Attention, learning and the dynamics of sentence processing: Event-related potential studies of individual differences

In this dissertation, I used the event-related potential (ERP) technique to investigate sentence comprehension and test predictions of syntax-first and constraint-based theories. In the syntax-first model, information is processed serially and modularly. Constraint-based models, however, argue that greater computational efficiency could be achieved if the system used context to constrain syntactic processing immediately and continuously. The two primary goals of this dissertation are to investigate the following issues: (1) whether early syntactic processing is encapsulated as proposed by syntax-first theories, and (2) whether individual variability in syntactic proficiency is correlated with different neural patterns during sentence processing. I used three experiments to re-evaluate the early encapsulated syntactic processing hypothesis. Experiment 1 was designed to replicate and extend the results from Hahne and Friederici (1999). My results suggested that at least in written English presentation, the ERP component hypothesized to index early encapsulated syntactic processing (the early left anterior negativity, or ELAN) is not encapsulated, as it was modulated by attentional manipulations. In Experiment 2, I manipulated local and global sentence coherence in order to test whether the ELAN would index competition between global and local coherence. Instead, anomalous sentences containing a local coherence generated larger LAN (slightly later than ELAN) and P600 (classic index of syntactic difficulty) responses than sentences without local coherences. Importantly, the materials we used are not subject to challenges that have been made to earlier behavioral studies of local coherence, and therefore these results bolster the self-organizing parser theory of Tabor and colleagues (e.g., Tabor, Galantucci, and Richardson (2004). In Experiment 3, I investigated whether an early ERP response can be elicited by non-outright category violation sentences, such as sentences that require challenging pronoun gender resolution. The result showed no early ERP response differences between pronoun gender match and mismatch conditions, but suggested that variation in an individual\u27s overt behavioral performance may be indexed in neural performance in ways that shed new light on individual differences in language processing abilities.