Neural categorization of visual words of alphabetic and non-alphabetic languages
- School of Psychological and Cognitive Sciences, PKU-IDG/McGovern Institute for Brain Research, Peking University, China
Figures
Figure 1
Illustrations of the repetition suppression (RS) paradigm and univariate/multivariate analyses of brain activities in response to words.
(a) The RS paradigm. Words of two different languages (e.g., Chinese vs. English, or English vs. German) were presented alternately in the Alt-Cond, and words of one language were presented repeatedly in the Rep-Cond. (b) Univariate analyses. The RS effect was quantified as the decrease of averaged neural responses to words of one language in the Rep-Cond vs. Alt-Cond. This RS effect manifested habituation of the integrated neural activities that supported classification of words into one or another category and occurred more frequently in the Rep-Cond (vs. Alt-Cond). (c) Multivariate analyses. Correlation distances between neural responses to two words were calculated to estimate how words of one language were clustered (i.e., intra-language similarity) and how words of two languages were separated (i.e., inter-language difference) during language-based word categorization.
Figure 2
Electroencephalography (EEG) results of Chinese speakers in Experiment 1.
(a, b) Results of univariate analyses (N = 34). Top panels illustrate electrophysiological responses to Chinese and English words in the Alt-Cond and Rep-Cond, respectively. Bottom panels show scalp distributions of significant repetition suppression (RS) effects on neural responses to words of each language. (c) Illustration of the procedure of computing neural representation dissimilarity matrices (RDMs) in the multivariate analyses of correlation distances between words. (d) Illustration of the 160 × 160 neural RDM. Triangles represent the neural RDMs corresponding to intra-language similarity. Squares represent the RDMs corresponding to inter-language difference. (e–g) Results of multivariate and multidimensional scaling analyses. The top two panels show the time courses of significant differences in correlation distances between words corresponding to intra-language similarity and inter-language difference between the Alt-Cond and Rep-Cond and the neural RDM in the two conditions, respectively. The bottom two panels illustrate clustered representations of the words in the two-dimensional (2D) word space built based on the first two dimensions of multidimensional scaling analyses of neural RDMs corresponding to intra-language similarity and inter-language difference, respectively, and the mean Euclidean distances in the 2D word space between two words of the same language and between two words of different languages. ***p < 0.001.
Figure 3
Magnetoencephalography (MEG) results combined Chinese and English words across Chinese and English speakers in Experiments 7a and 8a.
(a) Significant clusters of repetition suppression (RS) effects on magnetometer signals (the left two panels) and gradiometer signals (the right panel) (N = 68). (b) Results of the whole-brain source analyses. Shown are left, right, and ventral views of brain regions in which MEG signals in response to words of the same language decreased significantly in the Rep-Cond (vs. Alt-Cond). (c) Illustration of the regions of interest used in the following MEG data analyses. (d) Source-space MEG signals in the brain regions showing significant RS effects. Time windows of the significant clusters are shown for each brain region. LFULG = left fusiform and lingual gyrus; RFULG = right fusiform and lingual gyrus; LPHC = left parahippocampal cortex; RPHC = right parahippocampal cortex; LATL = left anterior temporal lobe; RATL = right anterior temporal lobe; LI = left insula; RI = right insula; LOFC = left orbital frontal cortex; ROFC = right orbital frontal cortex; LSTS = left superior temporal sulcus.
Figure 4
Results of the analyses of the global network representation dissimilarity matrix (RDM) across Chinese and English speakers in Experiments 7a and 8a.
(a) Illustration of the procedure to calculate network RDMs based on source-space magnetoencephalography (MEG) signals. (b–d) Results of multivariate analyses (N = 68). The top two panels show the time courses of significant differences in the correlation distances corresponding to intra-language similarity and inter-language difference between the Alt-Cond and Rep-Cond and the neural RDMs in the two conditions, respectively. The bottom two panels illustrate clustered representations of words in the two-dimensional (2D) word space built based on the first two dimensions of multidimensional scaling analyses of neural RDMs corresponding to intra-language similarity and inter-language difference, respectively, and the mean Euclidean distances in the 2D word space between two words of the same language and between two words of different languages. ***p < 0.001.
Figure 5
Results of the analyses of local representation dissimilarity matrices (RDMs) across Chinese and English speakers in Experiments 7a and 8a.
(a) Illustration of the procedure to calculate local RDMs based on source-space magnetoencephalography (MEG) signals. (b, c) Dynamic repetition suppression (RS) effects related to intra-language similarity of Chinese and English words in different regions of interest (ROIs) (N = 68). (d) Dynamic RS effects related to inter-language difference in different ROIs.
Figure 6
Results of GCA and linear regression analyses across Chinese and English speakers in Experiments 7a and 8a.
(a) Connectivity values above a threshold (GC values >0.001) corresponding to intra-language similarity and inter-language difference (N = 68). Each square represents connectivity from one brain region indicated by the x-axis to another region indicated by the y-axis. (b) Patterns of functional connectivity between two regions of interest (ROIs). Arrows indicate directions of information flow from one to another brain region. Lines with two arrows indicate information flow with mutual directions. (c) The mean connectivity strength in the whole word-categorization network related to the processing of intra-language similarity and inter-language difference. (d) Associations between the correlation distances corresponding to inter-language difference and intra-language similarity, and the neural index of language-based word categorization (δ) during the implicit and explicit language-based word categorization tasks. Shown are partial regression plots (e) Associations between the neural index of language-based word categorization (δ) and behavioral performances (response accuracies, reaction time, and inverse efficiency scores (IES)) during the explicit language-based word categorization tasks. ***p < 0.001; n.s. = no significance.
Appendix 2—figure 1
Illustrations of ERPs to non-target Chinese words at all electrodes in Chinese speakers in Experiment 1 (N = 34).
The N170 showed the largest amplitude over the left occipito-temporal electrodes.
Appendix 2—figure 2
ERPs to German and English words in the Alt-Cond and Rep-Cond at the frontal-central and central parietal electrodes in Chinese speakers in Experiment 1 (N = 34).
No significant difference was detected in the ERP amplitudes to English (or German) words between the Rep-Cond and Alt-Cond. n.s. = not significant.
Appendix 2—figure 3
Illustration of the neural repetition suppression (RS) effects of each individual participant in the Chinese–English session of Experiment 1 (N = 34).
The left and middle panels show the differences in the correlation distances of the representation dissimilarity matrices (RDMs) corresponding to intra-language similarity between the Rep-Cond and Alt-Cond at 180–206 ms for Chinese words and at 168–226 ms for English words, respectively. The right panel shows the differences in the correlation distances of the RDMs corresponding to inter-language difference between the Alt-Cond and Rep-Cond at 162–218 ms. Intra-CH similarity = intra-Chinese word similarity; Intra-EN similarity = intra-English word similarity; Intra-CHEN difference = inter-Chinese/English word difference.
Appendix 2—figure 4
ERPs to scrambled Chinese and English words in the Alt-Cond and Rep-Cond at the frontal-central and central parietal electrodes in Chinese speakers in Experiment 2 (N = 34).
Scramble words similarly elicited the N1/P2/N2/LPP components. However, there was no evidence for any significant repetition suppression (RS) effect on neural responses to scrambled Chinese or English words. n.s. = not significant.
Appendix 2—figure 5
Results in Experiment 3.
(a, b) ERPs to Chinese radicals and English letters in the Alt-Cond and Rep-Cond at the frontal-central and central-parietal electrodes in Chinese speakers in Experiment 3 (N = 34). (c–e) Results of the multivariate analyses.
Appendix 2—figure 6
Electroencephalography (EEG) results of English speakers in Experiment 4.
(a, b) Results of univariate analyses (N = 34). (c–e) Results of multivariate and multidimensional scaling analyses. The top two panels show the time courses of significant differences in the correlation distances corresponding to intra-language similarity and inter-language difference between the Alt-Cond and Rep-Cond and the neural representation dissimilarity matrices (RDMs) in the two conditions, respectively. The bottom two panels illustrate clustered representations of the words in the two-dimensional (2D) word space based on the first two dimensions of multidimensional scaling analyses of the neural RDM corresponding to intra-language similarity and inter-language difference, respectively, and the mean Euclidean distances in the 2D word space between two words of the same language and between two words of different languages. ***p < 0.001.
Appendix 2—figure 7
ERPs to German and English words in the Alt-Cond and Rep-Cond at the frontal-central and central-parietal electrodes in English speakers in Experiment 4 (N = 34).
No significant RS effect was observed. n.s. = not significant.
Appendix 2—figure 8
Electroencephalography (EEG) results of German speakers in Experiment 5.
(a, b) Results of univariate analyses (N = 34). Top panels illustrate electrophysiological responses to Chinese and English words in the Alt-Cond and Rep-Cond, respectively. Bottom panels show voltage topographies of the scalp distributions of significant repetition suppression (RS) effects on neural responses to words of each language. (c–e) Results of multidimensional scaling analyses. The top two panels show the time courses of significant differences in the correlation distances corresponding to intra-language similarity and inter-language difference between the Alt-Cond and Rep-Cond and the neural representation dissimilarity matrix (RDM) in the two conditions, respectively. The bottom two panels illustrate clustered representations of words in the two-dimensional (2D) word space built based on the first two dimensions of multidimensional scaling analyses of neural RDMs corresponding to intra-language similarity and inter-language difference, respectively, and the mean Euclidean distances in the 2D word space between two words of the same language and between two words of different languages.
Appendix 2—figure 9
ERPs to German and English words in the Alt-Cond and Rep-Cond at the frontal-central and central-parietal electrodes in German speakers in Experiment 5 (N = 34).
No significant repetition suppression (RS) effect was observed. n.s. = not significant.
Appendix 2—figure 10
Electroencephalography (EEG) results of Chinese speakers in Experiment 6.
(a, b) Results of univariate analyses of electrophysiological responses to Korean and Italian words in the Alt-Cond and Rep-Cond, respectively (N = 34). (c–e) Results of multivariate and multidimensional scaling analyses. The top two panels show the time courses of significant differences in the correlation distances corresponding to intra-language similarity and inter-language difference between the Alt-Cond and the Rep-Cond and the neural representation dissimilarity matrices (RDMs) in the two conditions, respectively. The bottom two panels illustrate clustered representations of the words in the two-dimensional (2D) word space based on the first two dimensions of multidimensional scaling analyses of neural RDMs corresponding to intra-language similarity and inter-language difference, respectively, and the mean Euclidean distances in the 2D word space between two words of the same language and between two words of different languages. ***p < 0.001.
Appendix 2—figure 11
Illustration of the neural repetition suppression (RS) effects of each individual participant in Experiments 7a and 8a.
The left and middle panels show the differences in the correlation distances of the representation dissimilarity matrices (RDMs) corresponding to intra-language similarity between the Rep-Cond and Alt-Cond averaged at 159–200 ms for Chinese words and at 151–198 ms for English words, respectively. The right panel shows the differences in the correlation distances of the RDMs corresponding to inter-language difference between the Alt-Cond and Rep-Cond averaged at 150–232 and 236–294 ms. Intra-CH similarity = intra-Chinese word similarity; Intra-EN similarity = intra-English word similarity; Intra-CHEN difference = inter-Chinese/English word difference.
Appendix 2—figure 12
Results of the ROI analyses of magnetoencephalography (MEG) signals to Chinese or English words in Chinese speakers in Experiment 7a (N = 34).
(a-c) The top two panels show the time courses of significant differences in the correlation distances corresponding to intra-language similarity and inter-language difference between the Alt-Cond and Rep-Cond, and the neural representation dissimilarity matrix (RDM) in the two conditions, respectively. The bottom two panels illustrate clustered representations of words in the two-dimensional (2D) word space built based on the first two dimensions of multidimensional scaling analyses of neural RDMs corresponding to intra-language similarity and inter-language difference, respectively, and the mean Euclidean distances in the 2D word space between two words of the same language and between two words of different languages. ***p < 0.001.
Appendix 2—figure 13
Results of the region of interest (ROI) analyses of magnetoencephalography (MEG) signals to Chinese or English words in English speakers in Experiment 8a (N = 34).
(a-c) The top two panels show the time courses of significant differences in the correlation distances corresponding to intra-language similarity and inter-language difference between the Alt-Cond and Rep-Cond, and the neural representation dissimilarity matrix (RDM) in the two conditions, respectively. The bottom two panels illustrate clustered representations of words in the two-dimensional (2D) word space built based on the first two dimensions of multidimensional scaling analyses of neural RDMs corresponding to intra-language similarity and inter-language difference, respectively, and the mean Euclidean distances in the 2D word space between two words of the same language and between two words of different languages. ***p < 0.001.
Appendix 2—figure 14
Results of the disruption analysis of the relationship between δ and inverse efficiency score (IES) (N = 68).
Shown are beta values of the regression analyses of the relationships between δ and behavioral categorization of words when the bilateral FULG, PHC, ATL, insula, OFC, and LSTS were removed from the network, respectively. The results showed that the association between δ and behavioral categorization of words was significant except when the bilateral FULGs were removed from the network. FULG = fusiform and lingual gyrus; PHC = parahippocampal cortex; STS = superior temporal sulcus; ATL = anterior temporal lobe; OFC = orbital frontal cortex. n.s. = no significance. *p < 0.05 (FDR correction).
Appendix 2—figure 15
Results of whole-brain sensor-space and source-space magnetoencephalography (MEG) signals to Korean and Italian words in Chinese speakers in Experiment 9 (N = 34).
(a-c) Whole-brain repetition suppression (RS) effects of the source-space signals were identified using a lenient threshold (a predefined threshold of p < 0.01, 10,000 iterations, and a cluster-level threshold of p < 0.05, one-tailed). RS effects in the regions of interest (ROIs) were identified using a lenient threshold (a predefined threshold of p < 0.05, 10,000 iterations, and a cluster-level threshold of p < 0.05, one-tailed).
Appendix 2—figure 16
Results of the region of interest (ROI) analyses of magnetoencephalography (MEG) signals to Korean and Italian in Chinese speakers in Experiment 9 (N = 34).
(a–c) Results of the ROI analyses of MEG signals to Korean or Italian words in native Chinese speakers in Experiment 9. The top two panels show the time courses of significant differences in the correlation distances corresponding to intra-language similarity and inter-language difference between the Alt-Cond and Rep-Cond, and the neural representation dissimilarity matrix (RDM) in the two conditions, respectively. The bottom two panels illustrate clustered representations of words in the two-dimensional (2D) word space built based on the first two dimensions of multidimensional scaling analyses of neural RDMs corresponding to intra-language similarity and inter-language difference, respectively, and the mean Euclidean distances in the 2D word space between two words of the same language and between two words of different languages. The repetition suppression (RS) effect of intra-language similarity of Italian words was identified using a lenient threshold (a predefined threshold of p < 0.1, 10,000 iterations, and a cluster-level threshold of p < 0.05, one-tailed) in the time window of 150–200 ms. ***p < 0.001
Appendix 2—figure 17
GCA results in Chinese speakers using Korean and Italian words in Experiment 9 (N = 34).
Shown are (a) connectivity values above a threshold (GC values >0.001), (b) the connectivity patterns, and (c) the mean connectivity strength in the whole word-categorization network related to the processing of intra-language similarity and inter-language difference. **p < 0.01
Appendix 2—figure 18
RS effects on neural responses to words in the frontal language network in Experiments 7a and 8a (N = 68).
(a) Illustration of the frontal language network that includes Brodmann areas 44, 45, 47, and 6. (b) Time courses of magnetoencephalography (MEG) source signals in each brain region of the frontal language network in the Alt-Cond and Rep-Cond. The Brodmann area (BA 44 and 45) was defined based on the PALS atlas (Van Essen, 2005) and is a key node of the domain-specific language network and domain-general multiple-demand network (Fedorenko and Blank, 2020). BA 47 is involved in syntactic and semantic processing (Ardila et al., 2017; Friederici, 2011). BA 6 is associated with the integration of language and action during reading and naming tasks (Duffau et al., 2003; Pulvermüller, 2005). Source-space MEG signals from the Alt-Cond and Rep-Cond were plotted and subject to cluster-based permutation t-tests (predefined threshold of p < 0.001 and a cluster-level threshold of p < 0.05, one-tailed, 10,000 iterations). No significant repetition suppression (RS) effect was observed in the analyses that combined source-space MEG signals to Chinese and English words across Chinese and English speakers.
Appendix 3—figure 1
Time-resolved split-half reliability of the neural representation dissimilarity matrices (RDMs).
(a, b) Mean split-half reliability (blue line) and ±1 SD (shaded) across 200 random splits in electroencephalography (EEG) and magnetoencephalography (MEG) experiments. Significance was assessed using a one-sided cluster-based permutation test against zero, restricted to 0–800 ms. Black bars indicate significant clusters greater than zero (predefined threshold of p < 0.001 and a cluster-level threshold of p < 0.001, 10,000 iterations).
Tables
Appendix 1—table 1
Demographic information about participants in Experiments 1–9.
| Experiments | Sample size | Sex | Age (years) | Age range (years) | Chinese proficiency (years) | English proficiency (years) | German proficiency (years) |
|---|---|---|---|---|---|---|---|
| Experiment 1 | 34 | 17 males | 20.41 ± 2.09 | 18–26 | Native | 12.74 ± 3.00 | / |
| Experiment 2 | 34 | 22 males | 21.21 ± 2.69 | 17–24 | Native | 12.65 ± 3.13 | / |
| Experiment 3 | 34 | 13 males | 20.26 ± 2.08 | 18–28 | Native | 13.00 ± 2.91 | / |
| Experiment 4 | 34 | 16 males | 22.15 ± 2.12 | 19–28 | 9.12 ± 6.67 | Native | / |
| Experiment 5 | 34 | 20 males | 23.76 ± 3.14 | 18–34 | 3.05 ± 3.29 | 13.53 ± 3.76 | Native |
| Experiment 6 | 34 | 17 males | 19.85 ± 2.09 | 18–28 | Native | 12.00 ± 2.45 | / |
| Experiments 7a and 7b | 34 | 17 males | 22.47 ± 3.60 | 19–32 | Native | 13.38 ± 3.09 | / |
| Experiments 8a and 8b | 34 | 17 males | 23.21 ± 2.74 | 19–31 | 9.40 ± 7.31 | Native | / |
| Experiment 9 | 34 | 15 males | 20.35 ± 2.06 | 17–25 | Native | 12.56 ± 2.68 | / |
-
Notes. The language proficiency was measured as the duration of formal language acquisition.
Appendix 1—table 2
Word stimuli used in Experiments 1 and 4–9.
| Chinese | English | German | Korean | Italian |
|---|---|---|---|---|
| 狮子 | Lion | Löwe | 사자 | Leone |
| 马匹 | Horse | Pferd | 말 | Cavallo |
| 山羊 | Goat | Ziege | 염소 | Capra |
| 鸭子 | Duck | Ente | 오리 | Anatra |
| 母牛 | Cow | Kuh | 소 | Mucca |
| 猴子 | Monkey | Affe | 원숭이 | Scimmia |
| 大象 | Elephant | Elefant | 코끼리 | Elefante |
| 猩猩 | Chimpanzee | Schimpanse | 침팬지 | Scimpanzé |
| 小鱼 | Fish | Fisch | 물고기 | Pesce |
| 狐狸 | Fox | Fuchs | 여우 | Volpe |
| 老鹰 | Eagle | Adler | 독수리 | Aquila |
| 小鹿 | Deer | Hirsch | 사슴 | Cervo |
| 野驴 | Donkey | Esel | 당나귀 | Asino |
| 袋鼠 | Kangaroo | Känguru | 캥거루 | Canguro |
| 黑熊 | Bear | Bär | 곰 | Orso |
| 青蛙 | Frog | Frosch | 개구리 | Rana |
| 小猫 | Cat | Katze | 고양이 | Gatto |
| 家犬 | Dog | Hund | 개 | Cane |
| 海豚 | Dolphin | Delfin | 돌고래 | Delfino |
| 公鸡 | Rooster | Hahn | 수탉 | Gallo |
| 电话 | Telephone | Telefon | 전화기 | Telefono |
| 钟表 | Clock | Uhr | 시계 | Orologio |
| 雨伞 | Umbrella | Regenschirm | 우산 | Ombrello |
| 汽车 | Car | Auto | 자동차 | Auto |
| 火车 | Train | Zug | 기차 | Treno |
| 地铁 | Subway | U-Bahn | 지하철 | Metropolitana |
| 飞机 | Airplane | Flugzeug | 비행기 | Aereo |
| 火箭 | Rocket | Rakete | 로켓 | Razzo |
| 胶水 | Glue | Leim | 접착제 | Colla |
| 桌子 | Table | Tisch | 테이블 | Tavolo |
| 椅子 | Chair | Stuhl | 의자 | Sedia |
| 按钮 | Button | Knopf | 단추 | Pulsante |
| 抽屉 | Drawer | Schublade | 서랍 | Cassetto |
| 冰箱 | Refrigerator | Kühlschrank | 냉장고 | Frigorifero |
| 钢笔 | Pen | Stift | 펜 | Penna |
| 尺子 | Ruler | Lineal | 자 | Righello |
| 书籍 | Book | Buch | 책 | Libro |
| 电视 | Television | Fernsehen | 텔레비전 | Televisione |
| 足球 | Football | Fußball | 축구공 | Calcio |
| 瓶子 | Bottle | Flasche | 병 | Bottiglia |
Appendix 1—table 3
Behavioral performances in the one-back task in Experiments 1–9.
| Experiment 1 (Chinese native speakers) | ||||
|---|---|---|---|---|
| Session | Chinese–English | |||
| Condition | Rep-CH | Rep-EN | Alt-CH | Alt-EN |
| Accuracy (%) | 96.08 ± 0.07 | 97.06 ± 0.05 | 98.23 ± 0.04 | 96.07 ± 0.07 |
| RT (ms) | 568.87 ± 62.93 | 577.87 ± 51.31 | 577.76 ± 70.90 | 579.90 ± 58.65 |
| IES | 597.86 ± 106.10 | 598.75 ± 78.67 | 589.87 ± 84.75 | 609.36 ± 101.43 |
| Session | German–English | |||
| Condition | Rep-GE | Rep-EN | Alt-GE | Alt-EN |
| Accuracy (%) | 96.08 ± 0.07 | 96.57 ± 0.07 | 94.51 ± 0.11 | 94.87 ± 0.07 |
| RT (ms) | 567.51 ± 51.93 | 575.90 ± 65.83 | 570.94 ± 68.88 | 572.01 ± 58.33 |
| IES | 597.07 ± 97.49 | 603.69 ± 120.12 | 619.10 ± 155.78 | 607.88 ± 89.40 |
| Experiment 2 (Chinese native speakers) | ||||
| Session | Scrambled Chinese–English | |||
| Condition | Rep-SCH | Rep-SEN | Alt-SCH | Alt-SEN |
| Accuracy (%) | 87.99 ± 0.12 | 85.29 ± 0.11 | 85.15 ± 0.15 | 87.35 ± 0.09 |
| RT (ms) | 572.50 ± 45.55 | 591.29 ± 65.46 | 590.64 ± 69.20 | 575.52 ± 55.99 |
| IES | 664.56 ± 116.13 | 706.74 ± 136.33 | 737.82 ± 298.61 | 668.66 ± 115.21 |
| Experiment 3 (Chinese native speakers) | ||||
| Session | Chinese radicals–English letters | |||
| Condition | Rep-RD | Rep-LT | Alt-RD | Alt-LT |
| Accuracy (%) | 92.89 ± 0.08 | 91.91 ± 0.13 | 92.68 ± 0.10 | 88.36 ± 0.18 |
| RT (ms) | 586.79 ± 74.49 | 604.58 ± 79.74 | 595.52 ± 78.41 | 602.78 ± 95.71 |
| IES | 640.48 ± 125.79 | 681.05 ± 193.05 | 658.99 ± 170.99 | 786.51 ± 626.46 |
| Experiment 4 (English native speakers) | ||||
| Session | Chinese–English | |||
| Condition | Rep-CH | Rep-EN | Alt-CH | Alt-EN |
| Accuracy (%) | 94.12 ± 0.09 | 94.36 ± 0.08 | 93.79 ± 0.08 | 95.22 ± 0.07 |
| RT (ms) | 568.62 ± 68.39 | 553.12 ± 70.98 | 572.36 ± 69.51 | 563.14 ± 76.27 |
| IES | 613.65 ± 123.58 | 594.68 ± 121.63 | 618.07 ± 116.99 | 595.45 ± 101.52 |
| Session | German–English | |||
| Condition | Rep-GE | Rep-EN | Alt-GE | Alt-EN |
| Accuracy (%) | 93.63 ± 0.07 | 94.36 ± 0.09 | 93.87 ± 0.09 | 95.29 ± 0.07 |
| RT (ms) | 565.77 ± 71.28 | 562.26 ± 74.15 | 571.89 ± 67.14 | 560.00 ± 78.80 |
| IES | 610.38 ± 104.48 | 603.51 ± 113.77 | 616.59 ± 103.86 | 592.48 ± 106.46 |
| Experiment 5 (German native speakers) | ||||
| Session | Chinese–English | |||
| Condition | Rep-CH | Rep-EN | Alt-CH | Alt-EN |
| Accuracy (%) | 89.71 ± 0.10 | 91.67 ± 0.12 | 92.95 ± 0.09 | 93.04 ± 0.10 |
| RT (ms) | 583.35 ± 60.74 | 575.23 ± 74.58 | 582.86 ± 68.93 | 578.84 ± 60.55 |
| IES | 660.29 ± 121.25 | 642.70 ± 145.40 | 633.36 ± 105.32 | 633.51 ± 125.41 |
| Session | German–English | |||
| Condition | Rep-GE | Rep-EN | Alt-GE | Alt-EN |
| Accuracy (%) | 93.38 ± 0.10 | 93.38 ± 0.11 | 92.61 ± 0.09 | 91.99 ± 0.12 |
| RT (ms) | 579.84 ± 84.87 | 574.56 ± 71.39 | 573.63 ± 85.37 | 579.65 ± 78.41 |
| IES | 637.01 ± 175.36 | 632.08 ± 168.75 | 628.80 ± 138.81 | 655.02 ± 204.02 |
| Experiment 6 (Chinese native speakers) | ||||
| Session | Korean–Italian | |||
| Condition | Rep-KR | Rep-IT | Alt-KR | Alt-IT |
| Accuracy (%) | 91.42 ± 0.09 | 95.83 ± 0.06 | 91.01 ± 0.10 | 94.92 ± 0.08 |
| RT (ms) | 567.54 ± 63.37 | 571.46 ± 47.05 | 557.62 ± 60.88 | 581.88 ± 51.57 |
| IES | 627.45 ± 97.54 | 597.98 ± 55.39 | 619.77 ± 94.45 | 618.14 ± 86.30 |
| Experiment 7a (Chinese native speakers) | ||||
| Session | Chinese–English | |||
| Condition | Rep-CH | Rep-EN | Alt-CH | Alt-EN |
| Accuracy (%) | 94.61 ± 0.08 | 92.89 ± 0.10 | 95.03 ± 0.07 | 92.47 ± 0.07 |
| RT (ms) | 583.01 ± 69.17 | 585.27 ± 68.04 | 590.05 ± 70.82 | 591.92 ± 68.54 |
| IES | 621.91 ± 97.81 | 641.58 ± 133.50 | 625.19 ± 94.31 | 646.45 ± 112.91 |
| Experiment 8a (English native speakers) | ||||
| Session | Chinese–English | |||
| Condition | Rep-CH | Rep-EN | Alt-CH | Alt-EN |
| Accuracy (%) | 92.16 ± 0.09 | 97.06 ± 0.05 | 95.64± 0.07 | 95.50 ± 0.06 |
| RT (ms) | 580.11 ± 69.21 | 566.32 ± 63.93 | 574.91 ± 56.82 | 551.50 ± 64.55 |
| IES | 640.99 ± 131.84 | 585.73 ± 79.17 | 605.90 ± 88.05 | 581.03 ± 88.65 |
| Experiment 9 (Chinese native speakers) | ||||
| Session | Korean–Italian | |||
| Condition | Rep-KR | Rep-IT | Alt-KR | Alt-IT |
| Accuracy (%) | 92.65 ± 0.07 | 94.12 ± 0.08 | 89.57 ± 0.11 | 94.78 ± 0.09 |
| RT (ms) | 564.02 ± 58.15 | 570.92 ± 68.39 | 554.83 ± 67.08 | 563.77 ± 52.02 |
| IES | 614.21 ± 92.16 | 613.14 ± 108.48 | 632.32 ± 138.80 | 602.62 ± 98.28 |
-
Notes. CH, Chinese; EN, English; GE, German; SCH, Scrambled Chinese; SEN, Scrambled English; RD, Radicals; LT, Letters; KR, Korean; IT, Italian; IES, inverse efficiency score.
Appendix 1—table 4
Statistical details of cluster-based permutation t-tests in Experiments 1–9.
| Experiment 1 | |||
|---|---|---|---|
| Session | Chinese–English | ||
| Analysis | Univariate analysis | ||
| Condition | Alt-CH vs. Rep-CH | Alt-EN vs. Rep-EN | Alt-Cond vs. Rep-Cond |
| Predefined p value | <0.05 | <0.05 | <0.05 |
| Cluster-level p value | <0.001 | <0.001 | <0.001 |
| Analysis | Multivariate analysis | ||
| Condition | Alt-CH vs. Rep-CH | Alt-EN vs. Rep-EN | Alt-CHEN vs. Rep-CHEN |
| Predefined p value | <0.025 | <0.025 | <0.025 |
| Cluster-level p value | 0.025 | <0.001 | <0.001 |
| Session | German–English | ||
| Analysis | Univariate analysis | ||
| Condition | Alt-GE vs. Rep-GE | Alt-EN vs. Rep-EN | |
| Predefined p value | <0.05 | <0.05 | |
| Cluster-level p value | 0.672 | 0.061 | |
| Experiment 2 | |||
| Session | Scrambled Chinese–English | ||
| Analysis | Univariate analysis | ||
| Condition | Alt-SCH vs. Rep-SCH | Alt-SEN vs. Rep-SEN | |
| Predefined p value | <0.05 | <0.05 | |
| Cluster-level p value | 1 | 0.144 | |
| Experiment 3 | |||
| Session | Chinese radicals–English letters | ||
| Analysis | Univariate analysis | ||
| Condition | Alt-RD vs. Rep-RD | Alt-LT vs. Rep-LT | Alt-Cond vs. Rep-Cond |
| Predefined p value | <0.05 | <0.05 | <0.05 |
| Cluster-level p value | 0.021 | 0.026 | 0.009 |
| Analysis | Multivariate analysis | ||
| Condition | Alt-RD vs. Rep-RD | Alt-LT vs. Rep-LT | Alt-RDLT vs. Rep-RDLT |
| Predefined p value | <0.025 | <0.025 | <0.025 |
| Cluster-level p value | 1 | 0.009 | <0.001 |
| Experiment 4 | |||
| Session | Chinese–English | ||
| Analysis | Univariate analysis | ||
| Condition | Alt-CH vs. Rep-CH | Alt-EN vs. Rep-EN | Alt-Cond vs. Rep-Cond |
| Predefined p value | <0.05 | <0.05 | <0.05 |
| Cluster-level p value | <0.001 | <0.001 | <0.001 |
| Analysis | Multivariate analysis | ||
| Condition | Alt-CH vs. Rep-CH | Alt-EN vs. Rep-EN | Alt-CHEN vs. Rep-CHEN |
| Predefined p value | <0.025 | <0.025 | <0.025 |
| Cluster-level p value | 0.005 | 0.021 | <0.001 |
| Session | German–English | ||
| Analysis | Univariate analysis | ||
| Condition | Alt-GE vs. Rep-GE | Alt-EN vs. Rep-EN | |
| Predefined p value | <0.05 | <0.05 | |
| Cluster-level p value | 0.131 | 0.332 | |
| Experiment 5 | |||
| Session | Chinese–English | ||
| Analysis | Univariate analysis | ||
| Condition | Alt-CH vs. Rep-CH | Alt-EN vs. Rep-EN | Alt-Cond vs. Rep-Cond |
| Predefined p value | <0.05 | <0.05 | <0.05 |
| Cluster-level p value | <0.001 | <0.001 | <0.001 |
| Analysis | Multivariate analysis | ||
| Condition | Alt-CH vs. Rep-CH | Alt-EN vs. Rep-EN | Alt-CHEN vs. Rep-CHEN |
| Predefined p value | <0.025 | <0.025 | <0.025 |
| Cluster-level p value | <0.001 | 0.004 | <0.001 |
| Session | German–English | ||
| Analysis | Univariate analysis | ||
| Condition | Alt-GE vs. Rep-GE | Alt-EN vs. Rep-EN | |
| Predefined p value | <0.05 | <0.05 | |
| Cluster-level p value | 0.634 | 0.310 | |
| Experiment 6 | |||
| Session | Korean–Italian | ||
| Analysis | Univariate analysis | ||
| Condition | Alt-KR vs. Rep-KR | Alt-IT vs. Rep-IT | Alt-Cond vs. Rep-Cond |
| Predefined p value | <0.05 | <0.05 | <0.05 |
| Cluster-level p value | <0.001 | <0.001 | <0.001 |
| Analysis | Multivariate analysis | ||
| Condition | Alt-KR vs. Rep-KR | Alt-IT vs. Rep-IT | Alt-KRIT vs. Rep-KRIT |
| Predefined p value | <0.025 | <0.025 | <0.025 |
| Cluster-level p value | 0.002 | <0.001 | <0.001 |
| Experiments 7a and 8a | |||
| Session | Chinese–English | ||
| Analysis | Univariate analysis (sensor space) | ||
| Condition | Alt-Cond vs. Rep-Cond | ||
| Predefined p value | <0.01 | ||
| Cluster-level p value | MAG Cluster 1 p < 0.001 | MAG Cluster 2 p < 0.001 | GRAD Cluster 1 p < 0.001 |
| Analysis | Univariate analysis (source space) | ||
| Condition | Alt-Cond vs. Rep-Cond | ||
| Predefined p value | <0.001 | ||
| Cluster-level p value | Cluster 1 p < 0.001 | Cluster 2 p < 0.001 | Cluster 3 p = 0.020 |
| Analysis | ROI analysis | ||
| Condition | Alt-Cond vs. Rep-Cond | ||
| Predefined p value | <0.001 | ||
| Cluster-level p value | All P values of clusters in the ROIs <0.003 | ||
| Analysis | Multivariate analysis (7a and 8a) | ||
| Condition | Alt-CH vs. Rep-CH | Alt-EN vs. Rep-EN | Alt-CHEN vs. Rep-CHEN |
| Predefined p value | <0.025 | <0.025 | <0.025 |
| Cluster-level p value | 0.003 | 0.001 | <0.001 |
| Analysis | Multivariate analysis (7a) | ||
| Condition | Alt-CH vs. Rep-CH | Alt-EN vs. Rep-EN | Alt-CHEN vs. Rep-CHEN |
| Predefined p value | <0.025 | <0.025 | <0.025 |
| Cluster-level p value | Cluster 1 p = 0.011; Cluster 2 p = 0.042; | 0.001 | Clusters 1 and 2 p < 0.001; Cluster 3 p = 0.034 |
| Analysis | Multivariate analysis (8a) | ||
| Condition | Alt-CH vs. Rep-CH | Alt-EN vs. Rep-EN | Alt-CHEN vs. Rep-CHEN |
| Predefined p value | <0.025 | <0.025 | <0.025 |
| Cluster-level p value | 0.011 | 0.014 | 0.002 |
| Experiment 9 | |||
| Session | Korean–Italian | ||
| Analysis | Univariate analysis (sensor space) | ||
| Condition | Alt-Cond vs. Rep-Cond | ||
| Predefined p value | <0.01 | ||
| Cluster-level p value | MAG Cluster 1 p = 0.018 | GRAD Cluster 1 p < 0.001 | GRAD Cluster 2 p = 0.018 |
| Analysis | Univariate analysis (source space) | ||
| Condition | Alt-Cond vs. Rep-Cond | ||
| Predefined p value | <0.01 | ||
| Cluster-level p value | Clusters 1 and 2 p < 0.001; Cluster 3 p = 0.030; Cluster 4 p = 0.037 | ||
| Analysis | ROI analysis | ||
| Condition | Alt-Cond vs. Rep-Cond | ||
| Predefined p value | <0.05 | ||
| Cluster-level p value | All p values of clusters in the ROIs <0.036 | ||
| Analysis | Multivariate analysis | ||
| Condition | Alt-KR vs. Rep-KR | Alt-IT vs. Rep-IT | Alt-KRIT vs. Rep-KRIT |
| Predefined p value | <0.025 | <0.025 | <0.025 |
| Cluster-level p value | Clusters 1 and 2 p < 0.001; Cluster 3 p = 0.041 | 0.021 | 0.006 |
-
Notes. CH, Chinese; EN, English; GE, German; SCH, Scrambled Chinese; SEN, Scrambled English; RD, Radicals; LT, Letters; KR, Korean; IT, Italian.
Appendix 1—table 5
Statistical details of t-tests in multivariate analysis.
| Experiment 1 | |||
|---|---|---|---|
| Condition | Alt-CH vs. Rep-CH | Alt-EN vs. Rep-EN | Alt-CHEN vs. Rep-CHEN |
| df | 779 | 779 | 1599 |
| t value | 8.497 | 18.771 | 27.150 |
| p value | <0.001 | <0.001 | <0.001 |
| Cohen’s d | 0.304 | 0.672 | 0.679 |
| 95% CI | [0.056, 0.090] | [0.118, 0.146] | [0.133, 0.154] |
| Experiment 4 | |||
| Condition | Alt-CH vs. Rep-CH | Alt-EN vs. Rep-EN | Alt-CHEN vs. Rep-CHEN |
| df | 779 | 779 | 1599 |
| t value | 19.483 | 6.726 | 22.643 |
| p value | <0.001 | <0.001 | <0.001 |
| Cohen’s d | 0.698 | 0.241 | 0.566 |
| 95% CI | [0.102, 0.124] | [0.045, 0.082] | [0.096, 0.114] |
| Experiment 5 | |||
| Condition | Alt-CH vs. Rep-CH | Alt-EN vs. Rep-EN | Alt-CHEN vs. Rep-CHEN |
| df | 779 | 779 | 1599 |
| t value | 20.841 | 10.221 | 61.125 |
| p value | <0.001 | <0.001 | <0.001 |
| Cohen’s d | 0.746 | 0.366 | 1.528 |
| 95% CI | [0.120, 0.145] | [0.050, 0.073] | [0.211, 0.224] |
| Experiment 6 | |||
| Condition | Alt-KR vs. Rep-KR | Alt-IT vs. Rep-IT | Alt-KRIT vs. Rep-KRIT |
| df | 779 | 779 | 1599 |
| t value | 12.454 | 13.851 | 39.241 |
| p value | <0.001 | <0.001 | <0.001 |
| Cohen’s d | 0.446 | 0.496 | 0.981 |
| 95% CI | [0.082, 0.113] | [0.081, 0.108] | [0.179, 0.198] |
| Experiments 7a and 8a | |||
| Condition | Alt-CH vs. Rep-CH | Alt-EN vs. Rep-EN | Alt-CHEN vs. Rep-CHEN |
| df | 779 | 779 | 1599 |
| t value | 11.327 | 8.549 | 29.376 |
| p value | <0.001 | <0.001 | <0.001 |
| Cohen’s d | 0.406 | 0.306 | 0.734 |
| 95% CI | [0.085, 0.120] | [0.066, 0.105] | [0.164, 0.188] |
| Experiment 7a | |||
| Condition | Alt-CH vs. Rep-CH | Alt-EN vs. Rep-EN | Alt-CHEN vs. Rep-CHEN |
| df | 779 | 779 | 1599 |
| t value | 11.901 | 14.297 | 23.370 |
| p value | <0.001 | <0.001 | <0.001 |
| Cohen’s d | 0.426 | 0.512 | 0.584 |
| 95% CI | [0.096, 0.134] | [0.131, 0.173] | [0.138, 0.164] |
| Experiment 8a | |||
| Condition | Alt-CH vs. Rep-CH | Alt-EN vs. Rep-EN | Alt-CHEN vs. Rep-CHEN |
| df | 779 | 779 | 1599 |
| t value | 12.791 | 4.655 | 20.260 |
| p value | <0.001 | <0.001 | <0.001 |
| Cohen’s d | 0.458 | 0.167 | 0.507 |
| 95% CI | [0.094, 0.129] | [0.028, 0.070] | [0.130, 0.158] |
| Experiment 9 | |||
| Condition | Alt-KR vs. Rep-KR | Alt-IT vs. Rep-IT | Alt-KRIT vs. Rep-KRIT |
| df | 779 | 779 | 1599 |
| t value | 10.206 | 3.853 | 14.925 |
| p value | <0.001 | <0.001 | <0.001 |
| Cohen’s d | 0.365 | 0.138 | 0.373 |
| 95% CI | [0.088, 0.130] | [0.020, 0.062] | [0.094, 0.122] |
-
Notes. CH, Chinese; EN, English; KR, Korean; IT, Italian.
Additional files
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)
Neural categorization of visual words of alphabetic and non-alphabetic languages
eLife 15:RP110320.
https://doi.org/10.7554/eLife.110320.3
