Tanya Wen, Ph.D.

I am a research scientist and with a cognitive neuroscience and psychology background currently working in the tech industry. With over 12 years of research experience (in vision and auditory perception science), I have utilized rigorous experimental design, inferential statistics, along with various tools to understand human behavior, including surveys, electrophysiology, eye-tracking, and virtual reality experiments. I am interested in applying my knowledge in real world settings, including UX research and product development.

I am currently an Applied Perception Scientist working at Meta Reality Labs Research - Audio (contract via Magnit). Previously, I worked as a Research Neuroscientist at the Naval Health Research Center. I earned a bachelor's degree in psychology and double-majored in life sciences at National Cheng Kung University in Taiwan between 2011-2015. Later, I earned my Ph.D. in medical science from the University of Cambridge from 2015-2019, under the supervision of Dr. John Duncan and Dr. Daniel Mitchell at the MRC Cognition and Brain Sciences Unit. From 2019-2022, I worked as a postdoctoral associate with Dr. Tobias Egner at the Center for Cognitive Neuroscience at Duke University.

Research

Visual Attention

I studied visual attention and am interested in how perception and cognition interact. I examined brain activity while participants viewed an illusion, which I find interesting as the percieved image is distorted despite it not being physically so. I also showed differences between perceptual (data-limited) difficulty and cognitive (resource-limited) difficulty, by increasing difficulty of visual discrimination, I found regions that are usually sensitive to cognitive difficulty to not accordingly increase activition. In another study, I used high-temporal resolution EEG/MEG to characterize the dynamic time-course of visual attention, from object processing to target recognition.

click to view details

click to view details

click to view details

Event Cognition and Memory

Task organization is often hierarchical, with smaller events forming larger temporal episodes, situation models, or semantic categories. For example, the goal of making a stew could involve smaller steps such as opening the fridge, wash vegetables, chop vegetables, and cook on the stove. These steps are different from another goal, like washing your face. A focus of my research is to understand how different brain regions represent these varying levels of information in human experience. I am also interested in how moving from one episode to another affects our temporal memory for the order in which our experiences occured.

click to view details

click to view details

click to view details

Cognitive Flexibility

Our world is ever-changing, therefore optimal regulation of task sets involves resolving a tradeoff between needing to implement the current task set and shielding it from distraction (cognitive stability) versus being ready to update (or switch) task sets in response to changing environmental contingencies (cognitive flexibility). The ability to dynamically adapt one’s flexibility level to suit varying environmental demands, i.e., meta-flexibility, facilitates optimal cognition. I am interested in how humans continuously monitor and integrate new information, and infer which task sets to use at a given time by observing environmental statistics. I am particularly interested in cognitive flexibility that translates to the real world.

click to view details

click to view details

click to view details

Skills

Matlab

Matlab is my first programming language, which I had self-taught myself during undergraduate, and have continued using thoughout my research career. I now have over 11 years of experience with Matlab. I have written many psychology experiments using Psychtoolbox, to present visual/auditory stimuli and collect repsonses from human participants. I use the SPM software package to analyze fMRI, EEG, and MEG data. I am also adept with using the EEGLAB toolbox to analyze EEG data. I also utlized packages such as LIBSVM, The Decoding Toolbox, and the RSA toolbox for multivariate analyses. I often write my own code to analyze both behavioral and neuroimaging data.

Web Development

This website is created using HTML and CSS, as well as the front-end framework Bootstrap to create a grid system layout and responsive site. In my research, I used JavaScript extensively, along with HTML and CSS to create online psychology experiments, where we present timed stimuli and collect responses. You can view a collection of my online experiments here. Many of my behavioral experiments are collected through Amazon Mechanical Turk. I have collected data from over 800 online participants for various experiments. For back-end, I wrote PHP scripts to send the data to our department server. I recently completed The Complete 2023 Web Development Bootcamp to learn more about full-stack web development, including Node.js, Express, RESTful APIs, PostgreSQL, MongoDB, and the React framework.

Python

I wrote scripts using Nipype to perform analyses on preprocessed fMRI data. I have also used Boto3, the AWS SDK for Python, in combination with my own code to accept and reject HITs on MTurk. In 2021, I participated in a programming club at Duke where we went through the lectures and homework in Data 100 by UC Berkeley. I have completed the Programming for Data Science with Python Udacity Nanodegree, which included SQL, Python, and Git. I have now used Python in projects including analyzing IMU data, wearable eye-tracking data, infrared camera videos, and Unity log outputs. I am experienced with python libraries including Numpy, Pandas, SciPy, Scikit-learn, Matplotlib, Seaborn, OpenCV, etc. I also experimented with pre-trained neural networks to analyze video data (e.g., YOLOv8 and Segment Anything Model).

R and RStan

I used R during my undergraduate course An Introduction to Statistical Learning: With Applications in R. I am familiar with R libraries such as lme4, tidyverse, and ggplot2. In my research, I used R and RStan to fit reinforcement learning models (Rescorla–Wagner model and its variants), as well as perform simulations and paramter recovery. I used hierarchical Bayesian modeling to estimate subject and population parameters.

Unity / C#

To align with my workplace's emphasis on realistic experiments, I learned a new programming language. I completed the "Unity Essentials Pathway" and "Unity Junior Programmer" certifications. Since then, I have built four different virtual reality tasks at work. I also learned how to integrate Phidgets into my Unity programs to make them interactive with hardware, collect sensor data, and enable timestamping of events.

Other

Microsoft Office (Word, Excel, Powerpoint), Adobe Illustrator, Adobe Photoshop, SPSS, Git and GitHub for version control, writing shell script to submit jobs to compute clusters, Docker and Singularity containers, graph theory metrics, Windows/Mac/Linux operating systems, Qualtrics online survey, pair programming, participant recruitment and communication, grant proposal writing, IRB (ethics committee) application, manuscript preparation, English and Mandarin Chinese writing/speaking.

Curriculum Vitae

View my academic CV!

View my Data Scientist resume!

View my UX Researcher resume!

Last updated - September 2024

Using functional magnetic resonance imaging to explore the flashed face distortion effect

Tanya Wen & Chun-Chia Kung

Abstract

The flashed face distortion (FFD) effect was coined by Tangen, Murphy, and Thompson (2011) in their second-place winner of the 2012 Best Illusion of the Year Contest. The FFD arises when people view various eye-aligned faces that are sequentially flashed in the visual periphery, and gradually the faces appear to be deformed and grotesque. In this functional magnetic resonance imaging (fMRI) study, participants were presented with four conditions: (a) one face pair changing only its illumination; (b) two and (c) three alternating face pairs; and (d) nonrepeated face pairs. Participants rated the magnitude of each illusion immediately after each block. Results showed that the receptive region of the early visual cortex (V1–V4), and category-selective areas such as the fusiform face area (FFA) and occipital face area (OFA), responded proportionally to the participants' rated FFD strength. A random-effects voxelwise analysis further revealed positively correlated areas (including the medial and superolateral frontal areas) and negatively correlated areas (including the precuneus, postcentral gyrus, right insula, and bilateral middle frontal gyri) with respect to participants' ratings. Time series correlations among these nine ROIs (four positive and five negative) indicated that most participants showed a clustering of the two separate ROI types. Exploratory factor analysis (EFA) also demonstrated the segregation of the positive and negative ROIs; additionally, two subsystems were identified within the negative ROIs. These results suggest that the FFD is mediated by at least two networks: one that is likely responsible for perception and another that is likely responsible for subjective feelings and engagement.

Click to view full paper

Response of the multiple-demand network during simple stimulus discriminations

Tanya Wen, Daniel J. Mitchell, John Duncan

Abstract

The multiple-demand (MD) network is sensitive to many aspects of task difficulty, including such factors as rule complexity, memory load, attentional switching and inhibition. Many accounts link MD activity to top-down task control, raising the question of response when performance is limited by the quality of sensory input, and indeed, some prior results suggest little effect of sensory manipulations. Here we examined judgments of motion direction, manipulating difficulty by either motion coherence or salience of irrelevant dots. We manipulated each difficulty type across six levels, from very easy to very hard, and additionally manipulated whether difficulty level was blocked, and thus known in advance, or randomized. Despite the very large manipulations employed, difficulty had little effect on MD activity, especially for the coherence manipulation. Contrasting with these small or absent effects, we observed the usual increase of MD activity with increased rule complexity. We suggest that, for simple sensory discriminations, it may be impossible to compensate for reduced stimulus information by increased top-down control.

Click to view full paper

The time-course of component processes of selective attention

Tanya Wen, John Duncan, Daniel J. Mitchell

Abstract

Attentional selection shapes human perception, enhancing relevant information, according to behavioral goals. While many studies have investigated individual neural signatures of attention, here we used multivariate decoding of electrophysiological brain responses (MEG/EEG) to track and compare multiple component processes of selective attention. Auditory cues instructed participants to select a particular visual target, embedded within a subsequent stream of displays. Combining single and multi-item displays with different types of distractors allowed multiple aspects of information content to be decoded, distinguishing distinct components of attention, as the selection process evolved. Although the task required comparison of items to an attentional “template” held in memory, signals consistent with such a template were largely undetectable throughout the preparatory period but re-emerged after presentation of a non-target choice display. Choice displays evoked strong neural representation of multiple target features, evolving over different timescales. We quantified five distinct processing operations with different time-courses. First, visual properties of the stimulus were strongly represented. Second, the candidate target was rapidly identified and localized in multi-item displays, providing the earliest evidence of modulation by behavioral relevance. Third, the identity of the target continued to be enhanced, relative to distractors. Fourth, only later was the behavioral significance of the target explicitly represented in single-item displays. Finally, if the target was not identified and search was to be resumed, then an attentional template was weakly reactivated. The observation that an item's behavioral relevance directs attention in multi-item displays prior to explicit representation of target/non-target status in single-item displays is consistent with two-stage models of attention.

Click to view full paper

Hierarchical representation of multistep tasks in multiple-demand and default mode networks

Tanya Wen, John Duncan, Daniel J. Mitchell

Abstract

Task episodes consist of sequences of steps that are performed to achieve a goal. We used fMRI to examine neural representation of task identity, component items, and sequential position, focusing on two major cortical systems—the multiple-demand (MD) and default mode networks (DMN). Human participants (20 males, 22 females) learned six tasks each consisting of four steps. Inside the scanner, participants were cued which task to perform and then sequentially identified the target item of each step in the correct order. Univariate time course analyses indicated that intra-episode progress was tracked by a tonically increasing global response, plus an increasing phasic step response specific to MD regions. Inter-episode boundaries evoked a widespread response at episode onset, plus a marked offset response specific to DMN regions. Representational similarity analysis (RSA) was used to examine representation of task identity and component steps. Both networks represented the content and position of individual steps, however the DMN preferentially represented task identity while the MD network preferentially represented step-level information. Thus, although both MD and DMN networks are sensitive to step-level and episode-level information in the context of hierarchical task performance, they exhibit dissociable profiles in terms of both temporal dynamics and representational content. The results suggest collaboration of multiple brain regions in control of multistep behavior, with MD regions particularly involved in processing the detail of individual steps, and DMN adding representation of broad task context.

Click to view full paper

Retrieval context determines whether event boundaries impair or enhance temporal order memory

Tanya Wen & Tobias Egner

Abstract

Meaningful changes in context create “event boundaries”, segmenting continuous experience into distinct episodes in memory. A foundational finding in this literature is that event boundaries impair memory for the temporal order of stimuli spanning a boundary compared to equally spaced stimuli within an event. This seems surprising in light of intuitions about memory in everyday life, where the order of within-event experiences (did I have coffee before the first bite of bagel?) often seems more difficult to recall than the order of events per se (did I have breakfast or do the dishes first?). Here, we aimed to resolve this discrepancy by manipulating whether stimuli carried information about their encoding context during retrieval, as they often do in everyday life (e.g., bagel-breakfast). In Experiments 1 and 2, we show that stimuli inherently associated with a unique encoding context produce a “flipped” order memory effect, whereby temporal memory was superior for cross-boundary than within-event item pairs. In Experiments 3 and 4, we added context information at retrieval to a standard laboratory event memory protocol where stimuli were encoded in the presence of arbitrary context cues (colored frames). We found that whether temporal order memory for cross-boundary stimuli was enhanced or impaired relative to within-event items depended on whether the context was present or absent during the memory test. Taken together, we demonstrate that the effect of event boundaries on temporal memory is malleable, and determined by the availability of context information at retrieval.

Click to view full paper

The functional convergence and heterogeneity of social, episodic, and self-referential thought in the default mode network

Tanya Wen, Daniel J. Mitchell, John Duncan

Abstract

The default mode network (DMN) is engaged in a variety of cognitive settings, including social, semantic, temporal, spatial, and self-related tasks. Andrews-Hanna et al. (2010; Andrews-Hanna 2012) proposed that the DMN consists of three distinct functional–anatomical subsystems—a dorsal medial prefrontal cortex (dMPFC) subsystem that supports social cognition; a medial temporal lobe (MTL) subsystem that contributes to memory-based scene construction; and a set of midline core hubs that are especially involved in processing self-referential information. We examined activity in the DMN subsystems during six different tasks: 1) theory of mind, 2) moral dilemmas, 3) autobiographical memory, 4) spatial navigation, 5) self/other adjective judgment, and 6) a rest condition. At a broad level, we observed similar whole-brain activity maps for the six contrasts, and some response to every contrast in each of the three subsystems. In more detail, both univariate analysis and multivariate activity patterns showed partial functional separation, especially between dMPFC and MTL subsystems, though with less support for common activity across the midline core. Integrating social, spatial, self-related, and other aspects of a cognitive situation or episode, multiple components of the DMN may work closely together to provide the broad context for current mental activity.

Click to view full paper

Transfer of learned cognitive flexibility to novel stimuli and task sets

Tanya Wen & Tobias Egner

Abstract

Adaptive behavior requires learning about the structure of the environment to derive optimal action policies, and previous studies have documented transfer of such structural knowledge to bias choices in new environments. Here, we asked whether people could also acquire and transfer more abstract knowledge across different task environments, in particular, expectations about demands on cognitive control. Over three experiments, participants performed a probabilistic card-sorting task in environments of either a low or high volatility of task rule changes (requiring low or high cognitive flexibility) before transitioning to a medium-volatility environment. Using reinforcement learning modeling, we consistently found that previous exposure to high task rule volatility led to faster adaptation to rule changes in the subsequent transfer phase. This transfer of expectations about demands on cognitive flexibility was both task- (Experiment 2) and stimulus- (Experiment 3) independent, thus demonstrating the formation and generalization of environmental structure knowledge to guide cognitive control.

Click to view full paper

Connectivity patterns in cognitive control networks predict naturalistic multitasking ability

Tanya Wen, De-Cyuan Liu, Shulan Hsieh

Abstract

Multitasking is a fundamental aspect of everyday life activities. To achieve a complex, multi-component goal, the tasks must be subdivided into sub-tasks and component steps, a critical function of prefrontal networks. The prefrontal cortex is considered to be organized in a cascade of executive processes from the sensorimotor to anterior prefrontal cortex, which includes execution of specific goal-directed action, to encoding and maintaining task rules, and finally monitoring distal goals. In the current study, we used a virtual multitasking paradigm to tap into real-world performance and relate it to each individual's resting-state functional connectivity in fMRI. While did not find any correlation between global connectivity of any of the major networks with multitasking ability, global connectivity of the lateral prefrontal cortex (LPFC) was predictive of multitasking ability. Further analysis showed that multivariate connectivity patterns within the sensorimotor network (SMN), and between-network connectivity of the frontoparietal network (FPN) and dorsal attention network (DAN), predicted individual multitasking ability and could be generalized to novel individuals. Together, these results support previous research that prefrontal networks underlie multitasking abilities and show that connectivity patterns in the cascade of prefrontal networks may explain individual differences in performance.

Click to view full paper