Prof. Karsten Steinhauer, McGill University

时间： 2015-11-19 12:50 - 14:50

地点： Room 1113, Wang Kezhen Building

Event-related brain potentials (ERPs) provide a great tool to investigate the real-time processing of language across the entire utterance. Over the past three decades ERP data have, therefore, played an important role in refining psycholinguistic models. For example, the intuition that syntactic structure rather than world knowledge should guide sentence interpretation (e.g., in ‘man bites dog’) led to influential ‘syntax-first’ parsing models (e.g., Frazier & Fodor, 1978). The subsequent discovery of very early ERP components for syntactic processing in the early 1990s, as well as ERP evidence demonstrating that certain syntactic violations can ‘block’ conceptual-semantic processing (Friederici, Steinhauer & Frisch, 1999), contributed important ‘hard brain data’ in support of ‘syntax-first’ accounts and inspired extremely influential neurocognitive models (e.g., Friederici, 2002, 2011). However, methodological problems – often present in ERP studies that contradicted ‘syntax-first’ models – may also be found in studies supporting them (e.g., Steinhauer & Drury, 2012).

In my talk I will try to give an introduction to how inappropriate stimulus materials, lack of control conditions, and inadequate ERP analyses can result in superficially compelling but invalid brain data (which, by the way, have sometimes been cited thousands of times). The phenomena include ERP effects that precede the linguistic stimuli whose processing they are supposed to reflect by several hundreds of milliseconds. I will also present data illustrating our recent problems replicating the ‘semantic blocking’ effect of Friederici et al.’s 1999 study. As we will see, some of the challenges in ERP data interpretation have to do with individual processing differences, possibly pointing to new approaches in data analysis. In conclusion, ERPs remain a very useful tool in advancing our understanding of online sentence processing, but it is crucial to design and analyze the studies with much care.

2015-11-19

2015-11-19