see for further details

Japanese Scrambling and Interpretation

Michiya Kawai (Huron University College, University of Western Ontario)

Japanese allows word-order variations known as scrambling. Scrambling poses two major challenges to the theory of movement within minimalism. Namely, scrambling appears to be optional, although movement is a last resort operation. Bošković and Takahashi (1998) propose that scrambling is a lowering operation of an item base-generated in a non-thematic position.

Consider (1), where the embedded object wh-operator nani-o appears to be scrambled out of the embedded clause.
(1) Nani1-o John-ga [Mary-ga t1 katta ka] shitte-iru.
What-acc John-nom Mary-nom bought Q knows
‘John knows what Mary bought.’
Nani-o is base-generated in its surface position, and obligatorily lowers into the embedded object position. Scrambling, under this analysis, is an obligatory operation, thereby maintaining the last resort nature of movement. This analysis commits itself to an unwanted position, however. That is, scrambling is semantically vacuous, suggesting that it is “undone” at the interface for interpretation (SEM)/Logical Form (LF). This paper argues that scrambling in Japanese is not semantically vacuous.

Scrambling affects the scope of the question operator and topic-focus articulation of sentences (Yanagida 1996, Kawai 1999, 2003, among others), as demonstrated in (2).
(2) a. Mary-wa Pari-de tokee-o katta-no?.
Mary-top Paris-loc watch-acc bought-Q
‘Did Mary buy a watch in Paris?’
b. Mary-wa tokee-o Pari-de katta-no?
‘Was it in Paris that Mary bought a watch?’
c. Un, Pari-de tokee-o katta-yo.
Yup, Paris-loc watch-acc bought
‘Yes, she bought a watch in Paris.’
d. Un, Pari-de katta-yo.
e. Un, tokee-o katta-yo. (Data from Yanagida 1996).
Under a neutral intonation, (2a), a sentence in default word order, can be answered by (2c). In (2b), tokee-o is short-distance scrambled to the left of Pari-de. Observe the interpretation shift: tokee-o moves out of the scope of the question. Therefore, (2d), but not (2c/e), is an appropriate answer for (2b).

That scrambling is not semantically vacuous has a wide range of implications. First, Bošković and Takahashi’s (1998) analysis is untenable, since scrambling must be visible at SEM/LF. Second, scrambling of the type observed in (2) is not optional, thereby maintaining its last resort nature. Third, this result sheds light on the clause internal structure of natural language. That is, in order to escape the question scope via leftward movement, the landing site of short-distance scrambling must be higher than VP, suggesting an intermediate phrasal projection (Polarity Phrase or Distributive Phrase) between Topic and the Q-marker. This supports the recent treatments of complex clausal left-periphery (e.g., É. Kiss 2002, Bayer 2002, Kawai 2003, and Borzdyko 2004), to be discussed in the paper.

References (abbreviated)
Bayer, J. 2002. Decomposing the left periphery.
Borzdyko, O. 2004. La disjonction, l’indefinition, et l’interrogation.
Bošković,Ž., and D. Takahashi. 1998. Scrambling and last resort.
Kawai, M. 1999. Pre-V focus position and focus licensing in Japanese short distance
Kawai, M. 2003. On multiple wh- and quantifier-constructions in Japanese. (ms.)
É. Kiss, K. 2002. The syntax of Hungarian.
Yanagida, Y. 1996. Deriving surface word order in discourse configuration languages.