The Role of the Multiple-Demand System in Fluid Intelligence and Selective Attention: Insights from Lesion and Neuroimaging Studies

Abstract

Our ability to flexibly and rapidly coordinate our thoughts and actions in line with our goals is a defining feature of human cognition. A prominent proposal posits the multiple-demand (MD) system, a distributed network of predominantly frontoparietal regions, plays a critical role in this process. A growing body of evidence shows the MD system is consistently recruited in response to a wide range of cognitive demands and preferentially represents information currently relevant to behaviour. Yet our understanding of how these regions support complex, goal-directed behaviour is still incomplete. In this thesis, I further elucidate the role of the MD system in facilitating purposeful behaviour, focussing on its role in fluid intelligence and selective attention. Across three empirical studies, I draw on evidence from patients with focal brain lesions and healthy adults using neuropsychology, functional neuroimaging (functional magnetic resonance imaging (fMRI) and magnetoencephalography (MEG)), and multivariate pattern analysis techniques to investigate the role of the MD system in fluid intelligence and selective attention. In Chapter 2, I analysed a large neuropsychological dataset from 174 patients with focal brain lesions to identify the brain networks critical for fluid intelligence. For this, I quantified the extent of grey matter damage and white matter network disconnection in seven canonical resting-state brain networks, and related these measures to an estimate of individual fluid intelligence loss. The data suggested that of all resting-state networks, frontoparietal network, a resting-state network closely aligned to the MD system, had the most prominent role in fluid intelligence. Even in this relatively large patient sample, it was not possible to dissociate the contribution of frontoparietal network from a broader contribution of total lesion volume. However, I showed that damage to this network was predictive of fluid intelligence deficit even after accounting for damage to other networks. In Chapter 3, I turned to the role of the MD system in supporting selective attention. I used fMRI, multivariate pattern analysis techniques, and MEG-fMRI fusion to investigate what task information is prioritised in the MD system when attention is directed both in space (left/right) and to particular object features (colour/shape). For this, I collected and analysed fMRI data from healthy adults (N=30). I found that MD codes were highly tuned to the task at hand, only reflecting behaviourally relevant information. These codes were organised to reflect both the difficulty of the current discrimination and the details of the attended visual features. Using MEG-fMRI fusion, I then compared the data to an MEG dataset from participants performing the same task, finding broad agreement between modalities. These data emphasised the role of the MD system in selective prioritisation of attended information. In Chapter 4, I combined the focal lesion approach with MEG and multivariate pattern analysis to explore how the representation of task-related information during selective attention is affected by parietal lobe damage. In this exploratory case-series, I examined the dynamic neural prioritisation of relevant over irrelevant information at an individual level in six patients with focal parietal lesions and 10 age-, gender-, and education-matched controls. I found that patients with parietal lesions can ultimately prioritise relevant information in their neural activity, commensurate with their high behavioural accuracy. However, for some patients, slower reaction times were associated with subtle deficits in neural prioritisation. This aligns with prominent suggestions of a causal role of parietal MD regions in attentional prioritisation and demonstrates the feasibility of combining MEG and MVPA to study individual differences in the dynamic prioritisation of information after brain damage. In conclusion, I observed that damage to frontoparietal network predicts fluid intelligence deficit, that MD regions selectively code behaviourally relevant information, and that damage to some of this network can affect the degree to which behaviourally relevant information is prioritised in the brain. Taken together, this body of work suggests that MD regions, or frontoparietal network more broadly, play an important role in fluid intelligence and selective attention, processes fundamental to intelligent human behaviour. My work also showcases new methodological approaches and suggests interesting directions for future research into how the MD system gives rise to our extraordinary capacity for complex, flexible behaviour.