Part I
Configurationality and the Pronominal Argument Hypothesis
The papers in this section draw together a rich body of work with data from Warlpiri, Navajo, Apache, Lummi, Yaqui, and others, arguing that languages that are non-configurational (in the sense described by Hale (1983)) are best explained by the view that they, to one degree or another, parametrically lack full DP arguments. Argument positions in these languages are occupied by pronouns, pronominal clitics, or rich agreement that indexes the pronominal arguments. Full DPs are adjuncts and donāt participate in traditional argument relations.
1 Empty Categories, Case, and Configurationalityā *
Eloise Jelinek
Editor's Introduction
Obvious challenges to constituency-based theories of syntax are found in languages, such as Warlpiri, that allow significant flexibility in word order. Warlpiri not only allows free ordering of constituents but also allows constituents to be non-contiguous. Hale (1983) proposed a configurationality parameter, whereby the languages of the world are divided into those that express grammatical and semantic relations via a constituent structure and those that do so through a morphological (e.g., case) mapping.
In this paper, Jelinek argues against Haleās configurationality parameter and introduces instead the influential and important Pronominal Argument Hypothesis (PAH). Using data from Warlpiri, she hypothesizes that non-configurational languages differ from configurational ones because DPs may not serve as arguments. The arguments in pronominal argument languages are always pronouns. These pronouns can be nullāindexed only by verbal agreement. Any DPs in the sentence function as adjuncts. This cluster of claims explains the range of properties exhibited by so-called non-configurational languages. The free word order follows from the fact that adjunct categories can attach to the syntactic structure at different levels and in different orders. The DPs thus just provide āadditionalā information about the referents in the sentence. A parallel in a configurational language might be left dislocation in English (The guy with the hat, he left earlier), where the pronoun serves as the argument of the verb, and the DP serves primarily as additional appositive information. Jelinek shows that this opposition between DPs and pronouns in terms of argument status explains a range of surprising binding and case-marking facts. In Warlpiri, pronouns exhibit properties we might construe as configurational. For example, they are subject to Principle B. But DPs behave in ways that make them look much more like adjuncts. For example, they appear to be exempt from Principle C.
The PAH has been a highly influential theory. It has been applied to a wide variety of languages. Its central insights underlie Bakerās (1996) seminal book The Polysynthesis Parameter. It has also been highly controversial. For example, within the literature on Salish and other Native American languages, the PAH continues to be a central bone of contention (see, for example, the critical review in Davis and Matthewson (2009)).
0. Introduction
Ken Haleās work on Australian and Native American languages has served to extend the data base of mainstream theoretical linguistics, and has made it necessary for a theory concerned with language universals to confront data from these typologically interesting languages.1 In a series of papers (1980, 1981, 1982, 1983) Hale has drawn attention to the problem of nonconfigurationality in Warlpiri; in the latest of these, āWarlpiri and the Grammar of Non-configurational Languagesā, his purpose is to define a configurationality parameter from which the cluster of properties seen in nonconfigurational languages would follow. I take issue here with Hale on the source of non-configurationality, and propose a different typological parameter, based on a reanalysis of Warlpiri data given in Haleās publications, and some observations on other non-configurational languages.2 An interesting result of this analysis is an explanation of the āergative splitsā frequently seen in nonconfigurational languages.
The properties common to non-configurational languages that Hale seeks to account for include the following: (1) āfreeā word order, (2) syntactically discontinuous expressions, and (3) ānull anaphoraā. In the following Warlpiri sentence, any word order is possible, with the provision that the AUX -clitic sequence appears in the second position.3
1) | Ngarrka-ngku | ka | wawirri | panti-rni. |
| man-ERG | AUX | kangaroo | spear-NONPAST |
| The man is spearing the kangaroo. (Hale 1983: 6) |
Thus, āfreeā word order. Furthermore, non-adjacent nominals may correspond to a single verbal argument, resulting in discontinuous expressions:
2) | Wawirri | kapi-rna | panti-rni | yalumpu. |
| kangaroo | AUX | spear-NONPAST | that |
| I will spear that kangaroo. (Hale 1983: 6) |
(This example is as given by Hale; the clitic - rna marks first-person singular subject.) Wawirri and yalumpu in (2) comprise a discontinuous expression. In (3) below, these nominals appear as a single (continuous) constituent, as can be seen by the fact that they precede AUX; only one word or a single constituent may occur before AUX.
3) | Wawirri | yalumpu | kapi-rna | panti-rni. |
| kangaroo | that | AUX | spear-NONPAST (Hale 1983: 6) |
By ānull anaphoraā Hale refers to āthe situation in which an argument (e.g., subject, object) is not represented by an overt nominal expression in phrase structureā (Hale 1983: 7). This is exemplified in (4) below: English exhibits none of these traits: Word order marks grammatical relations; constituents may not be discontinuous; and nominals are not optional. The primary goal of this paper will be to account for the fact that nominals are frequently āabsentā in Warlpiri sentences; once this aspect of Warlpiri syntax is clarified, we will also have an explanation for free word order and the apparent discontinuous expressions. Within the Government and Binding (GB) framework (Chomsky 1981, 1982) the Projection Principle precludes āmissingā nominal arguments:
Within the GB framework, there are no āmissingā nominals in English sentences; there are empty categories (ECs) that bear the relevant Īø-roles. The point is that nominals represented by ECs are recoverable, as in the case of PRO in the following example:
- 6) The man wants [[PRO] to spear the kangaroo].
The anaphoric relation between the subjects of the two clauses makes the reference of PRO in the embedded clause explicit.
Chomsky (1982: 78ā88) identifies pro as the āmissingā subject in āpro dropā languages; pro is free in its governing category and is a non-anaphoric pronominal, with independent (deictic) reference. Haleās claim is that neither PRO nor pro need be postulated in the analysis of Warlpiri main clauses; nominals are simply optional. Non-configurationality finds its origins in the nature of the relationship between phrase structure (PS) and lexical structure (LS), that is, in differences in the way the Projection Principle holds in the two language types.
By lexical structure, Hale refers to predicates and their argument arrays. These arrays correspond to variables specifie...