A-Scrambling Exists!

Martha McGinnis
University of Pennsylvania

The current literature on scrambling focuses on two questions: (i) whether scrambling is feature- driven and (ii) whether it involves A- or A-bar movement (or either, or neither). For example, Frank, Lee & Rambow (1996) argue that in some cases scrambling fails to create new binding relations, and that it also fails to respect the locality condition on A-movement; thus, they conclude that A-scrambling does not exist. In this talk, I will provide evidence that A-scrambling does exist and does respect the locality condition on A-movement. For concreteness, I adopt the Agrless bare phrase structure theory of Chomsky (1995) and the Attract F theory of movement.

A-SCRAMBLING NEED NOT VIOLATE LOCALITY

I begin by proposing that A-scrambling in Japanese involves movement of the scrambled argument into the second specifier of a syntactic head. Specifically, when an object (O) scrambles over the subject (S) (1a), I propose that it moves to a second specifier of vP, creating a structure similar to that underlying the O-S order in Icelandic (Jonas 1996) (1b).

     
(1)a.	[TP T [vP [John-to  Mary]-o  [otagai-no     sensei]-ga  [VP t	mita]]].
                   John-and Mary-ACC each other-GEN teacher-NOM	        saw
	'John and Mary, each other's teachers saw.'
   b.	[CP ½a lįsu [TP t   T [vP [¼essar bękur] aldrei [neinir stśdentar] [VP t ķ fyrra]]]].
	    there	read	these books	never	any students		last year
	'No students ever read these books last year.'

Miyagawa (1997) points out that object scrambling is impossible when the subject both has inherent (genitive) case and fails to trigger subject honorification. I propose that in such cases, T has default case/agreement. Jonas argues that the O-S order is possible only in languages (like Icelandic, unlike English) where T is ³strong² enough to target the features of the subject across the intervening object. In Japanese, T can be strong, allowing O-S order as in (1a), or weak (default), preventing O-S order. Since the object in (1a) is the most local argument for v to attract, moving it into the second specifier of vP does not violate locality.

A-SCRAMBLING RESPECTS GENERAL RESTRICTIONS ON ANAPHORA

As noted in the literature, scrambling can create new quantifier-pronoun binding relations, as well as new anaphor-binding relations. However, one argument that scrambles over another argument cannot bind it directly; the scrambled argument can only bind an anaphor embedded within the other argument (Miyagawa 1997). It can be shown that this restriction does not arise from a ban on nominative anaphors, subject anaphors, or nonsubject binders. Rather, the restriction appears to arise from a ban on anaphoric dependencies obtaining between arguments that occupy specifiers of the same head, at some stage in the derivation.
Frank, Lee & Rambow argue instead that the observed restriction on anaphoric dependencies arises because scrambling does not (ever) involve A-movement. However, the same restriction can be observed in A-movement to the subject position of a ditransitive passive in Albanian (Massey 1992). Here, the lower direct object (DO) first moves into a second specifier of the VP in which the indirect object (IO) is merged, then leapfrogs through this position to spec-TP (2). From this position, a quantifier DO can bind a pronoun embedded in the IO (2a), but the DO cannot bind an anaphoric IO (2b). I maintain that binding is ruled out here because the two arguments occupy specifiers of the same VP before the DO raises to the subject position.

     
(2)a.	[TP [Secili libėr] iu kthye [vP [VP2  tautorit tė tij [VP1 t ]]]].
        each book.NOM	CL returned.NACT	author.DAT its
	'Each book was returned to its author.'
   b.   *[TP Drita iu tregua  [vP [VP2	t vetes	[VP1 t ]] prej artistit]].
        Drita.NOM	CL show.NACT	self.DAT	by the.artist
	'Drita was shown to herself by the artist.'

A-SCRAMBLING RESPECTS LOCALITY

I present two arguments that A-scrambling respects feature-based locality; the head attracting the scrambled argument always attracts the closest DP (or PP). In general, the DO and the IO in Japanese can occur in either order, in part because the DO can scramble over the IO within VP if the IO is a PP (3a) (Miyagawa 1997). However, the IO can be associated with a quantifier only when it is a dative DP, in which case the DO can neither be generated above it nor scramble over it (3b). I show that in such a structure, the DO cannot be attracted nonlocally past the intervening dative DP to an A-scrambled position above the subject (4a). On the other hand, if the IO is a PP, the DO can first A-scramble past it within VP, then scramble again to a position above the subject (4b). (In this case, the scrambled DO cannot directly bind the subject or the IO.)

     
(3)a.	John-ga	[VP2 [pen-o]   Mary-ni [VP1 [t	ni-hon]	ageta]].
        J.-NOM	      pen-ACC  M.-DAT		2-CLS	sent
	'John gave two pens to Mary.'
   b.(???)Mary-ga [CD-o] [tomodati-ni futa-ri]	(t) okutta.
          M.-NOM  CD-ACC friend-DAT   2-CLS	    sent
	'Mary sent two friends a CD.'
     
(4) a.  *[Gakusei-o] [otagai-no     adobaizaa-ga] [VP2 sensei-ni   futari] [VP1 t syookaisita]].
        student-ACC  each other-GEN advisor-NOM	       teacher-DAT two		   introduced
	'The students, each other's advisors introduced to two teachers.'
    b.	[Gakusei-o] [otagai-no     adobaizaa-ga] [VP2 t	sensei-ni   [VP1 t syookaisita]].
        student-ACC each other-GEN advisor-NOM	        teacher-DAT	   introduced
        'Each other's advisors introduced the students to the teacher.'

The second argument for locality in A-scrambling arises from scrambling of idiom chunks. Since idiom chunks cannot long-distance scramble, it has been proposed that they can only undergo A-movement, which is local. Richards (1997) identifies an idiom in which both objects of a double-object construction are idiom chunks. Like (3b), this idiom requires the IO-DO base order. If only one object scrambles, it must be the more local IO (5a); the DO cannot be attracted nonlocally past the intervening DP (5b).

 (5)a.	Hi-ni	  Taroo-ga  [VP2 t  [VP1 abura-o  sosoida]].
        fire-DAT  T.-NOM		 oil-ACC  poured
	'Taroo made things worse.'
    b. *Abura-o	Taroo-ga [VP2 hi-ni    [VP1 t sosoida]].
        oil-ACC	T.-NOM	      fire-DAT        poured
        'Taroo made things worse.'
                 (ok on literal reading: 'Taroo poured oil on the fire.')

REFERENCES

Chomsky, Noam. 1995. Categories and transformations. In The minimalist program. Cambridge, MA: MIT Press.
Frank, Robert, Young-Suk Lee & Owen Rambow. 1996. Scrambling, reconstruction and subject binding. Rivista di Grammatica Generativa 21:67-106.
Jonas, Dianne. 1996. Clause structure and verb syntax in Scandinavian and English. Ph.D. dissertation, Harvard University.
Massey, Victoria Walker. 1992. Compositionality and constituency in Albanian. Ph.D. dissertation, University of North Carolina, Chapel Hill. MITOPL 3.
Miyagawa, Shigeru. 1997. Against optional scrambling. Linguistic Inquiry 28, 1­25.
Richards, Norvin. 1997b. What moves where when in which language? Ph.D. dissertation, MIT. Evidence for feature-driven A-scrambling


< Back to List of Abstracts | Back to PLC23 Home Page >

About the PLC23 Committee
Previously held Penn Linguistics Colloquium: PLC22 (1998), PLC21 (1997)

Penn Department of Linguistics
University of Pennsylvania