Article Figures & Data
Figures
- Figure 1.
Computation of text correlations between the text that was read and the content report subsequently written by the participants. A, Schematic procedure for text comparison. Each text and content report were translated into a semantic vector consisting of intermediate semantic features computed from word-collocation in a large text database. The correlation of the vectors represents the semantic similarity between the two texts rather than the appearance of specific keywords (see text for details). B, Performance of EEG and fMRI participants. Blue and red plots indicate the participants in the EEG and fMRI experiments, respectively. Horizontal and vertical axes denote the performance (the entire text correlation between the text read and the content report) and the surrogate text correlation, respectively. The fact that plots lie below the diagonal line shows that the content reports specifically reflected the texts read. The data from two participants, for whom the plots appeared along the diagonal line and whose correlations were below 0.6, were excluded from the analysis.
- Figure 2.
Results of fixation-related EEG analysis. A, FRPs of grand averaged signal over all electrodes. The upper plot shows the FRPs. The red and blue plots show FRPs during higher and lower text correlation between the read text and content reports, respectively. The “higher” and “lower” regression coefficients were defined by a median split of the regression coefficients. The gray plots show uncorrected FRP. Horizontal bars indicate time points showing significant effects (p < 0.05). Multiple comparisons were corrected by the nonparametric clustering permutation test (Maris and Oostenveld, 2007). The horizontal axis represents the time from fixation onset. The lower plot shows horizontal and vertical EOGs (HEOG and VEOG). B, Topographical head maps of regression coefficients of the sentence-length text correlation in the time periods 0.1–0.2 and 0.4–0.5 s. C, Time-frequency maps of raw (ocular-artifact-uncorrected) and corrected EEG power. The horizontal axis represents the time from fixation onset; vertical axis indicates frequency on a logarithmic scale. D, Time-frequency map of regression coefficients averaged over all channels for the paragraph-length text correlations between the text read and the content report. The values were masked by a cluster-based statistical value (p < 0.05). E, Topographical head maps of regression coefficients of the sentence-length and paragraph-length text correlations in θ-band (4–8 Hz) EEG. The maps of regression coefficients were averaged over the time periods 0–0.4 and 0.4–0.8 s.
- Figure 3.
Brain regions showing a significant decrease in BOLD activity during reading of texts that were highly correlated with the content reports subsequently written by the participants. The results of sentence-length text correlation are shown on the top and in the middle, and those relevant to paragraph-length text correlation are shown on the bottom. There were no significant regions showing positive subsequent memory effect (i.e., BOLD activity increasing during reading of texts that were highly correlated with the content reports).
Tables
- Table 1.
Brain regions showing a significant decrease in BOLD activity correlated with sentence-length and paragraph-length text correlations
Anatomical region MNI coordinates (mm) t value x y z Sentence length (NSME) R-insula (BA13) 50 0 -10 7.09 R-inferior frontal gyrus (BA47) 46 16 2 4.61 R-cingulate gyrus (BA32) 10 24 38 5.90 L-cingulate gyrus (BA32) -10 24 42 5.22 L-insula (BA13) -30 20 6 5.24 Paragraph length (NSME) L-hippocampus/parahippocampal gyrus (BA36) -22 -40 -2 6.18 R-precuneus/cingulate gyrus (BA31) 22 -56 30 6.03 R-intraparietal sulcus (BA7) 30 -72 38 4.76 p < 0.001 (uncorrected) with a cluster-wise FDR of p < 0.05.
- Table 2.
Volume overlap ratio of the NSME regions in this study with previously reported functional networks
SME/NSME networks (Kim, 2011) Functional networks (Shirer et al., 2012) Anatomical region SME NSME ASN dDMN vDMN PN LECN VSN Sentence length NSME Cingulate gyrus 0.92* 0.15 0.03 L-insula 0.40* R-insula/inferior frontal gyrus 0.41* 0.37* 0.05 Paragraph-length NSME R-precuneus/cingulate gyrus 0.18* 0.41* 0.20 0.37* 0.04 L-hippocampus/parahippocampal gyrus 0.01 0.04 R-intraparietal sulcus 0.35* 0.11 0.51* 0.37* SME/NSME networks were defined by multiple spheres of which locations and volumes were given by SME (verbal associative subgroup) and NSME (verbal item subgroup; Table 3 and 6 in Kim, 2011, respectively). Functional ROIs reported by Shirer et al. (2012) were analyzed. 6/14 functional networks showing significant overlap are listed in the table. Volume overlap was defined by (overlapped voxels)/(#voxels in the ROI). Each functional network was inflated ±1 voxel (3 mm) to give stable overlap. * indicates significance at the level of p < 0.05 with FDR correction (q < 0.05), with a null hypothesis of “volume overlap ratio between ROIs and functional regions is equal to the volume ratio of the region to the domain (voxels included in every network)”. ASN: anterior salience network; dDMN and vDMN: dorsal and ventral DMNs; PN: precuneus network; LECN: left-executive control network; VSN: visuospatial network.
In this issue
