In lieu of an abstract, here is a brief excerpt of the content:

  • Right Node Raising and Extraction in Tagalog
  • Joseph Sabbagh

1 Introduction

The correct analysis of right-node-raising (RNR) constructions, exemplified by the English example in (1), remains controversial.

  1. 1. Many people supported ___, but many people also denounced, that particular amendment.

Classic analyses of this construction assume a derivation involving across-the-board (ATB) rightward movement (Ross 1967, Postal 1974, Abbott 1976, Grosu 1976). According to this analysis, the shared constituent in (1) (henceforth, the pivot) moves out of each conjunct and right-adjoins to a position external to the coordinate structure.

Since the work of Wexler and Culicover (1980), this analysis has been subject to extensive criticism. In particular, Wexler and Culicover, and many authors working on RNR since then, offer evidence that the "gaps" in RNR constructions like (1) do not behave like ordinary gaps either of rightward movement, in particular, or of extraction (wh-movement, relativization, etc.) more generally (see McCawley 1982, Levine 1985, McCloskey 1986, Wilder 1997, Hartmann 2000). To mention just one argument, rightward movement in English cannot strand prepositions, as the ungrammaticality of (2a) shows (cf. Whose office were you looking for ___ all day?). As (2b) illustrates, on the other hand, preposition stranding is unexceptional with RNR.1

  1. 2.

    1. a. *We've been looking for ___ all day the dean's office.

    2. b. We've been looking for ___, but haven't yet found, the dean's office.

On the basis of observations like this, in-situ analyses of RNR have been proposed. One such approach proposes that the gap(s) in an RNR construction are derived by a (PF) deletion operation that operates backward (Kayne 1994, Wilder 1997, Hartmann 2000, Ha, [End Page 502] to appear). According to this approach, the shared element in (1) is crucially located in situ within the final and nonfinal conjuncts but deleted from the nonfinal conjunct. Another in-situ approach claims that there is just one occurrence of the RNR pivot, which is multiply dominated from a position within each conjunct of the coordinate structure (McCawley 1982, Levine 1985, McCloskey 1986, Phillips 1996, Wilder 1999, Bacharach and Katzir 2006).

I argue in this squib that with respect to at least one language, Tagalog (Western Austronesian), the classic analysis of RNR as ATB rightward movement is correct. The argument is based on the celebrated extraction restriction—iconic of many Western Austronesian languages—in which only subjects, certain oblique arguments, and various types of adverbs may be targets for (Ā-)movement in wh-question formation, relativization, topicalization, and so on. In addition to limiting movement in these environments, this restriction seems to limit the range of possible RNR constructions in Tagalog, a fact that provides compelling evidence for a movement analysis of RNR in this language.

2 The Subject Restriction

Like many Philippine languages, Tagalog observes a "subject-only" restriction on constructions involving (Ā-)movement. For present purposes, it will suffice to define the subject as the argument that controls the particular choice of voice morphology (i.e., agreement) on the verb and that is inflected with the case marker ang (or si, with proper names) (see Keenan 1976 and Schachter 1976 for further discussion). In (3a), for example, the subject is si Juan 'Juan', while in (3b) the subject is ang aklat 'the book'. Observe the change in agreement morphology on the verb, corresponding to the different choice of subject.2

  1. 3.

    1. a. Humahawak ng aklat si Juan.

      AGR.ASP.hold NS book S Juan

      'Juan is holding a book.'

    2. b. Hinahawak-an ni Juan ang aklat.

      ASP.hold-AGR  NS Juan S  book

      'Juan is holding the book.'

In constructions involving extraction (e.g., wh-movement, relativization), movement of nonoblique and nonsubject arguments is systematically prohibited. This is illustrated by the contrast between the (a) and (b) examples in (4) (from Rackowski 2002) and (5). The subject in (4b) is ang kotse mo 'your car', and the subject in (5b) is si Pedro. [End Page 503] The ungrammaticality of both (4b) and (5b) is the result of extracting the nonsubject argument of the clause.3

  1. 4.

    1. a. Sino ang n-agnakaw  ng kotse mo?

      who S  ACT.PERF-steal NS car  you(NS...