Biology subjects

Robinson, C. N.

Publications and source records attributed to Robinson, C. N..

3 recordsLinked to original sources

bioRxiv2026

Complementary frontoparietal and corticothalamic contributions to relational reasoning

Complex reasoning requires frontal, parietal, and thalamic systems to manage increasing relational demands, yet the underlying circuit mechanisms remain unclear. Here, we combined EEG with biologically grounded corticothalamic neural field modelling while participants solved relational problems of graded complexity. Successful reasoning was associated with dissociable frontoparietal dynamics. Frontal regions showed increased theta-band power, whereas parietal regions showed reduced alpha- and beta-band power. Neural field modelling linked these dynamics to complementary complexity-dependent circuit adaptations. Parietal regions showed modulation of intracortical and corticothalamic gains, intrathalamic inhibition, prolonged loop delays, and faster synaptic filtering, whereas frontal regions primarily adjusted intracortical gains in a manner consistent with maintaining local excitatory-inhibitory balance and supporting longer temporal integration windows. Together, these empirical and model-derived insights demonstrate that problem-solving under increasing reasoning demand relies not simply on greater frontoparietal engagement, but on additional, region-specific reconfigurations of cortical and corticothalamic circuits.

neuroscience↗

bioRxiv2025

Aligning transformer circuit mechanisms to neural representations in relational reasoning

Relational reasoning--the capacity to understand how elements relate to one another--is a defining feature of human intelligence, yet its computational basis remains unclear. Here, we combined human neuroimaging (7T fMRI) with artificial neural network modeling to identify circuit-level analogues of human reasoning computations. Using the Latin Square Task, we found that humans and transformers were able to generalize the task reliably, while standard architectures used in cognitive neuroscience could not. Analysing the transformer components revealed distinct computational roles: positional encoding captured the spatial structure of the task and aligned with representations in visual cortex, whereas attention encoded relational structure and mapped onto frontoparietal and default-mode networks. Attention weights tracked the relational complexity of the task, providing a computational analogue of reasoning demands. These results advance knowledge on the core algorithmic computations supporting complex reasoning, highlighting attention-based architectures as powerful models for investigating the neural and computational basis of higher cognition.

neuroscience↗

bioRxiv2024

Temporally delayed representations in alpha and beta rhythms in higher-order cortical networks track increasing relational integration demands

Relational reasoning is the ability to infer and understand the relations between multiple elements. In humans, this ability supports higher cognitive functions and is linked to fluid intelligence. Relational complexity (RC) is a cognitive framework that offers a generalisable method for classifying the complexity of reasoning problems. To date, increased RC has been linked to static patterns of brain activity supported by the frontoparietal system, but limited work has assessed the multivariate spatiotemporal dynamics that code for RC. To address this, we conducted representational similarity analysis in two independent neuroimaging datasets (Dataset 1 fMRI, n=40; Dataset 2 EEG, n=45), where brain activity was recorded while participants completed a visuospatial reasoning task that included different levels of RC (Latin Square Task). Our findings revealed that, spatially, RC representations were widespread, peaking in brain networks associated with higher-order cognition (frontoparietal, dorsal-attention, and cingulo-opercular). Temporally, RC was represented in the 2.5 - 4.1 seconds post-stimuli window and emerged in the alpha and beta frequency range. Finally, multimodal fusion analysis demonstrated that shared variability within EEG-fMRI signals within higher-order cortical networks were better explained by the theorised RC model, relative to a model of cognitive effort (CE). Altogether, the results further our understanding of the neural representations supporting relational processing, highlight the spatially distributed coding of RC and CE across cortical networks, and emphasise the importance of late-stage, frequency-specific neural dynamics in resolving RC.

neuroscience↗