Despina Papadopoulou & Harald Clahsen
University of Essex
Relative clause (RC) attachment preferences have been extensively discussed in recent sentence processing research with respect to general theoretical issues, such as the universality of the parser and the role of contextual cues in early parsing decisions. Several studies on English have shown that in an ambiguous sentence (e.g., The girl looked at the psychiatrist of the actress who was having tea) in which the relative clause can be attached either high to the first DP (the psychiatrist) or low to the second DP (the actress), the parser will preferentially attach the RC low. This is in line with what a universal Late Closure/Recency preference would predict. However, studies that tested similar kinds of sentences in a number of different languages found that a low attachment preference does not hold universally. Moreover, the question of whether attachment preferences are influenced by contextual information and lexical/thematic biases is controversial and has not yet been resolved.
In this paper, we will report results from acceptability judgment and self-paced reading experiments examining RC attachment preferences in native speakers of Greek. Results from four experiments will be presented. To test the role of lexical/thematic information for RC attachment, the materials in all four experiments were divided into two conditions with different RC antecedents, complex DPs with genitives as in (1), and complex DPs containing lexical PPs as in (2); otherwise the sentences were identical. The critical sentences were disambiguated by the form of the participle/adjective, through gender agreement.
| (1) |
Enas theatis kitakse tin athlitria tou proponiti pou itan apogoitevmenos ... |
|
"A spectator looked at the athlete(fem) of the trainer(masc) that was disappointed ..." |
|
| (2) |
Enas theatis kitakse tin athlitria me ton proponiti pou itan apogoitevmenos ... |
|
"A spectator looked at the athlete(fem) with the trainer(masc) that was disappointed ..." |
To examine context effects, the same sentences that were presented in Experiments 1 and 2 in isolation, were in Experiments 3 and 4 embedded in short stories yielding two conditions; in the high-attachment-biasing-context condition, two referents were introduced for the first noun of the complex DP, and in the low-attachment-biasing-context condition there were two referents for the second noun of the complex DP.
The four experiments revealed a very consistent picture, namely a high attachment preference for sentences such as (1) and a low attachment preference for sentences such as (2), showing that lexical and/or thematic properties of the antecedent affect the attachment of the RC. We will discuss explanations of these findings in terms of Construal (Frazier & Clifton, 1996) and Recency/Predicate Proximity (Gibson et al., 1996). Constraint-based approaches (e.g., Thornton et al., 1999) cannot explain our results, particularly the finding from Experiments 3 and 4 that biasing contextual information was unable to override the participants' parsing preferences.
References
Frazier, L., & Clifton, C. (1996). Construal. Cambridge, MA: MIT Press.
Gibson, E., Pearlmutter, N., Canseco-Gonzalez, E., & Hickock, G. (1996). Recency preferences in the human sentence processing mechanism. Cognition, 59: 23-59.
Thornton, R., MacDonald, M., & Gil, M. (1999). Pragmatic constraints on the interpretation of complex noun phrases in Spanish and English. Journal of Experimental Psychology: Learning, Memory, and Cognition, 25(6): 1347-1365.