Human intelligence relates to neural measures of cognitive map formation

Psychometric research on intelligence consistently identifies latent factors underlying performance correlations across cognitive tasks, with one general factor (g) explaining most variance and predicting general life outcomes. Their biological basis is yet unresolved, in particular with regard to the neural information processing mechanisms that may underlie intelligence. Here we test the hypothesis that interindividual differences in relational processing, supported by cognitive maps in the hippocampus, are related to fluid intelligence (gf), a statistical approximation of the general factor. Using standardized cognitive tests and fMRI of different mnemonic tasks, we demonstrate a positive correlation between gf and map-like encoding of object positions learned piecemeal. The behavioral and neural geometry of object representations in the lower gf range was less consistent with any two-dimensional representation, congruent with a lack of relational integration at encoding. The specificity of the link between cognitive maps and intelligence to relational processing was further supported by comparisons to non-relational mnemonic processing in the hippocampus during item recognition. These findings offer first empirical support for a link between neural mechanisms related to relational reasoning and general cognitive performance and probe the presumed relevance of hippocampal coding properties for cognition more broadly.