Back

RoSe-BaL: A neuroanatomically plausible model of routine action sequencing

Lewis, J.; Gurney, K.

2025-03-10 neuroscience
10.1101/2025.03.06.641869 bioRxiv
Show abstract

The performance of routine action sequences constitutes a significant proportion of human behaviour, and has received much attention in the cognitive psychology literature. However, a neuroanatomically plausible explanation of the cognitive processes underlying this routine sequential behaviour has hitherto remained elusive. This is despite wide acceptance that an established hierarchy of interconnected basal-ganglia thalamocortical (BGTC) loops appear to be heavily involved in organising and executing context-sensitive routine action. Here, we build on existing computational models of action selection in the basal ganglia to develop the 'Routine-Sequencing Basal-ganglia Loops' (RoSe-BaL) model of multiple basal-ganglia thalamocortical loops occupying associative and motor territories of the brain. We demonstrate this model in the sequential routine tasks of tea- and coffee-making. This model incorporates a novel hypothesis of the nature of the organisation of cognitive information in associative territories of the BGTC hierarchy, which we term the 'BG subset-selection hypothesis'. Erroneous behaviour made by the model under conditions of disruption shows similar trends to those observed in previous studies of action slips and action disorganisation syndrome. This finding provides support for the previously untested hypothesis that damaged temporal order knowledge and action schemas underlie many of the errors typically performed by humans. The neuroanatomical grounding of the model naturally reconciles two influential competing computational models of routine behaviour. Finally, we propose a novel interpretation of the widespread competitive queueing account, which we term 'suitability queueing'. This is shown to be consistent with existing cognitive, neurophysiological and neuroanatomical findings.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.