On concord and projection

ON CONCORD AND PROJECTION Giuliana Giusti Abstract: In the spirit of the copy theory of movement, I propose that Merge applies only to satisfy Selection or Modification. This proposal differentiates between Agreement (instantiated by an EPP-feature targeting the upper argument selected by the lexical head, in the usual way) and Concord, occurring between any head and its Specifier with no additional trigger. It also derives head-movement by proposing that the functional skeleton of a lexical head is one and the same bundle of features, projected according to a paradigm associated with the head. As a consequence, parametric variation can be reduced to properties of inflectional paradigms. Keywords: agreement, concord, merge, nominal expression, head-movement 1. Introduction This paper† sketches the general lines of an ambitious project (outlined in Giusti 2007, 2008, 2009) with the aim to reconcile general issues on syntactic structures, such as the motivation for the EPP-feature, the question whether feature-sharing is the result of a single operation, the (dis)advantages of assuming or eliminating head-movement, the status of the hierarchies investigated by the cartographic approaches. All these issues have been on the agenda in recent years and have until now received independent treatments.1 The ambition here is to give a unified answer to (some of) them, grounded on a principled theory of Merge. The general proposal is that Merge takes place to instantiate two basic relations: Selection and Modification. The former is triggered by selectional requirements that arise between a (lexical or functional) feature (which is realized in a head) and a maximal projection; the latter is the optional relation between a maximal constituent and a head. A crucial part of this proposal is to conceive the notion of head as a bundle of features (cf. Matushansky 2006) whose hierarchy is given by UG and whose * Università Ca’ Foscari di Venezia, Ca’ Bembo 1075 - 30123 Venezia, A first presentation of this research project was delivered at the the Bucharest Conference in 2007, and appeared in the University of Bucharest Working Papers in Linguistics. A second draft has been circulated as University of Venice Working Papers in Linguistics 2009. I presented parts of this project at the University of Leiden, October 2009, University of Amsterdam, the syntax circle April 2010, and at the Bucharest conference 2010. I thank all the many who commented on this, especially Lisa Cheng and Enoch Aboh, but my very special thanks go to Hans den Besten with whom I had a long discussion at dinner around a year ago and is no more here with us. 1 The work-in-progress status of this paper does not allow me to do justice to the literature, which obviously deserves the discussion that is missing here. † G i ul i a na Gi u st i 104 realization must be acquired through exposition in the frame of a paradigm. In this view the notion of paradigm becomes crucial. In one and the same language, the form of paradigms certainly shares many properties, but at the same time can leave the door open for idiosyncratic properties of individual (class of) items. This new notion of paradigm is taken to include not only the traditional forms of single words but also and crucially the free morphemes that realize their extended projection (in the sense of Grimshaw 1991), which in this proposal are taken to be part of the bundle and the (partial) realization of a scattered head (in the sense of Giorgi and Pianesi 1997). Agreement arises when an EPP feature is part of the bundle and is required when the proper interpretation of the extended constituent (clause or nominal expression) requires an intersection between two different reference values. In clauses this happens between the temporal reference of the situation and the individual reference of the subject, while in the nominal expression this happens when the individual reference of the head noun is restricted by the individual reference of an embedded nominal expression. Only the former is obligatory because it intersects reference values of different types (time and person) to produce a third type, namely propositional value. Concord arises every time an element is (re)-merged (for whatever reasons) as the modifier of a head. So abstract Concord always occurs between a modifier and its head, a remerged element can obviously concord with different heads, while a complex head which is remerged at different points of the extended projection concords with all its modifiers. The result is that Agreement is a unique relation while Concord is recursive. Furthermore, the same pair of elements may Agree and Concord at the same time. This paper is structured as follows: In section 2, I sketch the background proposals which have already been put forth in previous papers and talks and state the specific topic of this paper which regards the notions of projection and concord in the nominal expression. In section 3, I support my idea that Concord is not Agree concentrating on adjectival concord in nominal expressions in Italian and German, showing that some apparent adjectival concord is in fact the result of the overt realization of the scattered nominal head. In section 4, I draw some general conclusions highlighting strong points and hinting at open questions. 2. Background proposals The basic hypothesis from which all other properties follow is the quite reasonable one formulated in (1): (1) Merge operates to instantiate two relations: Selection and Modification. On concord and projection 105 Selection merges a lexical head (X or K) specified in the lexicon for selectional features with a fully fledged constituent, or “perfect projection”, that can satisfy such selectional features (KP and WP respectively). Modification merges a fully fledged constituent (GP, HP, LP) as a modifier of a lexical head K with a projection of K. The relations are instantiated one at a time obeying the relevant hierarchies. This has the consequence that the head needs to remerge each time a new argument or specifier is picked from the numeration. Let us take the structure in (2) as an abstract example. In (2) the head K is associated to two interpretable features [+z], [+y], and one uninterpretable features [−c].2 This means that the paradigm of K (take it to be a verb or a noun) is associated with interpretable features (e.g. Tense and Aspect if a V, gender and Number if a N) and an uninterpretable feature. If it is a V, in order to project to form a clause with propositional value, it must intersect with a person feature, so the well known EPP feature is merged as part of the paradigm. If it is a noun, in order to receive a theta-role it must be marked for Case). K has a theta-role (th) to assign and selects a maximal projection WP whose features are not given for ease of presentation. K remerges as many times (KP1, KP2, …) as necessary. Remerge may be necessary to realize all more than o (...truncated)