Comprehension of computer code relies primarily on domain-general executive brain regions

Abstract

Computer programming is a novel cognitive tool that has transformed modern society. What cognitive and neural mechanisms support this skill? Here, we used functional magnetic resonance imaging to investigate two candidate brain systems: the multiple demand (MD) system, typically recruited during math, logic, problem solving, and executive tasks, and the language system, typically recruited during linguistic processing. We examined MD and language system responses to code written in Python, a text-based programming language (Experiment 1) and in ScratchJr, a graphical programming language (Experiment 2); for both, we contrasted responses to code problems with responses to content-matched sentence problems. We found that the MD system exhibited strong bilateral responses to code in both experiments, whereas the language system responded strongly to sentence problems, but weakly or not at all to code problems. Thus, the MD system supports the use of novel cognitive tools even when the input is structurally similar to natural language.

The neural bases of program comprehension: a coordinate-based fMRI meta-analysis

“…The third was on the right middle frontal gyrus (MFG) extending to the adjacent right superior and inferior frontal gyri. In contrast to several previous studies (Siegmund et al, 2014; Liu et al, 2020; Ivanova et al, 2020; Peitek et al, 2021), left parietal regions such as angular gyrus and inferior parietal lobule were not observed in our MKDA result at the significant threshold.…”

“…The indicator maps taken from Ikutani et al (2020) and Peitek et al (2021) had slightly different patterns but still showed Language and Semantic within the top-3 largest correlations. Two indicator maps taken from contrasts of [Program output estimation > Output estimation from sentence] in Ivanova et al (2020) were grouped by the unique cluster in our HCA dendrogram, having very small correlations to the language-related terms. The indicator map of ScrachJr in Ivanova et al (2020) showed a different pattern than others.…”

“…The other suggests that program comprehension is supported by a frontoparietal network associated with symbol systems (i.e., logic and math) or domain-general executive functions. Two recent studies present evidence supporting this hypothesis by directly comparing activation maps elicited by program comprehension and other cognitive tasks (Ivanova et al, 2020; Liu et al, 2020). A meta-analysis would contribute to evaluating these representative hypotheses and seeking cognitive similarities between program comprehension and other human skills by bridging neuroimaging evidence obtained from program comprehension studies and diverse cognitive/psychological experiments.…”

“…However, individual neuroimaging studies of program comprehension often provided inconsistent results. One of the central contradictions is that multiple studies argued cognitive similarity between program comprehension and natural language comprehension (Siegmund et al, 2014, 2017; Prat et al, 2020; Ikutani et al, 2020), while others emphasized the remarkable difference between them (Floyd et al, 2017; Krueger et al, 2020; Ivanova et al, 2020; Liu et al, 2020).…”

“…For example, Siegmund et al (2014, 2017) used Java code snippets and contrasted brain activity during program output estimations against syntax error searches. Ivanova et al (2020) investigated brain activities measured while subjects read short programs written in Python or ScratchJr and contrasted those activities with content-matched sentence problems written in natural language. Such variability in adopted tasks across individual experiments makes it difficult to evaluate the brain regions consistently activated during program comprehension.…”