What I study
Where the eyes go, what the brain does, and what people report feeling do not agree with each other. Most of my work is about that disagreement.
Looking is not recognising
Fearful and neutral faces produce a first saccade to the eyes. Happy faces do not, because the mouth is already the informative feature. Invert the face and the bias reverses, which rules out low-level brightness and implicates facial configuration. None of this improves recognition: accuracy stays near ceiling regardless of which feature is fixated, at 150 ms and at 50 ms (Yang & Gamer, Scientific Reports, 2025). A fixation map records where attention went, not what was perceived. That is an uncomfortable conclusion when eye-tracking is half of your method.
Earlier work asked what the visual system needs before it can categorise something as a face at all. Rapid categorisation depends on a narrow band of spatial contrast, and the contrast statistics of natural images fall within that band (Liu-Shuang et al., eNeuro, 2022). When the evidence in the face is weak and the decision becomes effortful, drift-diffusion modelling and the ERP describe the same process from two directions (Yang et al., Frontiers in Human Neuroscience, 2020). This began with my dissertation at Université Paris-Saclay in 2018 and continues.
Gaze, and being left out
Exclusion is what is measured. Gaze is what changes it. In a Cyberball game where ball possession drops from 33% to 23%, the P3 to receiving the ball increases by about 2 µV when the co-players look directly at the participant, and does not change when they look away. The self-report measures give the opposite ordering: the averted-gaze group reported worse mood, and need-threat scores did not differ between the groups (Yang, Fang & Niedeggen, Scientific Reports, 2025). The electrophysiology and the questionnaire disagree about which form of exclusion is worse. That disagreement is the result.
The wider line of work concerns order effects. Prior exposure to one social threat alters the response to a second one (Cognitive, Affective, & Behavioral Neuroscience, 2024). Losing control before exclusion changes the ERP signature (Brain Sciences, 2022). Overinclusion after a threat produces effects of its own (Psychophysiology, 2025). From 2027 this becomes a funded project on how specific the gaze-mediated component of exclusion processing is.
How much of a result survives its analysis
One EEG dataset, 168 independent analyses, conclusions that do not converge (EEGManyPipelines, Journal of Cognitive Neuroscience, 2024). The full account of that analytic flexibility, and what it does to what gets published, is written up in Many pipelines, one dataset (2026). I work on the reporting side of that problem. ARTEM-IS is an agreed template for describing ERP methodology so that another lab can reuse it (Psychophysiology, 2025), developed alongside the EEG/ERP preregistration template. bidsMReye recovers gaze from MR images, which provides eye-tracking data for datasets recorded without an eye tracker.
Open science, mostly as community work
The piece I point people to first is a guide to open and reproducible neuroimaging written for early career researchers (Bhagwat, Urchs, Poline & Yang, Imaging Neuroscience, 2025). A decade of open tools, platforms and standards has produced more choice than guidance. That paper is the practical version of everything on this page: what to adopt, in what order, when you are the person who has to do it and nobody has given you extra time for it.
By count this is now the largest part of my record: the OHBM Brainhack and Open Science Room proceedings, with the 2023 hackathon paper under my lead; the Chinese Open Science Network; ARIADNE; a community-sourced glossary of open-scholarship terms (Nature Human Behaviour, 2022); a catalogue of questionable research practices; principles for generative AI in research (AI and Ethics, 2025); an assessment, for a German-speaking readership, of what AI can and cannot do seen from biological psychology and neuropsychology (Psychologische Rundschau, 2026); and an argument for sustainable neuroscience through open science (Nature Human Behaviour, 2026). Committee work that produces papers is still committee work, and I would rather say so than present it as a research programme.