Decoding Ruminative Inner Speech with Multimodal Brain-and-Articulatory Machine Learning

About the Project

Applications are invited for a fully funded PhD studentship in the Department of Psychology, Lancaster University, hosted by the Data Science and AI Institute (DSAIL) and funded by an EPSRC Doctoral Landscape Award. The project develops a wearable, AI-driven brain-computer interface that detects spontaneous inner speech - the silent verbal stream involved in self-talk, rumination, and intrusive thought - and classifies its emotional content from a combination of brain activity (EEG) and tongue movement (ultrasound tongue imaging).

Supervisory team: Dr Bo Yao (Psychology, Lancaster), Professor Hossein Rahmani (Computing and Communications, Lancaster), Dr Sam Kirkham (Linguistics and English Language, Lancaster).

PhD project description: Inner speech is functionally involved in memory, planning, and self-regulation. It also manifests as verbal rumination, a feature of depression and anxiety, and is implicated in voice-hearing experiences in psychosis. Existing approaches mostly use silent-speech paradigms in which inner speech is produced on cue, but ruminative and intrusive thoughts arise spontaneously, so capturing them requires methods that work in naturalistic conditions.

This PhD will develop such methods, combining wearable EEG with portable ultrasound imaging of the tongue. The student will work at the intersection of cognitive/computational neuroscience, machine learning, and articulatory phonetics. They will translate an existing inner-speech classifier from research-grade laboratory EEG to a wearable form factor, and develop deep-learning models that identify spontaneous inner-speech episodes and characterise their emotional valence. The project will also establish what is achievable with wearable EEG alone, a critical step towards future clinical applications.

Training and skills: Supervision is interdisciplinary, spanning cognitive neuroscience, machine learning, and articulatory phonetics. The student will develop hands-on skills in experimental design for EEG, ultrasound, and behavioural data collection; deep-learning model development; multimodal signal processing and fusion; articulatory speech analysis; and reproducible research practice with open-source software release.

Application details: The successful candidate will hold, or expect to obtain, at least an upper-second-class honours degree, and ideally a Masters, in computational neuroscience, computer science, or a closely related quantitative discipline. Strong Python programming skills and demonstrable machine-learning experience are essential.

Dates

• Application deadline: 30 June 2026
• Provisional interview date: week beginning 20 July 2026
• Studentship start date: 1 October 2026