Processing Multiple Filler-gap Sentences In Russian: Chain Avoidance

Abstract:

Lyn Frazier, University of Massachusetts at Amherst

In a written questionnaire 20 native speakers of Russian were presented with 28 sentences in a grammaticality judgment task. The 14 critical sentences contained two clauses. Each sentence began with a sequence of two fronted NPs, as in (1).

(1) Ja Volodja verju rasskazhet pravdu. I-NOM Volodja-NOM believe-1Sg will tell-3Sg the truth-ACC 'Volodja, I believe, will tell the truth.'

Overall the sentences were judged to be "pretty bad" (average 1.73 on a 5 point scale where 5 is "perfect"), presumably because of their syntactic complexity. The sentence with the highest rating (2) was ambiguous.

(2) Olja Masha dumaet uvazhaet Ivana. Olja-NOM Masha-NOM thinks-3Sg respects-3Sg Ivan-ACC (a) 'Olja thinks Masha respects ivan.' (b) 'Masha thinks Olja respects Ivan.'

Due to the particular number, person, and case features of the intial NPs,(2) can be analyzed as containing two chains, as in (3a), or just as a single chain, as in (3b).

(3)

Consistent with the Minimal Chain Principle (De Vincenzi 1989, 1991) preceivers apparently assigned (2) the structure in (3b).

To check our explanation of the high rating assigned to (2), ambiguous sentences like (2) were given to native Russian speakers who were asked to indicate their initial interpretation by choosing the preferred interpretation out of two possible ones. As expected, 85 of the responses indicated that subjects favored the structure in (3b).

The results provide further evidence for the existence of traces (see Pickering and Barry, 1991, for challenges to the theory that traces are generated in canonical positions from which they may raise, leaving a conidexed trace). In the absence of traces, there is no reason to expect (3a) to be structurally more complex than (3b). Further, since neither of the initial NPs can be combined with the verb until the verb has been encountered, there is no basis for claiming that interpretation (3a) imposes a greater memory burden than (3b).

Another possible alternative explanation is that (3a) is more complex than (3b) due to the crossing dependencies. Note, however, that multiple Scrambling in Russian within a single clause with a double-object construction can produce equally acceptable either nested or crossed dependencies. Further, intuitive evidence suggests that both nested and crossed dependencies are allowed in Long-Distance multiple Wh-Movement and Scrambling provided the additional prosodic stress and pausing are present. (An additional study is in progress to test these facts.) We thus maintain that the preference we have established follows automatically from the independently needed Minimal Chain Principle, and the preference for (3b) provides a simple but strong argument for the existence of traces and additional evidence for the cross-language validity of the Minimal Chain Principle.