Computer code comprehension shares neural resources with formal logical inference in the fronto-parietal network
- Johns Hopkins University, United States
Figures
Figure 1 with 1 supplement
Whole-brain contrasts.
Areas shown are p<0.05 cluster-corrected p-values, with intensity (both warm and cold colors) representing uncorrected vertex-wise probability. In the maps for each localizer contrast, both warm and cold colors indicate activated vertices in the contrast, with the cold color labelling the overlap with the code contrast.
Figure 1—figure supplement 1
Mean accuracy for (a) code comprehension and (b) localizer tasks.
Response time for (c) code comprehension and (d) localizer tasks. Error bars are mean ± SEM. *p<0.05. **p<0.01. ***p<0.001.
Figure 2
MVPA decoding accuracy in ROIs revealed by the code contrast.
(a) The four search spaces (IPS, pMTG, PFC, OCC in the left hemisphere) within which functional ROIs were defined for the MVPA. (b) The MVPA decoding accuracy in the four ROIs. Error bars are mean ± SEM. *p<0.05. ***p<0.001.
Figure 3
Overlap between the brain map revealed by the code contrast and each of the brain maps revealed by the localizer contrasts.
(a) Brain map with the activated regions in the five contrasts reported in Figure 1 overlain. The language network is shown in transparent blue, math in transparent red, and logic in transparent green. The regions activated in the MSIT contrast are enclosed in black outlines, and the code-responsive regions are enclosed in yellow outlines. (b) Cosine similarity between code contrast and each localizer contrast, in each hemisphere. Each dot represents the data from one participant. The dotted line on each bar indicates the null similarity between code contrast and the given localizer contrast. The yellow dashed line in each hemisphere indicates the empirical upper bound of the cosine similarity, the similarity between code comprehension and itself, averaged across participants. Error bars are mean ± SEM. *p<0.05. **p<0.01. ***p<0.001.
Figure 4
The lateralization index of the code contrast and the localizer contrasts.
(a) The lateralization index of the code contrast and the localizer contrasts. Each white dot stands for one participant, and the enlarged dots represent the mean values. (b) The lateralization indices of code contrast and language contrast are highly correlated.
Figure 5
The experiment design.
The FAKE function (bottom row) in this figure is created by scrambling the words and symbols in each line of the REAL function (top row). Note that for the purpose of illustration, the relative font size of the text in each screen shown in this figure is larger than what the participants saw during the actual MRI scan.
Additional files
-
Supplementary file 1
All the functions included in the fMRI study.
- https://cdn.elifesciences.org/articles/59340/elife-59340-supp1-v1.docx
-
Supplementary file 2
Activated clusters in each contrast.
- https://cdn.elifesciences.org/articles/59340/elife-59340-supp2-v1.docx
-
Supplementary file 3
FDR-corrected p-values for the post-hoc paired t-tests among the overlap between code contrast and the localizer contrasts.
- https://cdn.elifesciences.org/articles/59340/elife-59340-supp3-v1.docx
-
Transparent reporting form
- https://cdn.elifesciences.org/articles/59340/elife-59340-transrepform-v1.pdf
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Computer code comprehension shares neural resources with formal logical inference in the fronto-parietal network
eLife 9:e59340.
https://doi.org/10.7554/eLife.59340
