Funded Grants

Stability of mind in a dynamic brain: Neural principles of working memory for flexible human cognition

The ability to hold information in mind for short periods of time depends on working memory. Working memory provides the functional backbone to high-level cognition – by holding information in working memory, we are able to perform complex actions based on time-extended goals and contextual contingencies. Put simply, working memory frees action from direct stimulus dependency. My research explores the fundamental neural principles of working memory, and how they support flexible human behaviour.

The neuroscience of working memory faces a major challenge: brain activity is highly dynamic. At first glance these dynamics seem at odds with the very nature of working memory. How can we keep a stable thought in mind while brain activity is constantly changing? Indeed, some of the most influential models in neuroscience are built on the first-level intuition that stability of mind depends on stable brain activity. The standard account assumes that working memory is maintained by static patterns of neural activity, as if frozen in time to preserve a still-frame representation of the past. But new methods for measuring and analysing brain activity reveal a much more dynamic portrait: neural activity patterns are constantly changing, even when the cognitive state remains stable.

I am exploring a new theoretical framework to understand how working memory is implemented in a dynamic brain. According to a dynamic coding model, working memory is best understood as a temporary shift in how we process new information, rather than a representation of the past preserved in persistent activity. Based on insights from systems neuroscience, I propose that this temporary shift in coding properties is achieved by constructing new temporary pathways within the prefrontal cortex.

It is well-established that the brain’s wiring diagram adapts slowly over time. However connectivity is also modulated by short-lived neurophysiological changes. Transient patterns of effective connectivity are sometimes referred to as a ‘hidden’ state, because they are not directly observable using standard recording measures. For the same reason, neuroscience has been dominated by images of brain activity, even though we know from first principles that processing in the brain is defined by connectivity. Activity is simply easier to measure than connectivity, especially when measurement tools are limited by poor temporal resolution (e.g., fMRI).

Therefore, to test the role of dynamic coding in working memory, we have been developing new methods to infer the underlying connectivity states. Our research employs computational approaches to infer temporary patterns of connectivity during working memory using non-invasive whole-brain methods with real-time resolution (e.g., MEG). Zooming in to a more detailed level of analysis, we are also tracking memory-dependent changes in connectivity between individual neurons and local neural populations by teaming up with neurophysiologists and clinical neurosurgeons. This integrative approach is essential if we are to uncover fundamental neural principles of high-level human cognition.

Finally, a dynamic coding framework could also shed new light on individual differences in working memory. In particular, we hypothesise that working memory capacity is limited by encoding/retrieval rather than maintaining a stable activity state. This is important, because individual differences in working memory ability predict performance on laboratory tests of intelligence as well as real-world measures of academic progress. A better understanding of working memory will have broad implications for improving cognitive capacity in general.