The neural correlates of logical-mathematical symbol systems processing resemble those of spatial cognition more than language processing

The ability to use logical-mathematical symbols (LMS), encompassing tasks such as calculation, reasoning, and programming, is special to humans with recent emergence. LMS processing was suggested to build upon fundamental cognitive systems through neuronal recycling, with natural language processing and spatial cognition as key candidates. This study used meta-analyses and synthesized neural maps of representative LMS tasks, including reasoning, calculation, and mental programming, to compare their neural correlates with those of the two systems. Our results revealed greater activation overlap and multivariate similarity between LMS and spatial cognition than with language processing. Hierarchical clustering further indicated that LMS tasks were indistinguishable from spatial tasks at the neural level, suggesting an inherent connection. Our findings support the hypothesis that spatial cognition is the basis of LMS processing, shedding light on the logical reasoning limitations of large language models, particularly those lacking explicit spatial representations.