10rary Linguistics AUTOLEXICAL SYNTAX A Theory of Parallel Grammatical Representations Jerrold M. Sadock The University of Chicago Press Chicago & London4 Incorporation 4.1 Historical Background For purposes of the present chapter, I will use the term "incorporation" to refer to phenomena in natural languages (other than cliticization as described in chapter 3), where a proper subpart of a word can be shown to have the function of a formative, i.e., an atomic element, in the syntax. The details of the morphological joining are immaterial to my definition. The incorporated item may be a stem, either compounded with another stem or used as a base for affixation. It may be an affix or even a symbolic process, and may, in either case, be classified as derivational morphology or as inflectional morphology. In chapter 6, I will broaden the notion of incorporation to include examples of discrepancies between semantic organization and either morphological or syntactic structure, but for now my attention will be restricted to cases where it is clear that part of a morphological word is to be represented as a formative in syntax. In the absence of positive syntactic arguments to the contrary, I will assume that pieces of words are not to count as independent elements of the syntax. By itself, no purely semantic principle, such as Baker's 1988 Unifor-mity of Theta Assignment Hypothesis, will be taken as requiring an analysis in terms of noncoincident morphosyntactic bracketing, and thus the range of phenomena that will be discussed in this chapter will be considerably nar-rower than that of those phenomena labeled incorporation in Baker 1988. By far the most discussed putative instance of nonclitic interpenetration of morphology and syntax is noun incorporation, so called because a noun stem representing a syntactic argument of a verb (typically the object) is found as a part of the morphological verb. This phenomenon is frequently encountered in North American languages, -but it is not found either universally or exclu-sively there. Noun incorporation was introduced into the modern literature as evidence against the strict lexicalist hypothesis in work by Rischel (1971, 1972) and myself (Sadock 1980). In those languages where it has a clearly syntactic face, noun incorporation offers excellent evidence against the hierar-chical model of the relation between morphology and syntax, and it was pri-marily as an attempt to handle this striking phenomenon that I originally suggested the model of autolexical syntax. Historical Background 79 It has recently been reasserted, however, that noun incorporation is, after all, a strictly morphological process with no important syntactic ramifications (Mithun 1984; Hopper and Thompson 1984; Di Sciullo and Williams 1987; and most recently Rosen 1989), thus resurrecting a debate that took place early in this century. The term "incorporation" was first employed by Hum-boldt ([1836] 1988) who applied it to Nahuatl examples like ni-naca-qua '1-meat-eat'. Humboldt's argument was the distinctly nonrelativistic one that languages like Nahuatl confuse syntax and morphology and are thus less per-fect than those like Classical Greek or Sanscrit that clearly separate the two. [T]he METHOD OF INFLECTION in all its completeness, ... alone im-parts true inner fixity to the word for both mind and ear, and likewise separates with certainty the parts of the sentence, in keeping with the necessary ordering of thought. ... Compared with the INCOR-PORATIVE PROCEDURE, and that of loose ADDITION without true word unity, the METHOD OF INFLECTION appears as a principle of genius, born of a true intuition of language. (Humboldt [1836] 1988, 145) In apparent reaction to the by-then unpopular glottocentricity of the Humboldtian view, Kroeber (1911) attempted to ban the notion on general principles: In short, the term "incorporation" is a delusion, whether applied to pronoun or to noun. It must be relegated to the same category as other antiquated catch-words such as "agglutination," which have originated in the assumption that the languages of so-called un-civilized people must contain certain features of a kind totally dif-ferent from those characteristic of Europeans-and incidentally features of an inferior order,-and which have found their chief vogue and employment not among serious painstaking students of language but among doctrinaires, compilers, and those false popu-larizers who think to diffuse knowledge by giving a phrase instead of an idea. (Pp. 582-83) Kroeber (1909) offered the following definition of noun incorporation, and then proceeded to argue that the phenomenon did not (indeed, could not) exist: Noun incorporation is the combination into one word of the noun ob-ject of the verb and the verb functioning as the predicate of the sen-tence. It is essential that the resultant of incorporation is a single word, else the process is without limit and all syntactical relation may be construed as incorporation. (P. 37) Note that Kroeber's straw-man definition is essentially antilexicalist; it as-sumes (as I do) that the phenomenon of incorporation is definitionally one that80 Incorporation straddles the border between syntax and morphology. Soon after Kroeber pub-lished the brief paper from which the above passage is quoted, Sapir (1911) criticized him precisely because of the transmodular nature of his definition. Displaying a strikingly modern prejudice, Sapir argued that general consid-erations required that the definition of noun incorporation be either morpho-logical, or syntactosemantic, but not both at the same time. Examining this definition, we find that two things are required-a noun must combine with the verb-predicate into a word-unit, and the noun so combined must function as the object of the verb. The first requirement is morphologic in character, the second purely syntactic; in other words, the first calls for a certain type of word formation, while the second demands that a particular logical relation subsist be-tween the two independent elements that enter into this word forma-tion. Without denying the abstract right to set up such a definition, it - would seem that the combining of a morphologic requirement with an independent syntactic one yields, on general