Wh-Movement vs. Scrambling in Russian On-Line Processing
Irina Sekerina

1. Introduction. Recent developments in sentence processing theory have been investigating two main questions concerning empty categories (ECs), phonetically null place-holders for locally absent constituents. The first question is whether there is empirical evidence for the psychological reality of ECs, which may help tease apart the preditions of different theories of grammar, e.g., Minimalism vs. traceless theories. The second question bears on exclusively processing issue of complexity, under the assumption that filler-gap dependencies may result in extra cost. The goals of this study were (1) to look for processing evidence for postulating filler-gap dependency in these two constructions, and (2) to examine the processing complexity issue.
2. Design and Materials. Two experiments with Russian speakers using self-paced reading technique were conducted. Experiment 1 (33 subjects) had three conditions (1)-(2), (4), and the sentences were presented in isolation. Experiment 2 (88 subjects) had four conditions (1)-(4), and crucially differed from Experiment 1 in that every sentence was preceded by a one-sentence context. Each sentence out of 24 contained a double-object verb and was presented to subjects in one of the following conditions:

Context: Mama skazala, chto babushka zabolela, a v sadu uzhe sozreli jagody.
'Mother said that Grandmother fell ill but the berries have ripened in the garden.'

3. Predictions.

• Processing Cost. According to the current theories of sentence processing, the Garden-Path model (Frazier and Clifton, 1996), the Constraint-Based model (Tanenhaus and Trueswell, 1995), and the SPLT model (Gibson, 1998), the sentence processor will respect the grammar in the following way. If the grammar requires a movement analysis of a construction, this will add a processing cost (reflected in increased RTs). Thus, if Scrambling involves movement of the scrambled DP, both questions (4) and scrambled sentences (2) will be read slower than their respective controls (3) and (1).
• Locus of Processing Cost. The Trace Reactivation Account (Swinney at al., 1989) attributes processing cost to binding of ECs by their antecedents at the position of the gap (at the Indirect Object). The Direct Association Hypothesis (Pickering and Barry, 1991) equates processing cost with processing of dependents at their subcategorizer (at the Verb).
• Syntactic vs. Semantic Processing. If questions with wh-in-situ (3) involve covert LF-Movement, they may show increased RTs due to processing cost at the level of interpretation. In this case, both types of questions (with syntactic Wh-Movement (4) and without (3)) will be processed in a similar way.
• Discourse Effects. Functional accounts of Russian word order variation in grammar emphasize the discourse-oriented nature of Scrambling. Appropriate context is assumed to make scrambled word order felicitious. Translated into processing terms, presence of context coupled with unambiguous Case-marking on the scrambled DP should faciliate processing of such constructions. Such context is not required for questions. In this case, we will expect to find processing cost associated with scrambled sentences (2) in Experiment 1 (no context), in contrast to no such processing cost in Experiment 2 (context present).

4. Results.

|BACK TO PROGRAM|