ИСТИНА

Split-PPs in Russian have already been researched with experimental methods by I. Sekerina (1997). Nevertheless, there are several premises for our work. Firstly, some of I. Sekerina's experimental-based conclusions were lately disproven by A. Pereltsvaig (2008) in their examination of Colloquial Russian corpus data. Secondly, I. Sekerina only observed scrambled PPs, but other types of A'-movements can also trigger XP-splits, cf. (Fanselow & Ćavar 2002). Finally, splitting is in general a more resource-costly operation than full movement (Nunes 2004); however, there is no research on whether it is sensitive to the phonological weight or structural complexity of the parts of a constituent. In particular, there were no previous works on whether dislocating a head from its complement is acceptable in comparison with dislocating a left element. Thus, we conducted an experiment using the Likert scale 1–7 and self-paced reading tasks. We used preposition phrases containing noun phrases with elements on their left periphery as well as infinitival or PP-complements. The results are interesting in two ways. On the one hand, the extremely low ratings of split scrambling are not in line with the results of I. Sekerina's experiments. As scrambling is sensitive to the information structure, it can be hard to process without context. On the other hand, the acceptability of early and late splits differing for wh-movement and relativization is unexpected: one could suppose that all late splits would be rated lower because they form a meaningful combination of words on the left periphery. The observed difference is probably related to the information structure as well: wh-movement and relativization favorise different communicative statuses of moving parts, cf. Lyutikova (2019).