION Marianne Mithun State University of New York, Albany It has been suggested by Sadock 1986 that certain constructions in Greenlandic Eskimo and Southern Tiwa provide evidence against proposals of Mithun 1984 concerning the nature of noun incorporation. These objections are discussed, along with issues concerning the discourse salience, reference, and semantic interpretation of incorporated nouns. Sadock 1986 has devoted considerable effort to demonstrating that denominal verb formation in Greenlandic Eskimo shares none ofthe semantic or pragmatic functions of noun incorporation as generally understood. This is an important conclusion, although not completely astonishing: denominal verb formation is a different formal process. In N[oun] Incorporation], as commonly understood since Sapir 191 1 , a noun stem is compounded with a verb stem to yield a more specific, derived verb stem. The Greenlandic construction is based on a single noun stem with a derivational suffix. It is not entirely clear why one would refer to this as NI, since it is not obvious what such nouns are incorporated into. In incorporating languages, a verb minus its Incorporated] N[oun] is still a well-formed verb; but in Greenlandic, a denominal verb minus its noun stem would be no word at all.1 1 A number of authors used the term 'noun incorporation' in the standard sense before Sapir. However, Sapir was quite explicit in delimiting the scope of the term: 'It is this process of compounding a noun stem with a verb that it is here proposed to call noun incorporation' (1911:257). It is not the case that he simply failed to consider Eskimo, since he described the denominal verb construction at length (254): 'Many American languages form denominative verbs from noun stems by means of various derivative affixes of verbal, generally transitive meaning ... It can hardly be maintained, however , that verbs of this type have had much to do with a belief in the existence of noun incorporation, the process that they illustrate being a familiar one in Indo-Germanic. Eskimo, a language particularly rich in suffixes that verbify nouns, has been termed polysynthetic, but has not been employed by serious students as a source of examples of noun incorporation.' He concluded his article with a caveat (281-2): 'Lest it be thought, however, that noun incorporation is indeed the characteristic ofAmerican languages generally, it is well to point out that it is entirely absent in a large, perhaps the larger, number of them. Such are Athabascan, Salish, Chinookan, Yokuts, Siouan, and Eskimo ; and yet Athabascan and Eskimo might well be considered types of "polysynthetic" languages.' More recently, in two insightful articles, Hagège 1978, 1980 has shown (among other things) that incorporation and the Salish derivational process involving 'lexical suffixes' are formally distinct in a number of ways. Particularly pertinent is the coexistence of both incorporated nouns and lexical suffixes in the Sliammon dialect of Mainland Comox. Hagège points out (1978:67) that the lexical suffixes are the relics of an older system: 'It is clear from the foregoing that the lexical suffixes constitute an archaic system which is characterized by its coherence and its autonomy with regard to the nouns whose meanings are related to theirs. This alone would suffice to show the difference between these suffixes and the incorporated nouns in languages such as Náhuatl, Southern Paiute, Yana etc' 32 ON THE NATURE OF NOUN INCORPORATION33 Southern Tiwa, the other language cited by S, is a very different case: unlike Greenlandic, it does show canonical incorporation. S reports that, according to Allen et al. 1984, incorporation in Southern Tiwa is obligatory; so it cannot be a lexical process. In fact, as Allen et al. make clear, incorporation is a function of such features as animacy, humanness, and number. Apparently S has overlooked their concluding statement (310): 'Furthermore, overlay relations such as "new topic" may sometimes block incorporation where it is otherwise necessary.' It is observations of this kind that I addressed in Mithun 1984. Contrary to S's understanding, the purpose of my 1984 article is in fact not to prove 'the null hypothesis'—'that any individual word-building process does not interact with the syntax', and 'that NI is...