Deciphering cortical number coding from human brain activity patterns.

Area of Science:

• Cognitive Neuroscience
• Neuroimaging
• Computational Neuroscience

Background:

• The human parietal cortex is implicated in numerical processing, with intraparietal areas housing neurons tuned to numerosity.
• While neuroimaging has defined areas involved in numerical tasks, direct evidence of individual number coding via spatial patterns has been lacking.

Purpose of the Study:

• To demonstrate direct evidence of individual number coding by spatial patterns in the human brain.
• To investigate the format invariance of numerical representations using multivariate pattern recognition.

Main Methods:

• Utilized high-resolution functional magnetic resonance imaging (fMRI) data.
• Applied multivariate pattern recognition techniques to decode fine-scale signals evoked by different individual numbers.
• Examined activation patterns for both symbolic (digits) and nonsymbolic (dot sets) number formats.

Main Results:

• Accurately discriminated and generalized parietal activation patterns for individual numerosities across stimulus parameter changes.
• Found distinct patterns for symbolic and nonsymbolic number formats, with dot sets decoded more accurately than digits.
• Demonstrated that nonsymbolic numerosity could be predicted from digit-evoked patterns, but not vice versa, suggesting an orderly layout of number representations.

Conclusions:

• Findings support partial format invariance of individual number codes, consistent with computational models postulating broader tuning for nonsymbolic numbers.
• Illustrates the potential of fMRI pattern recognition for understanding detailed representational formats within semantic categories.
• Highlights the applicability of these methods beyond sensory cortical areas.