Steinhauer, K. 1, 2 & Drury, J. E. 1, 3
1 Centre for Research on Language, Mind and Brain. McGill University. Montreal, Canada
2 School of Communication Sciences and Disorders. McGill University. Montreal, Canada
3 Department of Lingustics. Stony Brook University. Stony Brook, USA.
Within the framework of an extremely influential serial syntax-first model of sentence processing proposed by Friederici (1995, 2002), the early left-anterior negativity (ELAN) in ERPs has been taken as an important marker of very early automatic parsing processes. Here we attempt to systematically evaluate both the empirical support for the model and some methodological issues of the experimental paradigms used to study phrase structure (PS) violations. For example, we will show that asymmetrical designs in PS violation studies (especially those manipulating the pre-target word context while keeping the target word constant) may result in ERP spill-over effects and DC offset artifacts (baseline problems). Other questions include the following: (1) Are ELAN findings in reading studies comparable to those in auditory studies? (2) Do ELANs elicited in phase 1 of Friederici’s model really occur earlier than LANs elicited in phase 2? (3) Do ELANs reliably reflect ‘outright violations’ of phrase structure? (4) Do PS violations block subsequent processes in phase 2 of the model, such that ELANs never co-occur with LANs or N400s? (5) Are the processes reflected by ELAN components independent of strategies? (6) Do ELAN latencies shift depending on the point in time when word category information becomes available? (7) Are local ELANs (100-300 ms) and (subsequent) sustained frontal negativities (300-1500+ ms) independent effects? - We will discuss empirical evidence as to why, in our view, the answer to all of these above questions is ‘no’. In addition, we will present a fully balanced (symmetrical) PS violation paradigm in English avoiding confounds of previous studies. ERP data illustrate that context effects are real and can be misinterpreted as early violation effects. Implications for Friederici’s model and for future ERP research on syntactic online processing will be discussed.