31 May 2002

The "Sam I Am" Phenomenon: Brain Responses to Leftward Displacement of Constituents

Robert Kluender

University of California, San Diego

The difference between subject and object dependencies has long been a vexing question for linguistic theories. As is well known, in many cases the formation of long-distance subject dependencies into island con- figurations results in a greater degree of ill-formedness than the formation of corresponding long-distance object dependencies. It is less well known that the formation of object dependencies into non-island embedded clauses disturbs readers more than subject dependencies into the same embedded contexts. In these cases, standard linguistic grammaticality judgements reveal no obvious difference in well-formedness, whereas more fine-grained psycho- and neurolinguistic measures do. The focus of this talk will be on just such cases.

Event-related brain potentials (ERPs) constitute one measure that has been fruitfully applied in this domain. We know from a number of studies over the past ten years that there are fairly reliable indices of long-distance A'-dependency formation in the brain's responses to such structures. These indices have been taken to reflect the additional working memory costs associated with processing constituents displaced from their canonical positions.

In this talk I will review these findings and report on the most recent study from our lab, which extends this research paradigm to contexts of Japanese scrambling. In this study, we simply permuted the subject and object constituents of Japanese sentences from SOV to OSV word order. Nonetheless, the ERPs revealed familiar indices of long-distance dependency formation related to working memory costs. This suggests that the disruption of canonical word order even at a fairly local level results in an additional processing load for Japanese readers. Related findings from other ERP studies suggest further that the brain response to long-distance dependencies may be more compatible with an approach based on the processing of canonical word order than with an approach based on the processing of specific phrase structural positions.