Agreement and Head Movement: Clitics, Incorporation, and Defective Goals

3. Cliticization

3 Cliticization 3.1 Introduction: Minimal Phases In this chapter, I develop and motivate at length an analysis of Romance clitics. Following an idea that has become fairly standard since Muysken ’s (1982) original proposal, I treat clitics as simultaneously maximal and minimal elements. As such they are able to move both as XPs and as heads (as was first pointed out by Sportiche (1996)). I propose that cliticization is head movement, and develop a general account of why that is both allowed and required. The account involves two central ideas: that of a word as phase, adopted and adapted from Marantz 2001, 2006, 7 (see also Mavrogiorgos 2006), and that of defective goal. This account of the mechanism of cliticization as head movement falls under (12a) of chapter 2. As such, we are dealing with a case of head movement where the A-over-A Condition is not relevant, because clitics are simultaneously minimal and maximal categories. But the interest of clitic movement is that, if clitics can be shown to move as minimal categories, and to form derived structures as such, then we are able to see, independently of the A-over-A Condition, how movement of a minimal category works. In particular, we can see what, if any, special conditions apply. The two central ideas just described are each based on a core intuition. The account of clitics to be proposed is based on the intuition that they show special syntactic behavior because they are defective elements. This idea was influentially pursued by Cardinaletti and Starke (1999), as well as by Déchaı̂ne and Wiltschko (2002). The technical concept of a defective goal, to be developed below, can be seen as a way of articulating this intuition. The other idea, that of minimal phases, is based on the very traditional idea that words are treated as units by the interfaces, and that they are the basic unit of sound-meaning correspondence. As such, it is natural to see them as phases; in fact, as Marantz (2001, 2006) suggests, we may be able to understand the pretheoretical concept of word in terms of the theoretical concept of phase. Here I will follow Marantz ’s leading idea, but execute it in a rather di¤erent way. We can justify treating words, or at least words of certain categories, as phases in terms of bare phrase structure. Bare phrase structure does not make a primitive distinction between minimal and maximal categories, and so, if Xmax is a phase, then, in principle, so is Xmin . One immediate consequence of treating (some) words as phases is that we expect them to be opaque domains for syntactic operations.1 In this way, lexical integrity more generally might follow from the PIC rather than being a separate stipulation.2 Another consequence is that, as phases, words should have a left edge that is transparent to the outside; this is where the link to clitics comes in. The principal reason clitics and cliticization have been of interest for theories of syntax and morphology is that they seem to straddle the boundary between the two components: clitics seem to behave as if they are part of the word formed by their host in that they often form a phonological unit with the host. As Matthews (1991, 217) puts it: ‘‘Clitics are word-like in their grammar, but phonologically must lean for support’’ (also see Julien 2000, 16¤., for a summary of the evidence that clitics have a special phonological status in relation to their host). Further, they seem to move syntactically with the host (see the discussion of French proclitics in subject-clitic inversion in section 1.3) and seem to be unable to separate from their host or appear independently from it. On the other hand, clitics di¤er from bound morphemes in the following respects: (i) they show relative freedom of distribution, in particular in that in many languages, including Romance, one and the same clitic can be either pro- or enclitic, depending on context; (ii) many second-position clitics show indi¤erence to the category of the elements they attach to; they generally (but certainly not always) attach outside clear inflectional and derivational morphology, and so on. (For discussion of these issues from various theoretical standpoints, see, among other publications, Klavans 1985; Van Riemsdijk 1999; Spencer 1991, 350–351; Zwicky 1977; Zwicky and Pullum 1983; and Anderson 2005.) Here I am only concerned with what Zwicky referred to as special...