|
Reconsidering ERP evidence for early syntactics encapsulation| old_uid | 9128 |
|---|
| title | Reconsidering ERP evidence for early syntactics encapsulation |
|---|
| start_date | 2010/10/07 |
|---|
| schedule | 11h-12h |
|---|
| online | no |
|---|
| summary | This project is inspired by a paradox in the sentence processing literature. Eye tracking studies have found repeatedly that non-syntactic constraints are used by comprehenders as quickly as we can measure their impact, with impact proportional to constraint strength. The constraints that have been studied include visual context (Tanenhaus et al., 1995), object and instrument affordances (Chambers et al., 2008), and expectations based on conversational pragmatics and contingencies within an experiment (Magnuson et al., 2008). However, a large number of ERP studies are consistent with Friederici’s (2002) theory of sentence comprehension, which posits an early, encapsulated syntactic processing phase (Phase 1, 100-300 msecs) for each word followed shortly by integration. The primary index of Phase 1 is the early left anterior negativity (ELAN), which is observed in response to immediately detectable form class violations. This component is the linchpin of the theory, as it provides the strongest evidence for an early, context-blind mechanism that compares a word's possible syntactic category to the category or categories expected given the emerging syntactic structure of an ongoing sentence. Hahne & Friederici (1999) reported that a simple attentional manipulation -- the proportion of anomalous sentences used -- did not affect the ELAN; it was observed in response to an easily detectable error when anomaly proportion was high or low. In contrast, a later component that the theory posits indexes a later integration stage (the P600, observed in response to syntactic errors even when they are not immediately detectable) was greatly weakened when anomaly proportion was high. We revisited this result because we find a special-purpose mechanism for checking syntactic expectations implausible (as opposed to the ELAN reflecting the operation of mechanisms that are actually engaged in the "work" of sentence processing). We conducted an ERP study modelled after Hahne & Friederici (1999), but with three levels of anomaly proportion (.2, .5, .8). Our results differed strikingly from Hahne & Friederici. We observed weak and strong ELANs for low and medium anomaly proportions, and a reversed effect for the high anomaly condition (as though correct sentences contained an error, since anomalies were so strongly expected). It appears that the ELAN is sensitive to attentional manipulation, and it can be reversed, such that it is triggered by a word that is perfectly grammatical when it occurs in a construction that is rare within the experimental session. These findings pose serious challenges for Friederici's (2002) early-encapsulation, late-integration model. |
|---|
| responsibles | Pélissier |
|---|
| |
|