About

Ellen Lau is an Associate Professor in the Department of Linguistics and a member of the Maryland Language Science Center at the University of Maryland. She is also a Co-Director of the KIT-Maryland MEG Lab and faculty in the Program in Neuroscience and Cognitive Science. Her research expertise encompasses neurolinguistics and psycholinguistics, with a focus on understanding neural and cognitive mechanisms underlying language processing. Her work challenges traditional lexicalist models of language by exploring shared neural computations for morphological and syntactic structures, providing evidence for non-lexicalist theories of language comprehension. Lau investigates the boundaries of the binding problem beyond vision, examines processing differences in ergative languages such as Georgian, and studies parallel models of speech perception that incorporate local and global linguistic contexts. Her research also includes the online processing of adjunct control and reference resolution, contributing to a deeper understanding of syntactic and structural influences on language comprehension.

Research topics

• Computer Science
• Natural Language Processing
• Artificial Intelligence
• Psychology
• Cognitive psychology
• Speech recognition
• Neuroscience
• Communication
• Linguistics
• Cognitive science

Selected publications

• How single-neuron computation matters for cognitive neuroscience

  2025-03-31

  articleOpen access1st authorCorresponding

  A traditional and pervasive view in neuroscience is that memory and computation are wholly supported by weighted connections between neurons (synapses).As in a 'neural net' architecture, computation occurs when a pattern of activity is instantiated at an input layer, which initiates a chain reaction: each neuron's activity can affect others to which it is 'connected' by a synapse and this reaction propagates through the network, determining the output of the function which is given by the resulting values of the neurons designated as the 'output layer'.Experience-driven memory is constituted by changes to the weights on the connections, that is the 'strength' of the synapse.Many considerations, however, militate against this 'synaptocentric' view of memory and computation.One is

  PublisherOA PDFDOI

• Features fade, pointers persist: dissociable parietal mechanisms in visual working memory formation and maintenance

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-07-03

  preprintOpen accessSenior author

  Abstract An emerging body of work has adopted the theoretical construct of pointers: an object is mentally represented by a content-free pointer binding the corresponding features together. Indeed, fMRI studies have highlighted dissociable parietal regions sensitive to pointer and feature load respectively. Specifically, while the superior IPS (intraparietal sulcus) is sensitive to feature load, inferior IPS is only sensitive to the number of pointers – i.e., the number of objects. However, the spatiotemporal dynamics remain unclear, and therefore it is unknown whether these effects reflect visual working memory (VWM) formation, maintenance, or both, especially given limited temporal resolution of fMRI. In our current MEG (magnetoencephalography) study, participants memorized visual arrays with different numbers of objects (object/pointer load: two, four), as well as different features and feature-per-object (color, orientation, bifeatural) in a VWM task. We observed a dissociation between inferior and superior IPS in the temporal dynamics of feature-sensitive and pointer-sensitive responses. While pointer-sensitive signals persisted across VWM formation and maintenance in inferior IPS, feature-sensitivity was only transiently manifested during VWM formation in superior IPS. This spatiotemporal dissociation may reflect a representational architecture optimized for efficiency, reducing the need for sustained neural activity to maintain features once they are bound to pointers. In revealing the spatiotemporal profile of pointer and feature representations, our results provide novel evidence on how pointers underlie energy-efficient neural representations in VWM.

  PublisherOA PDFDOI

• A finite set of content-free pointers in visual working memory: magnetoencephalography (MEG) evidence

  Neuroreport · 2025-01-29 · 6 citations

  articleSenior author

  Human visual working memory (VWM) is known to be capacity-limited, but the nature of this limit continues to be debated. Recent work has proposed that VWM is supported by a finite (~3) set of content-free pointers, acting as stand-ins for individual objects and binding features together. According to this proposal, the pointers do not represent features within themselves, but rather bind features represented elsewhere together. The current study set out to test if neural hallmarks resembling these content-free pointers can be observed with magnetoencephalography (MEG). Based on two VWM delay-match-to-sample experiments (N = 20 each) examining memory for simple and complex objects, we report a sustained response in MEG over right posterior cortex whose magnitude tracks the core hypothesized properties of this content-free pointer system: load-dependent, capacity-limited, and content-free. These results provide novel evidence for a finite set of content-free pointers underlying VWM.

  PublisherDOI

• Unleashing the potential of OPM-MEG to study event-related fields against low-frequency artifacts: the case of sentence processing

  bioRxiv (Cold Spring Harbor Laboratory) · 2025-03-15

  preprintOpen access

  Abstract OPM-MEG (optically pumped magnetometers-magnetoencephalography) offers unprecedented opportunity in its proximity to the brain and in mobility, outperforming other human MEG systems. However, movements induce low-frequency artifacts (up to ∼ 3 Hz) in this system, calling for pipelines that could efficiently reduce these low-frequency noises. Although a high-pass filter of e.g., 4 Hz may minimize these artifacts, it may also eliminate many event-related field (ERF) components, such as the N400 response in sentence processing studies. Moreover, as an emerging technology with expensive sensors, many labs are starting with an experimental setup with fewer sensors (e.g., ∼ 10), making it challenging to reject principal components based on visually inspecting the topographic field maps. In the current paper, we show that a combination of moderate high-pass filtering (1 Hz) and evoked-biased denoising source separation (evoked-biased DSS) can effectively reveal typical N400 deflections and effects (between nouns and verbs), with a nine-sensor OPM-MEG setup. Our current pipeline paves the way to studying ERFs with OPM-MEG in more budget-friendly setups with a small number of sensors.

  PublisherOA PDFDOI

• Electrophysiological hallmarks for event relations and event roles in working memory

  Frontiers in Neuroscience · 2024-01-23 · 4 citations

  articleOpen accessSenior author

  The ability to maintain events (i.e., interactions between/among objects) in working memory is crucial for our everyday cognition, yet the format of this representation is poorly understood. The current ERP study was designed to answer two questions: How is maintaining events (e.g., the tiger hit the lion) neurally different from maintaining item coordinations (e.g., the tiger and the lion)? That is, how is the event relation (present in events but not coordinations) represented? And how is the agent, or initiator of the event encoded differently from the patient, or receiver of the event during maintenance? We used a novel picture-sentence match-across-delay approach in which the working memory representation was "pinged" during the delay, replicated across two ERP experiments with Chinese and English materials. We found that maintenance of events elicited a long-lasting late sustained difference in posterior-occipital electrodes relative to non-events. This effect resembled the negative slow wave reported in previous studies of working memory, suggesting that the maintenance of events in working memory may impose a higher cost compared to coordinations. Although we did not observe significant ERP differences associated with pinging the agent vs. the patient during the delay, we did find that the ping appeared to dampen the ongoing sustained difference, suggesting a shift from sustained activity to activity silent mechanisms. These results suggest a new method by which ERPs can be used to elucidate the format of neural representation for events in working memory.

  PublisherOA PDFDOI

• A finite set of content-free pointers in visual working memory: MEG evidence

  bioRxiv (Cold Spring Harbor Laboratory) · 2024-07-26 · 2 citations

  preprintOpen accessSenior author

  Abstract Human visual working memory (VWM) is known to be capacity-limited, but the nature of this limit continues to be debated. Recent work has proposed that VWM is supported by a finite (∼ 3) set of content-free pointers, acting as stand-ins for individual objects and binding features together. Here, based on two visual working memory experiments (N=20 each) examining memory for simple and complex objects, we report a sustained MEG response over right posterior cortex whose magnitude tracks the core hypothesized properties of this system: load-dependent, capacity-limited and content-free. These results provide novel evidence for a finite set of content-free pointers underlying VWM.

  PublisherOA PDFDOI

• Exploring the effectiveness of reward-based learning strategies for second-language speech sounds

  Psychonomic Bulletin & Review · 2024-08-07 · 1 citations

  reviewOpen access
  PublisherOA PDFDOI

• Same set of visual pointers for biological and non-biological objects in working memory

  Visual Cognition · 2024-09-13 · 3 citations

  articleSenior author

  Data and script for "Same set of visual pointers for biological and non-biological objects in working memory", Xinchi Yu & Ellen Lau

  PublisherDOI

• Same set of visual pointers for biological and non-biological objects in working memory

  2024-01-06

  preprintOpen accessSenior author

  It has been well-established that visual working memory (VWM) is implemented by a limited set of ~ 3 pointers/indexicals, binding features together. Most of these studies used non-biological objects (e.g., shapes) as stimuli. However, two separate lines of recent work on infants and adults have prompted a proposal that biological objects recruit a separate set of visual pointers in working memory, distinct from the classical ~ 3 pointers for non-biological objects (the separate set hypothesis). The alternative possibility is that biological and non-biological objects share the same set of pointers (the same set hypothesis). In the current paper, we provide evidence supporting the same set hypothesis with a series of three experiments involving biological motions and complex shapes, as well as arguments reconciling evidence that seemed to support the separate set hypothesis.

  PublisherOA PDFDOI

• Shared neural computations for syntactic and morphological structures: evidence from Mandarin Chinese

  bioRxiv (Cold Spring Harbor Laboratory) · 2024-02-01 · 4 citations

  preprintOpen accessSenior author

  Abstract Although psycho-/neuro-linguistics has assumed a distinction between morphological and syntactic structure building as in traditional theoretical linguistics, this distinction has been increasingly challenged by theoretical linguists in recent years. Opposing a sharp, lexicalist distinction between morphology and syntax, non-lexicalist theories propose common morpho-syntactic structure building operations that cut across the realms of “morphology” and “syntax”, which are considered distinct territories in lexicalist theories. Taking advantage of two pairs of contrasts in Mandarin Chinese with desirable linguistic properties, namely compound vs. simplex nouns (the “morphology” contrast, differing in morphological structure complexity per lexicalist theories) and separable vs. inseparable verbs (the “syntax” contrast, differing in syntactic structure complexity per lexicalist theories), we report one of the first pieces of evidence for shared neural responses for morphological and syntactic structure complexity in language comprehension, supporting a non-lexicalist view where shared neural computations are employed across morpho-syntactic structure building. Specifically, we observed that the two contrasts both modulated neural responses in left anterior and centro-parietal electrodes in an a priori 275:400 ms time window, corroborated by topographical similarity analyses. These results serve as preliminary yet prima facie evidence towards shared neural computations across morphological and syntactic structure building in language comprehension.

  PublisherOA PDFDOI

Recent grants

• NIH Grant F32HD063221

  NIH · $99k · 2013

• Doctoral Dissertation Research: Morphosyntactic and Interpretive Dependency Formation in Agreement Attraction

  NSF · $6k · 2017–2018

• Sustained Activity Mechanisms for Neurally Encoding Linguistic Structure

  NSF · $387k · 2018–2024

Frequent coauthors

• Gina R. Kuperberg

  71 shared

• Kirsten Weber

  McMaster University

  65 shared

• Alexandre Gramfort

  50 shared

• Matti Hämäläinen

  Aalto University

  44 shared

• Colin Phillips

  25 shared

• Benjamin Stillerman

  New York University

  20 shared

• Ben Stillerman

  Athinoula A. Martinos Center for Biomedical Imaging

  12 shared

• Candida Ustine

  Medical College of Wisconsin

  12 shared

Labs

• CLIPPI