Representation Dynamics Underlying Human Information Compression in Visual Working Memory

Profoundly limited in both capacity and duration, short-term memory (STM) is a crucial indicator of high-level cognitive functioning. While traditionally regarded as separate from long-term memory (LTM), more recent proposals challenged this convention citing structures, such as the medial temporal lobe
(MTL), and more notably the hippocampus, which have been linked to both STM and LTM processing. However, the nature of limiting factors and interactions between LTM and STM are a subject of critical dispute.
A novel information theoretical posits that information in STM is represented as a code that is shaped by previous information in LTM. Specifically, knowledge about the statistical structure of the input allows for compression and therefore more efficient storage in STM. Understanding STM capacity in terms of
information rather than objects or features could redefine the limits of STM capacity, additionally providing an account of how information is transferred between STM and LTM. Still, the representational format of compressed information and the mechanism of LTM involvement are unclear. This project endeavours to address those limitations, utilising a mixed approach of behavioural, computational, and neuroimaging methods.