Rachel Brown Abstracts
Rachel Brown
Ph.D. Candidate
Arid Lands Resource Sciences GIDP
32nd Annual CUNY Conference on Human Sentence Processing
Boulder, Colorado
March 29-31, 2019
While context influences language comprehension [see 1 for review], the interaction between nearby local context and long-distance global context remains unclear [2,3,4]. The current study investigates the relative role of local and global context during reading comprehension of ambiguous phrases using EEG imaging to examine processing independent of external behavioral responses. This study shows that discourse-level context effects influence immediate parsing, followed by effects of local cues. This supports ‘Top Down’ models of sentence processing where long-distance context plays a key role.
We examine the contextual processing of ambiguous phrases with varying semantic properties. For example, the biased stimuli wrecked commuter ship prefers the Right Branching interpretation commuter ship that is wrecked over the Left Branching interpretation ship with wrecked commuters, while automatic copying machine has not preferred reading generally (i.e., Right Branching: copying machine that is automatic/ Left Branching: machine for automatic copying). Brain activity was recorded at the 2nd and 3rd words of the phrases, while the 1st and 1st+2nd words served as the local context respectively. The phrases are preceded by global contexts that correspond to the possible interpretation of the ambiguous phrase (i.e., ship of wrecked passengers or passenger ship that is wrecked). They are designed to influence the ease of interpreting the critical words with immediately preceding local context. For example, global context ship of wrecked passengers facilitates the processing of ‘wrecked commuter’ later on because wrecked explicitly integrates with similar word to passenger.
24 native English-speaking subjects read sentences with the ambiguous phrases in context in RSVP format. ERPs were generated at 300-500ms and 500-800ms after critical word onset midline electrodes (Fz, Cz, Pz). Separate analyses of 2nd and 3rd word ERPs were conducted using repeated measures ANOVAs with the factors: 2 Global Contexts x 4 Local Contexts. At the 2nd word, there was no significant Global*Local interaction and no main effect of Local Context.
The Global Context manipulation resulted in a N400 and anterior negativity sustained across both time windows where the critical word is significantly more negative when preceded by a Right Branching Global Context relative to preceded by a Left Branching Global Context. Decreased negativities in the Left Branching Global Context condition suggest that explicit integration of the local context and critical word facilitates integration in the ambiguous phrase. At the 3rd word, there was no significant Global*Local interaction and no main effect of Global Context. The Local Context manipulation resulted in a N400 where the critical word is significantly more negative when preceded by an inconsistent local context (i.e., wrecked commuter + ship) in comparison to when the local context is consistent (i.e., furry tiger + painting).
Abstract for Lay Audience
While context influences language comprehension, the relative roles between nearby local context and long-distance global context remains unclear. The current study investigates the relative role of local and global contexts during reading comprehension of ambiguous phrases (i.e., phrases with more than one possible meaning) using EEG imaging (i.e., electroencephalography imaging) to record brain activity.
This experiment examines the reading of ambiguous phrases with varying meaning properties. For example, the biased stimuli wrecked commuter ship prefers the interpretation commuter ship that is wrecked over the interpretation ship with wrecked commuters. In contrast, furry tiger painting prefers the opposite interpretation (i.e., painting of a furry tiger over tiger painting that is furry). Brain activity was recorded at the 2nd and 3rd words of the phrases (e.g., commuter and ship), while the 1st and 1st+2nd words served as the local context respectively (e.g., wrecked and wrecked commuter). The phrases are preceded by global contexts that correspond to the possible interpretation of the ambiguous phrase (i.e., ship of wrecked passengers or passenger ship that is wrecked). They are designed to influence the ease of combining critical words with immediately preceding local context. For example, global context ship of wrecked passengers facilitates the understanding of ‘wrecked commuter’ later in the sentence because wrecked explicitly combines with similar word to passenger.
24 native English-speakers read sentences with ambiguous phrases in global contexts. Brain signals called ERPs (i.e., event-related potentials) were recorded. The Global Context manipulation resulted in significant differences at the 2nd word of the phrase. This suggests that explicit combination of the local context and critical word within the prior global context facilitates the reading of the ambiguous phrase. The Local Context manipulation resulted in significant differences at the 3rd word of the phrase. This suggests that reading is harder when the critical word is preceded by an inconsistent local context (i.e., wrecked commuter + ship) in comparison to when the local context is consistent (i.e., furry tiger + painting). Thus, while inconsistencies between the critical word and the Local Context occur at both the 2nd and 3rd word, the results suggest that Local Context effects appear at the 3rd word only when Global Context effect fades. This study shows that long-distance global context effects influence the interpretation of ambiguous phrases, followed by effects of local immediate cues to the intended meaning. This supports models of sentence comprehension where long-distance global context plays a key role.
These findings are consistent with prior work showing that global context cues override local cues. However, this study adds to the literature by investigating ambiguous phrases – phrases where global context plays a pivotal role in determining the intended meaning. In addition, previous research used striking overall incongruences between the global context and the critical words. The proposed project employs carefully controlled global contexts with either identical or highly related words to the critical words to test how subtle incongruences between context and ambiguous phrases influence language processing. The findings show that even subtle long-distance cues from previous global context can impact of processing of ambiguous phrases and can even override the immediately preceding contextual cues.