Ina Bornkessel,1 Brian McElree2 & Matthias Schlesewsky3
1 Max Planck Institute of Cognitive Neuroscience, Leipzig, 2 New York University, 3 University of Potsdam
Reanalysis processes in sentence comprehension are typically associated with the P600 event-related brain potential. However, a non-preferred (object-initial) resolution of subject-object ambiguities in German sentences such as (1a) has been shown to elicit an N400 component in comparison to control sentences resolved towards a subject-initial reading (1b; Bornkessel, Schlesewsky & Friederici, 2001).
| (1) | a. | Gestern wurde erzählt, dass Maria Sängerinnen
folgen. yesterday was told that Maria-AMB.SG singers-AMB.PL follow-PL "Yesterday, it was said that singers follow Maria." |
| b. |
Gestern wurde erzählt, dass Maria Sängerinnen folgt. yesterday was told that Maria-AMB.SG singers-AMB.PL follows-SG "Yesterday, it was said that Maria follows singers." |
The present study examined the hypothesis that the N400 observed for sentences such as (1a) may indeed be interpreted as reflecting reanalysis processes by means of the multiple-response speed-accuracy tradeoff (SAT) method (McElree, 1993). In this paradigm, participants continually judge the acceptability of a sentence while reading (by means of a button press every 350 ms). The time-course functions derived from this method may be described by three parameters: (a) an asymptote, reflecting the level of acceptability reached with functionally unlimited processing time; (b) an intercept, describing when performance departs from chance level; and (c) a rate of rise indexing how quickly accuracy increases from chance to the asymptote. If the N400 found in processing structures like (1a) reflects reanalysis, these structures should be associated with a slower processing speed than minimally contrasting control structures (1b) because additional time is required to compute the non-preferred reading. The slower processing speed will be reflected in differences in SAT intercept or rate. In contrast, if the N400 simply reflects that structures like (1a) are overall less plausible than control structures, differences should be reflected in SAT asymptote alone.
Results showed that object-initial sentences were associated with both a lower asymptotic value and a delayed intercept relative to subject-initial counterparts. The finding of a difference in processing dynamics (i.e., the intercept difference) indicates that the processes reflected in the N400 observed for sentences such as (1a) may indeed be interpreted in terms of reanalysis. We assume that this particular type of reanalysis differs from a P600-eliciting reanalysis of subject-object ambiguities in that disambiguation is effected by a semantically contentful verb rather than a clause-final auxiliary, i.e., at a stage of processing when no associations between the predicate and its arguments have as yet been made.
Reference
Bornkessel, I, Schlesewsky, M., & Friederici, A.D. (2001). The application of universal hierarchies during sentence processing: Evidence for incremental interactive thematic processing. Paper presented at the Architecture and Mechanisms for Language Processing Conference, Saarbrücken.