|Title:||Split scrambling: barriers as violable constraints|
|Comment:||WCCFL XX proceedings, 2001|
|Abstract:||In contrast to English (3), Russian allows movement out of a DP (1,2). This kind of movement is called split scrambling. Both Topics (1) and Foci (2) can split scramble, Topics moving to sentence-initial position, and Foci to sentence-final position.
(1)[Bol’šaja t DP] ukusila men’a gorilla.
big bit me gorilla
A big gorilla (Focus) bit me.
(2) Bol’šuju my videli [t gorillu DP] vèera (Russian)
(3) *Big we saw gorilla yesterday
We saw a big (Topic) gorilla.
I argue that (1) and (2) are grammatical because the DP is a barrier in English but not Russian. The central question of this paper is what determines the status of phrases as barriers cross-linguistically. I claim that barriers are violable, universal constraints in an Optimality Theoretic grammar, and that the ranking of the barrier constraints forms a fixed hierarchy from smallest to largest: DP BARRIER, PP BARRIER >> VP BARRIER. The analysis predicts that no language allows movement out of DP or PP but bans the same type of movement out of VP. If a constraint that requires movement dominates DP BARRIER or PP BARRIER, it also dominates VP BARRIER. Russian and Polish allow movement to cross all three barriers, while English tolerates crossing VPs and PPs but not DPs, and Malagasy bans movement even out of VPs (Keenan 1976).
Split scrambling arises from the conflict of two opposing forces: constraints that demand movement on the one hand, and constraints that oppose it on the other. In Russian, DP Barrier is dominated by the requirements to place Topics and Foci in their positions (ALIGN-R FOCUS (Grimshaw and Samek-Lodovici 1998), ALIGN-L TOPIC, and CONTRASTIVE FOCUS SPEC). Thus, movement out of DPs is not only possible but necessary for Topics and Foci. In English, DP BARRIER dominates the Focus/Topic constraints, so Topics and Foci cannot move out of a DP.
The constraints that oppose movement and those that require it can be interspersed, e.g. MOVE WH >> DP BARRIER >> ALIGN-L TOPIC >> PP BARRIER >>VP BARRIER. It is thus predicted that barriers for some kinds of movement are not necessarily the same as barriers for other kinds of movement in the same language. Thus, Mandarin Chinese allows free Topicalization out of DP, but no Wh-movement. This and other cases of variable barrier status within the same language are straightforwardly explained in the framework I propose.
This work also addresses the motivation of pied-piping, which I claim to be a repair of BARRIER constraint violations. What gets pied-piped will vary from language to language, depending on which BARRIER constraints are highly ranked. This work also bears on the question of whether or not Alignment constraints are gradient, and argues that they are.