Project Outline

Working memory (WM) describes the ability to maintain and manipulate information that is no longer available in the environment. It provides a flexible mental workspace that scaffolds most higher cognitive functions. However, despite its fundamental role, little consensus exists about the brain mechanisms that underpin WM.

Traditionally, WM has been thought to rely on active maintenance of sensory representations, implemented via persistent neocortical activity. However, brain activity often returns to baseline levels during memory delay periods, suggesting that active maintenance is not essential. Consequently, new theories have been developed proposing that WM could be achieved via synaptic weight changes that are not reflected in brain activity but shape how the brain processes new input. In support of this view, recent studies have shown that impulse stimuli can reinstate otherwise no-longer measurable memory signals (similar to active sensing, where the contours of a hidden structure are inferred from changes in “echoes” evoked by a constant impulse).

An alternative explanation for such activity silent memory maintenance is the recruitment of episodic memory systems. From this perspective, maintenance does not rely on short-lived traces in cortical areas processing stimulus features, but on durable traces in the hippocampus. Distinguishing between these accounts will have great impact upon longstanding debates on the architecture of memory systems and could also open the door toward new translational research on sources of memory deficits (e.g., in ageing). At present, however, it is impossible to adjudicate between these accounts, as the neural sources of reactivated memory signals remain unknown.

Research Objectives

This PhD project will leverage multimodal brain recordings to fill this gap and reveal the neural sources of reactivated memories. Human subjects will perform WM tasks, while their brain activity is measured with simultaneous EEG-fMRI recordings, and high-contrast visual impulses are presented to reactivate activity-silent memory signals. In previous studies, we have established that across-trial variance in EEG impulse responses tracks changes in memory precision. Here, we will link this variance with BOLD signals using advanced signal processing and machine learning approaches to reveal brain areas whose activity states predict the magnitude of memory reactivation.

The integration of memory signals in space and time will close a fundamental gap in theories of human memory and provide a rich testbed for translational research. Following the initial proof of principle, the candidate will have the opportunity to develop this line of research into new directions, e.g., by testing if changes in task design can shift the neural sources of memory reactivation or by applying the paradigm to the study of memory deficits.

Research team and centre

The project brings together expertise in memory research, neural pattern analyses (Dr Paul Muhle-Karbe), multimodal brain imaging (Dr Karen Mullinger), advanced signal processing, and translational research (Prof Andrew Bagshaw). The student will receive in-depth training in concepts from Psychology, Neuroscience, and Machine Learning, allowing them to develop a wide range of valuable skills for a career in academia or data science. The training will moreover benefit from the vibrant research community and world-class infrastructure at the Centre for Human Brain Health in Birmingham.

Contact

Interested candidates should contact Dr Paul Muhle-Karbe ( ) for informal discussion before submitting an application.

References

Beukers, A. O., Buschman, T. J., Cohen, J. D., & Norman, K. A. (2021). Is activity silent working memory simply episodic memory?. Trends in Cognitive Sciences, 25(4), 284-293.

Muhle-Karbe, P. S., Myers, N. E., & Stokes, M. G. (2021). A hierarchy of functional states in working memory. Journal of Neuroscience, 41(20), 4461-4475.

Techniques

• experimental task design (retrospective cuing task with external impulse stimuli)
• programming (MATLAB/R/Python)
• computational modelling
• simultaneous EEG-fMRI recordings