This repository contains MATLAB scripts needed to generate the results in our paper "Neuroimaging evidence for a network sampling theory of human intelligence"
Contact
Please contact Dr. Eyal Soreq through Github for queries relating to this code.
Abstract
Here we use a combination of machine-learning and psychometrics to determine in an unbiased data-driven manner how individual variability and the factor structure of human intelligence relate to cortical dynamic-network connectomics. We report that 12 sub-tasks from an established intelligence test can be multi-way classified with high accuracy (74%, chance$\approx8.3%$) from the network states that they evoke. The tasks recruit densely overlapping but unique configurations of brain regions. We find evidence of this multivariate coding mechanism even within multiple-demand brain regions that are most commonly active across cognitive tasks. Critically, the network states of higher-performing individuals are classifiable with superior accuracy. Furthermore, the proximities of the tasks within behavioural-psychometric space reflect the similarity of their corresponding network states. These results demonstrate that the human brain uses a high-dimensional network-sampling mechanism to flexibly code for diverse cognitive tasks. Population variability in intelligence reflects the fidelity of expression of these task-optimised network states.
Reference