Computer code comprehension shares neural resources with formal logical inference in the fronto-parietal network

Area of Science:

• Cognitive Neuroscience
• Neuroimaging of Programming
• Computational Linguistics

Background:

• The cognitive and neural underpinnings of programming code comprehension remain largely unexplored.
• Existing theories propose code comprehension may reuse neural mechanisms for natural language or rely on networks for other symbolic systems like logic and math.
• Understanding these mechanisms is vital given programming's societal importance.

Purpose of the Study:

• To investigate the cognitive and neural bases of code comprehension in expert programmers.
• To determine whether code comprehension recruits language-specific, general symbolic, or executive control networks.
• To compare the neural overlap between code comprehension and formal logic, symbolic math, and executive functions.

Main Methods:

• Functional magnetic resonance imaging (fMRI) was used to scan expert programmers (average 11 years experience) during code comprehension and memory control tasks.
• Participants also completed tasks assessing formal logic, symbolic math, executive control, and language processing (language localizer).
• Activity patterns in the fronto-parietal network were analyzed for code comprehension and compared with other task performances.

Main Results:

• Code comprehension activated a left-lateralized fronto-parietal network.
• Neural activity patterns within this network differentiated between 'for' loops and 'if' conditional statements.
• Significant overlap was observed between code comprehension and formal logic networks, with lesser overlap with symbolic math; overlap with executive control and language was minimal, though individual language and code laterality covaried.

Conclusions:

• Code comprehension is supported by a distinct fronto-parietal cortical network, shared with other cultural symbol systems like formal logic and mathematics.
• This network appears specialized for processing abstract symbolic information, rather than primarily recycling language mechanisms.
• The findings highlight a unique neural basis for programming, emphasizing its role as a sophisticated cultural invention.