Extraction from Noun Phrases and the Determiner Phrase Analysis in Slavic: A Comparative Perspective

Gilbert C. Rappaport, University of Texas, Austin

Within the "Principles and Parameters" framework of generative grammar (and retained in the current "Minimalism" instantiation of the theory), it has been proposed that syntactic structure is built around two types of category (and their phrases): lexical categories and functional categories.

Lexical categories assemble lexical items from the dictionary together with their arguments and modifiers; lexical categories encompass such familiar word classes as nouns, verbs, adjectives, adverbs, and prepositions. Functional categories may be instantiated by a limited inventory of words, but may instead by associated with unpronounced formal features. By definition, functional categories provide landing sites for syntactic movement. For example, the overt wh-movement of question words removes a word from its original position in a phrase headed by a lexical item and places it in a phrase projected from the functional category C (for Complementizer) (in some cases, C is pronounced, as in Russian li). It is also assumed that there is covert movement, a mechanism for performing certain morphosyntactic operations such as subject-predicate agreement and case marking. For example, a functional category T (for Tense) is the locus of agreement features (person, number, gender), and the overt or covert movement (depending upon the language) of the subject and verb from the Verb Phrase to the Tense Phrase makes it possible to "check" agreement in a finite clause.

This paper addresses the issue of the functional category of Determiner, which is to the lexical category of nominals what Tense is to verbs. In some languages (e.g., Turkish and Hungarian) there is nominal agreement checked by raising N to D (like V to T in the finite clause). And this category is realized by an article in some languages (e.g., Germanic, Romance). Since neither of these properties holds of most Slavic languages, it has been an issue of some controversy as to whether to assume a Determiner phrase in Slavic. In particular, we defend the existence of this category in Slavic against a reasonable argument that has been used against it.

After a survey of the arguments for (e.g., Engelhardt and Trugmann 1998, Progovac 1998, Rappaport 1998) and against (Corver 1992 and Zlatic 1997) (see also Fowler and Franks 1994), we focus on the issue of extraction from Noun Phrases. Corver noted that Polish and Czech permit the extraction of agreeing modifiers (violations of the "Left Branch Condition") and correlated this with the absence of a Determiner Phrase in these languages, in contrast to, say, Germanic and Romance languages, in which a Determiner Phrase would block such extraction as a case of Subjacency (Stepanovic 1998 has extended this argument to a wider range of modifiers). Pursuing this line, Zlatic has proposed a typological correlation: a language has a Determiner Phrase if and only if it has articles.

We will show that the typical case in Slavic is that non-agreeing modifiers and complements in Slavic contrast with agreeing modifiers in that the former cannot be extracted from a Noun Phrase, but the latter can be; this does not follow from the Determiner Phrase/Subjacency approach at all. Extraction in Bulgarian and Macedonian will be examined, because these languages have articles and only vestigial case, thus resembling English in these typological parameters more than they do the other Slavic languages. It will be seen that the variation in extraction is correlated not with the presence vs. absence of a Determiner Phrase, but rather with the case properties of the language: agreeing modifiers can be extract in a language only if that language has overt case marking. This typological observation will be discussed in the context of the theory assumed.