In this paper, I argue that scrambling in Japanese is strictly clause-internal, and that all the alleged instances of long-distance preposing in Japanese should be reanalyzed as cases of pseudo-raising, argument transfer, or extraposition. If this analysis is correct, then it becomes possible to maintain that scrambling in a language like Japanese is essentially the same phenomenon as scrambling in a language like German (pace Fanselow (1990) and many others).
(S. Yatabe, "The boundedness of scrambling," in Soonja Choi, ed., Japanese/Korean Linguistics 3, CSLI, Stanford, 1993, pp. 204-220.)