Abstract/Summary

Cognitive flexibility enables individuals to respond adaptively to an ever-changing world. Neurally, flexibility is underpinned by involvement from across the cerebrum, and there is evidence from animal and human neuroscience suggesting that integration of cortical and thalamic signals in the striatum is necessary for appropriate behavioural control. A commonly used assay of flexibility is reversal learning, an associative learning task with high inter-species translatability. Evidence from animal literature has clearly defined the importance of the striatal cholinergic system in regulating striatal activity and output from the basal ganglia, and there is nascent evidence suggesting this system operates in a similar way in humans. However, there is a need to further disentangle the role of cortical, striatal, and thalamic regions during reversal learning in humans to better understand how the system works, and whether it has heterogeneous functionality in different contexts. Furthermore, as studying these processes is not trivial, further methodological work is required to enable us to understand the system. In chapter two we systematically assess an automated parcellation technique for identifying specific thalamic nuclei. Despite generally being treated as a homologous structure in neuroimaging work, nuclei within the thalamus have dissociable roles, and have diverse contributions to cognitive functioning, including reversal learning. We found mixed efficacy for segmentations across the thalamus, with some regions being more accurately defined relative to a “gold standard” atlas than others. Crucially, we find that the centromedian and parafascicular nuclei, which have an important role in reversal learning, are clearly defined and have little overlap with contiguous regions. These results show we can use this automated parcellation technique to identify specific thalamic nuclei that are relevant for cognitive flexibility and use these parcellations to study functionally relevant processes. Recent work has demonstrated that the functional relevance of the striatal cholinergic system can be studied in vivo using magnetic resonance spectroscopy by separating the peaks of different metabolites. But this non-conventional approach has not yet been widely adopted, and work is needed to determine its reliability. Chapter three presents test-retest reliability data on the use of magnetic resonance spectroscopy to study cholinergic activity in the striatum and cortex. We find measures of choline containing compounds are highly correlated when peaks are separated and when they are not. Across time we find that choline concentrations are relatively inconsistent, and that this was due to changes in the functionally relevant metabolite choline. Conversely, metabolites that we think are not functionally relevant were stable over time. We believe these differences may underly differences in acetylcholine function over time and may explain some intra-individual behavioural variability. In chapter four we use functional magnetic resonance imaging and psychophysiological interaction analysis to study corticostriatal and thalamostriatal connectivity during serial reversal learning. Functional connectivity between the centromedian-parafascicular nuclei of the thalamus and the associative dorsal striatum, and between the lateral-orbitofrontal cortex and the associative dorsal striatum was related to processing feedback during reversal learning. Specifically, thalamostriatal connectivity was found across the task, and may reflect a general error signal used to identify potential changes in context. Conversely, corticostriatal connectivity was found to be specific to when behaviour changed and suggests this may be a mechanism for the implementing adaptive change. We also show findings from exploratory work that may explain further how the cortex supports flexibility during reversal learning. Lastly, we used magnetic resonance spectroscopy to investigate whether the state of the cholinergic system at rest is related to reversal learning performance and latent measures of behaviour using computational modelling. Choline concentrations at rest showed significant functional relevance to our measures of reversal learning. More specifically, we found that errors during reversal learning, and learning rates for positive and negative prediction errors, explained significant variance in choline. However, the relationship between choline levels and task performance presented here differ from previous work which instead used a multi-alternative reversal learning task, and suggests that the striatal cholinergic system may have dissociable roles in different contexts. Overall, we show that the striatum, its cholinergic interneuron system, and its afferent projections from the cortex and thalamus, are associated with performance during serial reversal learning. Moreover, these findings suggest that the system may operate in separable ways in different contexts which may be dependent on internal representations of task structure.