Mapping the neural mechanism that distinguishes between holistic thinking and analytic thinking

Holistic and analytic thinking are two distinct modes of thinking used to interpret the world with relative preferences varying across cultures. While most research on these thinking styles has focused on behavioral and cognitive aspects, a few studies have utilized functional magnetic resonance imaging (fMRI) to explore the correlations between brain metrics and self-reported scale scores. Other fMRI studies used single holistic and analytic thinking tasks. As a single task may involve processing in spurious low-level regions, we used two different holistic and analytic thinking tasks, namely the frame-line task and the triad task, to seek convergent brain regions to distinguish holistic and analytic thinking using multivariate pattern analysis (MVPA). Results showed that brain regions fundamental to distinguish holistic and analytic thinking include the bilateral frontal lobes, bilateral parietal lobes, bilateral precentral and postcentral gyrus, bilateral supplementary motor areas, bilateral fusiform, bilateral insula, bilateral angular gyrus, left cuneus, and precuneus, left olfactory cortex, cingulate gyrus, right caudate and putamen. Our study maps brain regions that distinguish between holistic and analytic thinking and provides a new approach to explore the neural representation of cultural constructs. We provide initial evidence connecting culture-related brain regions with language function to explain the origins of cultural differences in cognitive styles.


Introduction
Culture, encompassing a collective set of shared values and meaning systems among individuals in a society or group (Han et al., 2013;Kitayama and Salvador, 2017), not only influences people's cognitive processes and behavior but also shapes the functional organization of the brain (Han et al., 2013;Kwon et al., 2021).Neural activities related to various cultural constructs exhibit notable differences among individuals from distinct cultural backgrounds.For example, in a Japanese cohort, there is a significant negative correlation between interdependent self-construal (SC) and the gray matter volume of the orbitofrontal cortex (Kitayama et al., 2017).Chinese participants demonstrate a significant positive correlation between interdependent SC and gray matter volume in the ventromedial prefrontal cortex (PFC), right dorsolateral PFC, and right anterolateral PFC (Wang et al., 2017).Culture also influences people's recognition of emotional faces.When exposed to excited faces as opposed to calm faces, Western participants display greater activation in the bilateral ventral striatum and left caudate, while Chinese participants exhibit the opposite pattern (Park et al., 2016).
Numerous studies in cultural psychology highlight the influence of culture on individuals' thinking styles, subsequently shaping cognitive processes (Uskul et al., 2008).Easterners typically exhibit holistic thinking, emphasizing interconnectedness and attentiveness to contextual cues.Conversely, Westerners often demonstrate analytic thinking, focusing on object independence and properties to create category features for a more profound understanding and prediction of object behavior (Cohen, 2009;Spencer-Rodgers et al., 2010).Holistic versus analytic thinking entails different levels of psychological and behavioral consequences.Holistic thinkers categorize objects based on relationships and similarities, whereas analytic thinkers prioritize object properties (Ji et al., 2004).Holistic thinkers encounter difficulties in disregarding the impact of object surroundings when making decisions (Masuda et al., 2008) and in the processes of making attributions (Morris and Peng, 1994).They also tend to consider the antecedents and consequences of events more thoroughly compared to analytic thinkers (Maddux and Yuki, 2006).
Several studies have delved into the neural mechanisms underlying holistic and analytic thinking.In a recent resting-state fMRI study by Wang, Chen, and Sui (2022), participants' self-reported scores on the Dialectical Self Scale (Spencer-Rodgers et al., 2004) were significantly and positively correlated with coupling between the dorsal anterior cingulate cortex (dACC) and the default mode network (DMN).The dACC, a key node for processing conflicting information (Carter and van Veen, 2007), is affiliated with the salience network and plays a role in detecting and integrating emotional and sensory stimuli (Goulden et al., 2014).In addition, a graph theory analysis revealed that holistic thinking, as measured by the Analysis-Holism Scale (Choi et al., 2007), was negatively correlated with the degree and nodal global efficiency of the left dorsal superior frontal gyrus, a crucial DMN node (Luo et al., 2022).Degree is defined as the number of direct connections a node has with other nodes in the network, while the node's global efficiency is calculated by taking the inverse of the harmonic mean of the length between a node and all other nodes in the network (Bullmore and Sporns, 2009).Furthermore, holistic thinking was correlated positively with the degree and nodal global efficiency of the bilateral amygdala and putamen, which are vital for emotion processing and regulation, particularly in societies that prioritize interpersonal relationships (Luo et al., 2022).Holistic thinking derives from interdependent SC to some extent, which may predict higher gray matter volume in the parahippocampal place area (Yu et al., 2021).These preliminary results suggest that the neural substrates of holistic and analytic thinking involve multiple cortical brain regions, especially the DMN and subcortical nuclei.However, resting-state fMRI studies often rely on correlations between scale scores and various brain metrics within regions of interest (ROI) (Bacha-Trams et al., 2018;Wang et al., 2022).This approach raises several issues.First, the scores on self-reported scales may be affected by external contexts (Riemer et al., 2014), which raises uncertainty about the effectiveness of distinguishing between holistic and analytic thinkers.Second, previous studies often selected ROI for thinking styles based on single functions, overlooking their multifunctionality.For example, Wang, Chen, and Sui (2022) hypothesized dACC as a ROI, based on its role in processing conflicting information.However, other cognitive functions within the dACC (Heilbronner and Hayden, 2016) may also influence holistic and analytic thinking.Finally, correlational studies cannot determine causal relationships between thinking styles and specific brain regions.They also cannot ascertain the involvement of these brain regions in individual task performance related to holistic and analytic thinking.
In tasks related to holistic and analytic thinking, brain activation patterns differ significantly both across and within cultures.These thinking styles affect even basic visual processing.When viewing pictures, Americans activated more brain regions involved in object processing than East Asians (Gutchess et al., 2006); while Chinese showed greater activation in lateral occipital cortex to incongruent scenes than to congruent scenes relative to Americans (Jenkins et al., 2010).When viewing the same video, holistic thinkers exhibit significantly higher inter-subject correlations (ISC) of neural activity in occipital, prefrontal, and temporal cortices, compared to analytic thinkers (Bacha-Trams et al., 2018).In the classical attention-related paradigm known as the frame-line task (Hedden et al., 2008), holistic thinkers allocate more attentional resources to the absolute judgment task, resulting in increased activation in frontal and occipital lobes.This heightened activation contrasts with their default processing mode, as the absolute judgment task requires disregarding background information.Conversely, analytic thinkers exhibit increased activation in these brain regions during relative judgment tasks (Hedden et al., 2008).In addition to paying more attention to contextual factors, holistic thinkers also place greater emphasis on relationships and similarities between objects when further classifying them (Uskul et al., 2008).In the triad task (Ji et al., 2004), another classic paradigm of holistic thinking, East Asians activated the frontal-parietal network associated with controlled executive processes during semantic strategy tasks compared to control tasks.Americans, on the other hand, showed increased engagement of the temporal lobes and the cingulate, possibly reflecting the processing of semantic content conflict (Gutchess et al., 2010).In addition, individuals with higher levels of holistic thinking had smaller volumes in the bilateral nucleus accumbens and right amygdala than those with lower levels of holistic thinking.Hierarchical regression analyses showed that the holistic tendency negatively predicted left nucleus accumbens and right amygdala volumes, which play a crucial role in processing emotions, particularly in cultures that prioritize interpersonal relationships (Li and Hu, 2022).The above evidence for the frame-line and triad tasks suggests that the neural activity underlying holistic and analytic thinking involves multiple brain regions, including the frontal-parietal network, occipital lobes, and subcortical nuclei.However, prior task-state fMRI studies investigating holistic thinking have relied on single-task designs, which may confound the relationship between the brain regions truly related to holistic and analytic thinking versus task-related brain regions.
Behavioral studies have demonstrated that holistic and analytic thinking are distinct cognitive processing modes operating at higher levels (Luo et al., 2022;Wang et al., 2022).However, studies investigating the neural underpinnings of these processes remain rare, motivating the present fMRI study to address this gap.To avoid potential involvement of spurious low-level regions, the current study employs two well-established paradigms, the frame-line task and the triad task, to identify convergent neural representations of holistic and analytic thinking.The frame-line task, designed to evaluate an individual's ability to incorporate and ignore situational influences in a single nonsocial area of attention, boasts objective performance criteria (Kitayama et al., 2003).Holistic thinking and analytic thinking, high-level top-down cognitive styles, are also reflected in the more complex triad task.The triad task was designed to assess whether people classify based on holistic or analytic thinking.The brain stores information according to specific rules, which can be influenced by cultural factors (Ji et al., 2004).Classification is a fundamental principle for organizing information, enabling us not only to identify objects but also to understand how new and unfamiliar objects fit into existing knowledge structures (Gutchess et al., 2010).Thus, the brain areas involved in the triad task may include areas beyond those used for holistic and analytic thinking.The frame line task, though only a basic attention task, avoids interference from semantic information storage and extraction required for the triad task.The frame-line task and the triad task each have their own unique advantages and disadvantages.Their combination, however, offers a complementary approach for mapping the brain regions distinguishing holistic thinking and analytic thinking.
It should be noted that the distinction between holistic thinking and analytic thinking also exists within cultures (de Oliveira and Nisbett, 2017).Individuals exhibit varying degrees of both thinking styles, which Y. Teng et al. can be temporarily primed (Ma-Kellams et al., 2011).Our study intentionally focuses on a single cultural context to avoid confounding variables associated with different cultural backgrounds (Bacha-Trams et al., 2018).We explore the behavioral and neurological mechanisms underlying holistic and analytic thinking by examining participants who exhibit a range of thinking styles.In particular, we use multivariate pattern analysis (MVPA) to search brain regions that can specifically differentiate between holistic thinking and analytic thinking.Previous fMRI studies on cultural constructs relied on limited univariate fMRI analyses, such as activation analysis (Gutchess et al., 2010;Hedden et al., 2008), focusing on the differences in average brain activity.These methods may be suboptimal, as regional BOLD activity may not reflect subtle cortical specializations, which may be necessary to generate perceptual differences between cultures (Ksander et al., 2018).MVPA can show discrepancies in the representation pattern of multiple brain voxels under different conditions (Haxby, 2012).Compared to activation analysis, MVPA is considered a more sensitive and informative method for revealing the functional organization of the cortex, as it aligns with the notion that people's perception of the world is inherently multivariate in nature (Haxby et al., 2014).In previous studies, MVPA was only used for a single task.Here, we first combine two tasks through MVPA, identifying brain regions that show significantly different patterns between conditions representing holistic and analytic thinking.
In summary, the present study aims to delineate the foundational brain regions that distinguish between holistic and analytic thinking.To achieve this goal, we conduct two classic experiments, the frame-line task and the triad task, to eliminate brain regions associated solely with the tasks but not with the thinking styles.Based on previous studies, we hypothesized that holistic thinking and analytic thinking involve multiple brain regions, including frontal parietal network, occipital lobes, and subcortical nuclei (Gutchess et al., 2010;Hedden et al., 2008;Li and Hu, 2022).We then employ MVPA to identify brain regions that exhibit significantly different neural representations between holistic thinking and analytic thinking.By eliminating interference from the two tasks themselves, this methodological innovation facilitates a more fundamental and accurate understanding of the neural mechanisms underlying holistic and analytic thinking.

Participants
Fifty Chinese participants completed fMRI experimental tasks, including the frame-line task and the triad task.Task order was counterbalanced across participants.For the triad task, four participants were excluded because of too many missed responses (more than 10).For the frame-line task, six participants were excluded because of an error in the experimental procedure, and four participants were excluded because of too many nonresponse trials.As a result, when analyzing the triad task and the frame-line task separately, the sample sizes were divided into 46 (22 females; age range: 18-30 years; mean age: 23 years) and 40 (20 females; age range: 18-30 years; mean age: 23 years), respectively.In the matched convergence analysis, 38 participants completed both experiments.The experiment was approved by the Institutional Review Board of the Institute of Psychology, Chinese Academy of Sciences.All participants signed informed consent.

Experimental design
The Frame-Line experiment (Hedden et al., 2008) is a block design including two variables: instruction (absolute task; relative task) and congruency (congruent stimuli; incongruent stimuli).The experimental stimuli are the combination of stimuli with a line in the box.The absolute instruction asks participants to judge whether the line in the present stimulus is as long as the line in the previous stimulus, which reflects analytic thinking.The relative instruction requires them to judge whether the combination of the box and the line is scaled proportionally to the previous combined stimulus, which reflects holistic thinking.Thus, in the relative instruction, participants need to focus on both the box and the line, while in the absolute instruction, participants only need to focus on the line.In the congruent blocks, responses are the same regardless of the instruction (match or non-match), while in the incongruent blocks, the answers are the opposite.Participants completed blocks, with a total of six blocks per condition.Each block consisted of seven stimuli, and participants responded from the second stimulus onwards (Fig. 1).Within each block, there was a 250-millisecond masking stimulus between each presentation, designed to conceal the preceding stimulus and create a distinct separation before the next stimulus was presented.Intervals between blocks were jittered around an average interval of approximately 15 s.In the triad task (Talhelm et al., 2014), participants need to select pictures according to the instructions.Selected items include two types.One type belongs to the same abstract category as the target picture (for example, trains and buses belong to the transportation category), which reflects analytic thinking, while the other has a functional relationship with the target picture (trains run on the track), which reflects holistic thinking (Fig. 1).The experiment is a block design that includes four conditions: free choice blocks ('Please select the image that you think matches the target image.');forced selection category blocks ('Please select the image that belongs to the same category as the target image.');forced selection relationship blocks ('Please select the picture that interacts with the target picture.');and control blocks ('Please choose your favorite picture.').The experiment consists of 16 blocks, four blocks per condition, and each block consists of seven trials.Each trial first presents a target image for one second, followed by a selection image for 2 s, with a random inter-trial interval of 1000 to 3000 ms.

fMRI preprocessing
We used the DPARSF (the Data Processing Assistant for Resting-State fMRI) module within the toolbox for Data Processing & Analysis for Brain Imaging (DPABI) software suite (Yan et al., 2016) to preprocess fMRI data.The preprocessing included the following processes.1) The initial 10-time points were discarded.2) Slice-timing correction based on the time-shift method was performed by shifting the signal measured for each slice relative to the acquisition of the slice at the mid-time of each TR. 3) Head motion correction was based on a six-parameter (rigid body) linear transformation with a two-pass procedure: first step correction to the first time point and subsequent correction to the average of all images.4) Individual T1-weighted structural images were co-registered to the mean functional images using a six-degree-of-freedom linear transformation.5) T1-weighted structural images in the individual space were segmented into grey matter, white matter, and cerebrospinal fluid.6) The DARTEL (Diffeomorphic Anatomical Registration Through Exponentiated Lie algebra) tool was used to normalize images from individual space to MNI space.7) The functional images were smoothed with a 4 mm FWHM (full-width at half-maximum) Gaussian kernel.
Y. Teng et al.

Analysis of multivariate pattern analysis (MVPA)
We used the CoSMoMVPA toolbox (http://www.cosmomvpa.org)(Oosterhof et al., 2016) for pattern classification.First, we used SPM12 for individual-level analyses, and a general linear model (GLM) was estimated for each voxel.The design matrix consists of each block and the six-movement parameters computed at the realignment stage; each block was used as a regressor.The regressors were convolved with the SPM canonical hemodynamic response function.After GLM estimation, we obtained the beta-map of the brain corresponding to each condition block.Second, based on the conditional beta-map, we used the CoS-MoMVPA for model classification.(1) Frame-line task: for the frame-line task, we used the searchlight (radius = 5) method to identify brain regions in the whole-brain activation model that could classify absolute and relative conditions by the varying weights of multiple voxels.Linear Discriminant Analysis (LDA) classifier was used.The frame-line task consisted of 12 blocks for both the absolute and relative conditions; hence, we performed a twelve-fold cross-validation.(2) Triad task: we classified categories and relationships based on the forced selection blocks for the triad task.The analysis was performed in the same way as the frame line experiment, but a four-fold cross-validation was employed, as this task had four blocks for each of the two conditions.(3) Convergence of the two tasks.We first overlapped the two brain maps obtained in frame-line task and triad task (Fig. S5).Then we combined the two experiments to classify holistic and analytic thinking (Fig. 2).Specifically, we labeled the absolute conditions in the frame-line task and the forced choice categories in the triad task as analytic thinking, while the relative conditions in the frame-line task and the forced choice relationships in the triad task were labeled as holistic thinking.We did not use participants' subjective responses in the free choice condition due to the prevalence of holistic thinking among Chinese participants, which would lead to serious deviations in the number of trials of different thinking responses.As the number of blocks for the frame-line and triad tasks differed, we randomly selected four blocks for each condition for the frame-line task to match the number of blocks in the triad task.In addition, the randomly selected four blocks for each condition (absolute/relative) consisted of two congruent blocks and two incongruent blocks.We conducted four-fold cross-validation.Finally, we performed group-level statistical analyses using one-sample t-tests to detect brain regions with classification accuracy greater than 0.5.Multiple comparisons were corrected using the Gaussian Random Field (GRF) method (voxel-level p < 0.001, cluster-level p < 0.05, one-tailed).
To further validate the accuracy of the searchlight analysis results (Etzel et al., 2013), we conducted additional ROI-based MVPA using Schaefer's 400 parcellations.The analysis process was generally consistent with the searchlight-based analysis described above.We first got the conditional beta-map using GLM in SPM12.Then we also labeled the absolute conditions in the frame-line task and the forced choice categories in the triad task as analytic thinking, while the relative conditions in the frame-line task and the forced choice relationships in the triad task were labeled as holistic thinking.Within the ROI, LDA classifier was used to distinguish the different neural representations of holistic thinking label and analytic thinking label.The cross-validation was conducted.Finally, we performed one-sample t-tests (classification accuracy greater than 0.5) at the group level and applied FDR (q < 0.05) for multiple comparison correction.
For the triad task, we first calculated the choice preferences for participants in different conditions (Fig. 4A).Participants chose more relational pairings (category: M = 9.152, SD = 7.152; relationship: M = 18.174,SD = 7.337; t = 4.237, p < 0.001) than categorical pairings under the free choice condition.Under the forced choice condition, the Fig. 2. Schematic diagram of a participant's MVPA analysis.First, the general linear model (GLM) was estimated for each voxel in SPM12 for individual-level analysis; each block was used as a regressor.Because of the inconsistent number of blocks in the two experiments, we randomly selected four blocks for the relative and absolute conditions, respectively, of the frame-line experiment.Then, the relative conditions in the frame-line task and the forced choice relationships in the triad task were labeled as holistic thinking, whereas the absolute conditions in the frame-line task and the forced choice category in the triad task were labeled as analytic thinking.The searchlight (radius = 5) method and LDA classifier were used to identify brain regions in the whole-brain activation model that could classify holistic thinking and analytic thinking by different weights of multiple voxels, and we further conducted four-fold cross-validation.Finally, one-sample t-tests for group-level statistical analysis were used to detect brain regions with classification accuracy greater than 0.5 (one-tailed, GRF correction, voxel-level p < 0.001, cluster-level p < 0.05).Furthermore, we calculated the ratio of relationship pairs selected by participants in the free-choice condition to the total number of choices to categorize them into holistic thinking or analytic thinking groups based on their behavioral performances.We found that 31 participants had a relationship selection ratio higher than 0.5 (M = 0.823, SD = 0.129), 13 participants had a relationship selection ratio lower than 0.5 (M = 0.308, SD = 0.120), and two participants had a relationship selection rate equal to 0.5.Based on the relational selection rate, we divided the participants into two groups: the relationship group and the category group.Then we conducted a two-factor mixed ANOVA to compare differences in response time between the two groups of participants in the free choice, forced selection category, and forced selection relationship conditions.The main effect of condition was significant (F (2, 84) = 23.752,p < 0.0001, partial η 2 = 0.361).Multiple comparisons showed that the differences in response time across all three conditions were significant for participants with relationship tendency, with the slowest response time in the free choice condition (M = 918.07,SD = 161.73),followed by the forced choice relationship condition (M = 851.11,SD = 144.41),and finally, the forced choice category condition (M = 746.49,SD = 144.13)(Fig. 4C).The category tendency participants followed the same ranking, but differences did not survive multiple comparison correction (free choice: M = 911.24,SD = 126.97;forced choice relationship: M = 892.30,SD = 160.78;forced choice category: 802.79,SD = 147.22)(Fig. 4D).The interaction effect was not significant (F (2, 84) = 1.254, p = 0.291, partial η 2 = 0.029) (Fig. S3).
With regard to task-fMRI analysis, both tasks showed directionality of brain regions.In the frame-line task, the absolute condition activated more brain voxels than the relative condition (Fig. S3A).In the triad task, the forced category task activated more brain voxels than the forced relationship task (Fig. S3B).

MVPA results
We used the LDA classifier for the frame line experiment to classify brain activity differences in absolute and relative conditions based on a whole brain searchlight approach.For differentiating between the two task conditions (absolute vs. relative), significant brain areas involved extensive bilateral occipital, bilateral frontal, bilateral temporal, and bilateral parietal lobes.In addition, bilateral fusiform, bilateral lingual gyrus, bilateral angular gyrus, bilateral precuneus, right insula, and sensorimotor areas were involved (Fig. 5B, Table 1).The results of the MVPA analysis using incongruency condition only and congruency condition only are shown in Fig. S4.Consistent with the frame-line task, the triad task involved bilateral frontal lobes, bilateral temporal lobes and right superior parietal gyrus, bilateral fusiform, bilateral lingual gyrus, right angular gyrus, bilateral precuneus, bilateral insula, and sensorimotor areas.In particular, the triad task involved the limbic system widely, including the cingulate gyrus, left hippocampus, bilateral parahippocampus, thalamus, bilateral caudate, bilateral putamen, and bilateral pallidum (Fig. 5C, Table 1).The overlapping brain regions for the two experiments included the bilateral occipital lobe, bilateral inferior frontal gyrus, right temporal lobe, bilateral fusiform, bilateral lingual gyrus, and right insula (Fig. S5, Table S1).
The aforementioned separate analyses of these two paradigms revealed that the brain areas involved differed by the specific task.To eliminate the influence of task-specific effects, we integrated the two tasks to identify brain areas that showed the same pattern of differences across the two tasks corresponding to the neural mechanisms underlying holistic and analytic thinking.More precisely, the absolute conditions in the frame-line task and the forced choice categories in the triad task were classified as indicative of analytic thinking, whereas the relative conditions in the frame-line task and the forced choice relationships in the triad task were classified as indicative of holistic thinking.This allowed us to identify brain regions associated with distinguishing holistic and analytic thinking (Fig. 5A, Table 1).These areas contain features of both tasks and involve more meaningful brain regions than the overlap map derived from the two tasks, including bilateral frontal lobes, bilateral parietal lobes, bilateral precentral and postcentral gyrus, bilateral supplementary motor areas, bilateral fusiform, bilateral insula, bilateral angular gyrus, left cuneus, and precuneus and left olfactory cortex.The temporal and occipital lobes showed significant left lateralization.In addition, the limbic system was involved, including the cingulate gyrus, right caudate, and putamen.Based on the ROI-based MVPA analysis, which align closely with the searchlight analysis results, we examined the correspondence of the significant parcellations with the seven Yeo networks.For detailed brain region information, please refer to the supplementary material.

Discussion
In the current study, we explored the neural mechanisms underlying holistic and analytic thinking using the classic frame-line task and triad task paradigms.Behavioral results revealed that Chinese participants in the frame-line task were more accurate in the relative condition than in the absolute condition, and responded faster in the congruent than in the incongruent condition, as expected because most Chinese individuals are holistic thinkers and the congruent condition is generally easier than the incongruent condition.In the triad task, participants were more inclined to classify according to relationship.This is consistent with the expectation that Chinese participants are more prone toward holistic thinking.MVPA results indicated distinct representation patterns in bilateral frontal lobes, bilateral parietal lobes, bilateral precentral and postcentral gyrus, bilateral supplementary motor areas, bilateral fusiform, bilateral insula, bilateral angular gyrus, left cuneus, and precuneus and left olfactory cortex between different conditions for both tasks.Additionally, the triad task also included more involvement of the limbic system.The results confirmed our hypothesis that holistic thinking and analytic thinking, as high-level cognitive functions, involved multiple brain regions including the frontal, parietal, temporal and subcortical nuclei.These findings lay the foundation for an understanding of the neural mechanisms of holistic and analytic thinking, suggesting that cultural attributes and thinking styles may shape different representations of the aforementioned brain regions.

The behavioral results
Our experimental findings concur partially with prior research (Ji et al., 2004;Kitayama et al., 2003).Regarding the behavioral outcome of the frame-line task, participants in the study by Hedden et al. (2008) exhibited faster and more accurate responses in the congruent condition, which is consistent with our results, indicating that both Eastern and Western participants allocate fewer cognitive resources and perform better in less demanding conditions.In the study by Kitayama and colleagues, Japanese participants demonstrated higher accuracy in the relative task (Kitayama et al., 2003), aligning with our findings and indicating a greater tendency for individuals from East Asian cultural backgrounds to engage in holistic thinking.Moreover, our study revealed slower response times during the relative task compared to the absolute task.This could be attributed to the need to consider contextual information during judgments, resulting in significant lengthening of response times.Regarding the triad task behavioral outcomes, participants consistently favored pairings based on relationships under free choice conditions, in line with prior studies (Ji et al., 2004), indicating a proclivity for holistic thinking among Chinese participants.Participants' response time was shorter during forced selection than during free choice selection, presumably because forced selection decreased the response options.Interestingly, response times were significantly longer during relational classification than category classification.This difference may be attributed to the need for participants to engage in inductive reasoning about various relational aspects (Lin and Murphy, 2001), such as temporal, spatial, causal, or functional relationships, leading to prolonged response times in relationship-based classification.However, reasons for differences in response time still need to be further explored.

MVPA results
MVPA of the frame-line task revealed distinct neural patterns for holistic thinking and analytic thinking in various brain regions, including bilateral occipital, frontal, temporal, parietal, fusiform, lingual gyrus, angular gyrus, precuneus, right insula, and sensorimotor areas.These brain regions partially overlapped with previous studies (Goh et al., 2010;Gutchess et al., 2006;Hedden et al., 2008).Most previous studies are cross-cultural studies, assuming that Westerners are more inclined to analytic thinking while Easterners are more inclined to holistic thinking.These findings need to be treated with caution, as there may be other cultural factors confounding the results between the different cultural groups and a particular thinking style is not an inherent attribute of a particular cultural group (Bacha-Trams et al., 2018;Julie Spencer-Rodgers et al., 2010).Our study addressed this issue by recruiting participants from the same cultural background, minimizing confounding variables.By comparing our findings with existing literature, we provided a more convincing complement to previous studies.
In the study by Hedden et al. (2008), compared to Westerners, Eastern participants showed stronger activation in the parietal, frontal, temporal, fusiform, and precentral gyrus when completing the absolute task.The fusiform is the final stage of object recognition (Yi et al., 2006) and the frontal and parietal lobes are the control brain regions for attention and working memory (Badre and Wagner, 2004).For Eastern participants, who are more inclined to take background information into account, considering only local objects and ignoring the surroundings require more attentional control.Similarly, researchers found that Eastern participants had higher activation in parietal-related brain regions when performing absolute tasks compared to Westerners, using near-infrared spectral imaging (Murata et al., 2015).Furthermore, compared to Easterners, Westerners activated the angular gyrus and right insula more during object processing (Gutchess et al., 2006).When viewing faces (object information) and houses (background information), Easterners displayed greater activation in the lingual gyrus than Westerners (Goh et al., 2010).Goh and colleagues found that younger East Asian individuals activated bilateral occipital lobes during object recognition compared with older East Asian individuals, while this effect was not present in younger and older Western individuals (Goh et al., 2007).Another study examining the frame-line task among older and younger individuals from different cultures found that older East Asian individuals exihibited lower levels of holistic thinking compared to younger individuals, while this effect was also absent in Western participants (Zhang et al., 2014).These two behavioral and brain imaging experimental results converge to support that the occipital lobe of young Chinese participants showed different representational patterns across conditions in the frame-line task in our study.In addition, in the rs-fMRI study, the degree of the bilateral precuneus was significantly and negatively correlated with participants' AHS scores (Luo et al., 2022).Our results similarly validate the role of the precuneus in holistic and analytic thinking.
In particular, the MVPA results for the triad task revealed additional limbic system areas, including the cingulate gyrus, left hippocampus, bilateral parahippocampus, thalamus, bilateral caudate, bilateral putamen, and bilateral pallidum, overlapping with previous studies (Luo et al., 2022;Sachs, Weis, Krings, et al., 2008;Wang et al., 2016).Compared to Easterners, Westerners activated more left parahippocampal gyrus when classifying according to relationship, and activated more left precuneus and right thalamus when classifying according to category (Sachs, Weis, Zellagui, et al., 2008).In addition, Wang and colleagues found that self-reported scores were significantly and positively correlated with neural activity in the anterior cingulate gyrus, as well as with functional connectivity between the anterior cingulate gyrus and hippocampus and caudate nucleus (Wang et al., 2016).As another example, the degree and nodal global efficiency of putamen were significantly and positively correlated with AHS scores (Luo et al., 2022).All these studies provide supporting evidence for the involvement of the limbic system in the triad task.
Combining both paradigms using MVPA revealed unique brain areas, such as the bilateral precentral and postcentral gyrus, bilateral supplementary motor areas, and left olfactory cortex, which were not identified in separate analyses.Easterners had stronger activation in the precentral and postcentral gyrus than Westerners when classifying according to relationship (Gutchess et al., 2010).In addition, in the self-referential paradigm, another classical paradigm of dialectical thinking, individuals had higher activation in the precentral and postcentral gyrus when dealing with contradictory information versus non-contradictory information processing (Wang et al., 2016).These studies all confirm the involvement of the anterior and posterior central gyrus in the neural activity of holistic and analytic thinking.Supplementary motor areas play an important role in the programming and conscious execution of detailed actions, linking medial limbic cortex with primary motor cortex (Goldberg, 1985).In the previous literature, holistic and analytic thinking involved the limbic cortex and primary motor cortex (Wang et al., 2016).This may explain why supplementary Note: Brain regions with voxel counts greater than 20 were shown based on AAL3 template (Rolls et al., 2020).
Y. Teng et al.
motor areas are involved in the neural activity of holistic and analytic thinking.As an ancient homologous cortex, some olfactory cortex cells have been found to participate in encoding spatial information, a key hippocampal function (Poo et al., 2022).This result may partly explain why olfactory cortex are involved in holistic and analytic thinking processing.

MVPA results from functional network perspective
Through MVPA, we delineated a whole-brain scale brain map of holistic and analytic thinking, mitigating the impact of unique task characteristics.While our results implicate multiple brain regions, the actual voxel count remains relatively modest.To better explain the multiple neural and cognitive functions of the holistic and analytic thinking brain map, we calculated its proportion from a functional network system perspective (Yeo et al., 2011), operationalized as the number of resultant voxels per network/total number of resultant voxels (Fig. S6).The results showed that the visual network, somatomotor network (SMN), and frontoparietal network (FPN) accounted for 55.6 %, 15.9 %, and 11.6 % of the identified voxels, respectively.The prominence of visual networks aligns with expectations, given that holistic and analytic thinking represent top-down thought patterns influencing information processing, primarily originating from the visual cortex.Previous studies have found significant cross-cultural differences in visual cortex representations (Brodmann 18 and 19) through MVPA, during the observation of everyday pictures (Ksander et al., 2018).There are also significant cultural differences between Easterners and Westerners in the initial eye fixation of face recognition.Though this conclusion has been questioned (Or et al., 2015), the consensus is that Eastern individuals differ in eye fixation patterns during face recognition compared to Western individuals (Tardif et al., 2017).These findings support our conclusion that holistic and analytic thinking have a profound impact on visual information processing.In addition to visual information, according to the embodied cognition theory, people also interact and transmit information from the external environment through the sensory and motor experiences of the body (Novembre et al., 2019).Therefore, holistic thinking and analytic thinking, as high-level thinking modes, also affect the processing of physical and sensory information.Behavioral experiments showed that holistic and analytic thinking affect people's division of continuous sensory experience.Compared with Indian individuals, American individuals tend to segment sensory memories into a greater number of events (Swallow and Wang, 2020).Finally, the executive control function of FPN, which plays an important role in engagement in top-down control strategies, is indispensable to the influence of holistic and analytic thinking on multi-channel information processing (Gutchess et al., 2010).

The cognitive functions of MVPA results using neurosynth
We used Neurosynth to decode the cognitive implication of MVPAsignificant brain regions via NiMARE (Salo et al., 2018) in Python.We performed a functional meta-analysis of the significant brain regions obtained by MVPA using the Neurosynth database (Yarkoni et al., 2011).Please refer to the supplementary materials for detailed procedures.The brain map based on the convergence of the two tasks shows more relevant cognitive terms, including visual, reading, word, encoding, phonological, objects, production, semantic, recognition, lexical, etc. (Fig. S7A).For the frame-line task, the implicated brain regions showed more relevant cognitive terms, including visual, reading, word, encoding, objects, retrieval, memory, phonological, etc. (Fig. S7B).For the triad task, the implicated brain regions showed more relevant cognitive terms including retrieval, encoding, episodic, motor, memory, engagement, reappraisal, etc. (Fig. S7C).
This additional analysis allows us to clarify that holistic thinking and analytic thinking are complex high-level thinking styles, involving multiple interrelated successive cognitive processes.Whether individuals engage in holistic or analytic thinking, the relevant brain regions provide cognitive resources for processing external information, be it visual or semantic, ultimately generating distinct thinking modes in the task.Furthermore, the brain regions involved in holistic thinking are also responsible for processing language, such as 'reading', 'word', and 'phonological'.This interesting conclusion reminds us of the Sapir-Whorf hypothesis, which articulated the central idea that culture influences thinking through language (Ji et al., 2004).Language, an abstract system of meaning that organizes and internalizes thought, plays a crucial role in explaining cultural variations in thinking styles (Chen et al., 2014).Logan argues that reading and writing with the alphabet is not just for social communication, but provides an analytical framework for understanding reality anew (Logan, 1986).Phonetic alphabets are abstract, logical, and systematic, and provide the basis for the development of Western analytic thinking.Conversely, Chinese characters are pictorial and concrete, rather than abstract concepts, creating a distinct thinking style compared to the Western approach.This provides a potential explanation for the relationship between brain regions involved in holistic and analytic thinking and language processing.In addition, we found differences in the cognitive terms associated with brain areas corresponding to the two paradigms.In addition to the mentioned processing functions, cognitive processes like 'memory' are integral to the frame-line task.This is likely due to its connection with attentional processing, where recalling the original stimulus is necessary for comparison and judgment.Similarly, the triad task encompasses 'retrieval,' 'episodic,' 'encoding,' and 'motor' processes.This is essential for classifying pictures, requiring the extraction of stored information about objects and subsequent analysis for classification based on episodic information.Therefore, the differences in brain regions involved in the two paradigms are likely to be more related to the nature of the specific task.Future research should explore the precise reasons for the differences in brain areas involved in the two paradigms.

Contributions, limitations and future expectations
Our study contributes to the field of cultural neuroscience.First, by using two classical holistic thinking paradigms, we identify the overlapping and fundamental brain regions involved in holistic and analytic thinking.We use MVPA to clarify which brain regions can effectively distinguish between two different modes of thinking based on brain activity patterns under two different cognitive conditions.We further explain the cognitive functions of these brain regions from the perspective of functional brain networks.Furthermore, we provide initial neural evidence for the Sapir-Whorf hypothesis, revealing the neural basis of culture-related thinking styles involved in language function, suggesting that the differences in thinking styles between East and West may arise from their different language systems.Finally, our research provides a new approach to examine the representation of cultural constructs in the brain, deepening our understanding of how culture shapes the brain and sheds light on the sociocultural nature of the human brain (Han et al., 2013).
Nevertheless, our study has some limitations.First, our participants were all Chinese, and future studies should conduct cross-cultural comparisons to better contrast holistic and analytic thinking.Second, we only used two classical holistic thinking paradigms.Future work should include additional paradigms to validate and extend our results.Third, our analysis focused on brain regions associated with holistic and analytic thinking.However, cognition and behavior are seen as the result of brain synergy (de Schotten and Forkel, 2022).Future studies should further consider the associations among brain regions to construct neural networks underlying holistic thinking and analytic thinking.Finally, the current study attempted to identify brain regions associated with both holistic and analytic thinking by integrating experiments at two different cognitive levels.However, it remains unclear which of the identified foundational brain regions play a more direct role in holistic thinking and which are more directly involved in analytic Y. Teng et al. thinking.In other words, it remains to be determined whether certain brain regions are exclusively associated with holistic thinking, while others are solely linked to analytic thinking.Additionally, the present study did not succeed in predicting individual tendencies toward holistic or analytic thinking.Further research is needed to delve deeper into these aspects.

Conclusion
Combining two classical holistic and analytic thinking paradigms, the frame-line task and the triad task, we delineated the fundamental brain regions involved in distinguishing holistic and analytic thinking through MVPA, including the bilateral frontal lobe, bilateral parietal lobe, bilateral anterior and posterior central gyrus, bilateral supplementary motor area, bilateral fusiform gyrus, bilateral insula, bilateral angular gyrus, left precuneus, left olfactory cortex, cingulate gyrus, right caudate and putamen.Using Neurosynth, we further identified the cognitive functions associated with these brain regions and decoded this complex cultural construct of holistic and analytic thinking from physiological and cognitive levels.

Fig. 1 .
Fig. 1.Schematic diagram of the procedure for the task-fMRI experiment.

Fig. 4 .
Fig. 4. Triad task behavioral results.(A) Comparison of frequency distributions for selecting categories versus relationships under different conditions.(B) Differences in response times between forced selection categories and forced selection relationships.(C) Comparison of response time differences between the freechoice and the two forced-choice conditions for participants with relationship tendency (N = 31).(D) Comparison of response time differences between the freechoice and the two forced-choice conditions for participants with category tendency (N = 13).Data were means with 95 % CI. *** p < 0.001; **** p < 0.0001.

Fig. 5 .
Fig. 5. MVPA results.The brain maps of T-values showed results of one-sample t-tests on accuracy rate (Base = 0.5, one-tailed, GRF correction, voxel-level p < 0.001, cluster-level p < 0.05).The top three graphs were the results of the searchlight analysis, and the lower three graphs were based on the results of the Schaefer 400 parcellation.Searchlight, Radius = 5; classifier: LDA.(A, C) Results of the holistic and analytic condition classification by converging the two tasks.(B, D) Results of the absolute and relative condition classification of the frame line task.(C, F) Results of the forced choice category and forced choice relationship classification on the triad task.