Wing-Yee Chow  

Research

I am interested in how humans process language in real-time. My research incorporates behavioral as well as electrophysiological techniques to examine the representations and processing mechanisms that underlie real-time language comprehension.

 

Predictive computations in real-time language comprehension

Relation between real-time syntactic processing and semantic interpretation

Memory representations and retrieval mechanisms in anaphora resolution

 


Predictive computations in real-time language comprehension

Much evidence suggests that fast and robust language understanding depends on predictive mechanisms. Just as a tennis player’s ability to anticipate a flying ball’s trajectory increases his/her chances of hitting the ball, our ability to anticipate likely upcoming input allows us to recognize that input more rapidly, or more reliably when it is degraded. Recent research has gathered a growing body of evidence for the prevalence of predictive computations in language comprehension, but much less is known about how we succeed or why some speaker groups struggle. By exploring when and how different sources of contextual information feeds linguistic predictions, my research aims to build an explicit model of how linguistic predictions are computed during real-time comprehension.

(Studies in Mandarin Chinese in collaboration with Suiping Wang at South China Normal University)

Methods: Event-related potentials (ERP); Reading eye-tracking.

Back to top

 

Relation between real-time syntactic processing and semantic interpretation

Is real-time semantic interpretation ever decoupled from syntactic analysis? Recent reports of unexpected brain potential responses to grammatically well-formed but thematically role-reversed sentences (the “Semantic P600” phenomenon) have been taken as evidence for syntax-independent semantic interpretation. Through basic research on the functional significance of language-related brain responses, my research examines alternative interpretations of the phenomenon and suggests that this phenomenon can provide insights into how syntactic information informs the computation of expectations about thematic relations.

Methods: Event-related potentials (ERP).

Back to top

 

Memory representations and retrieval mechanisms in anaphora resolution

Since pronouns and reflexives require a referent that is present (or implied) earlier in the discourse, the task of accurately interpreting these anaphors in real-time requires retrieving the target referent from memory and might be susceptible to interference from inappropriate antecedents in memory.

 

What representation(s) do we access when we interpret a pronoun?

Previous research has shown that the processing of a pronoun is modulated by various properties of its antecedent. In my collaboration with Sol Lago and Ewan Dunbar, we examine the nature of the mental representation that underlies pronoun resolution.

Methods: Reading eye-tracking; distributional analyses of eye-movement data.

 

What are the roles of grammatical and pragmatic knowledge in pronoun resolution?

Linguistic theories proposed that the dependency between a pronoun and its referent can be established through distinct mechanisms (variable binding vs. coreference). This distinction has interesting implications for language processing, since the two kinds of dependencies are subject to different constraints and are construed as operations on distinct levels of representations. In my collaboration with Shevaun Lewis, we examine how comprehenders employ their grammatical and pragmatic knowledge to identify the referent of a pronoun in real-time.

Methods: Self-paced reading; reading eye-tracking.

 

What are the memory retrieval mechanisms that underlie real-time interpretation of anaphors?

Previous as well as my own research has demonstrated impressive grammatical accuracy in online pronoun interpretation. Such observations (the lack of interference from inappropriate antecedents in memory) are perhaps unexpected according to models that assume general content-addressability in memory retrieval operations during language processing. In my collaboration with Brian Dillon (UMass), Matt Wagers(UCSC), Ming Xiang (UChicago), Fengqin Liu and Taomei Guo (Beijing Normal University), we examine the hypothesis that, upon encountering a long-distance reflexive (e.g., ziji in Mandarin Chinese), comprehenders employ a retrospective memory search mechanism that queries a targeted set of structural positions serially to retrieve the antecedent of the anaphor.

Methods: Speed-accuracy tradeoff analysis; self-paced reading.

Back to top