Functional connectivity with medial temporal regions differs across cultures during post-encoding rest

Abstract

Connectivity of the brain at rest can reflect individual differences and impact behavioral outcomes, including memory. The present study investigated how culture influences functional connectivity with regions of the medial temporal lobe. In this study, 46 Americans and 59 East Asians completed a resting state scan after encoding pictures of objects. To investigate cross-cultural differences in resting state functional connectivity, left parahippocampal gyrus (anterior and posterior regions) and left hippocampus were selected as seed regions. These regions were selected, because they were previously implicated in a study of cultural differences during the successful encoding of detailed memories. Results revealed that left posterior parahippocampal gyrus had stronger connectivity with temporo-occipital regions for East Asians compared with Americans and stronger connectivity with parieto-occipital regions for Americans compared with East Asians. Left anterior parahippocampal gyrus had stronger connectivity with temporal regions for East Asians than Americans and stronger connectivity with frontal regions for Americans than East Asians. Although connectivity did not relate to memory performance, patterns did relate to cultural values. The degree of independent self-construal and subjective value of tradition were associated with functional connectivity involving left anterior parahippocampal gyrus. Findings are discussed in terms of potential cultural differences in memory consolidation or more general trait or state-based processes, such as holistic versus analytic processing.

It is now well-established that the brain at rest, instead of being quiet and static, is in fact actively engaged (Raichle et al., 2001). Resting-state functional connectivity (RSFC) has been proposed to reflect the intrinsic functional-anatomical networks of the brain and is associated with individual differences in cognitive capabilities (Stevens & Spreng, 2014). On a more spontaneous level, RSFC also may support the dynamic unfolding of mnemonic and other cognitive processes, including reprocessing past experiences and preparing for future tasks depending on the goals and priorities of the individual (Buckner & Vincent, 2007; Foster & Wilson, 2006; Gordon et al., 2014; Meyer et al., 2019; Miall & Robertson, 2006; Raichle & Snyder, 2007; Stevens et al., 2010; Stevens & Spreng, 2014; Tambini et al., 2010; Zabelina & Andrews-Hanna, 2016).

One way in which past experiences could shape resting state activity of the brain is through culture. People from different cultural backgrounds have been found to vary in perceptual, socioemotional, and cognitive processes (Ksander et al., 2018; Markus & Kitayama, 1991; Masuda et al., 2008; Rule et al., 2013), including memory specificity (Leger & Gutchess, 2021; Millar et al., 2013; Paige et al., 2017b). Extended, immersive experiences associated with a culture, potentially including which information and styles of information processing are prioritized, could potentially shape the connectivity of networks in distinct ways across cultural groups. Little research thus far has examined the ways in which neural connectivity may differ across cultures, with a lack of research associating these patterns with cognitive processes. Further examining cultural influences during rest could help to characterize ways in which mental experiences differ across cultures. In terms of cognition and memory, preceding tasks can shape the engagement of networks during rest, including strengthening memory traces (Lewis et al., 2009; Meyer et al., 2019; Stevens et al., 2010; Tambini et al., 2010). Using resting state functional connectivity (FC) analysis, the present study examined cultural differences in connectivity at rest, with a focus on medial temporal lobe regions previously identified as exhibiting cultural differences during the encoding of specific memories.

Cultural influences on RSFC

Existing research in the field of cultural neuroscience have shown compelling evidence for cultural differences in the engagement of relatively localized brain regions during cognitive tasks. However, the ways in which connectivity between brain regions at rest, assessed using RSFC, could be shaped by chronic cultural experiences has been rarely investigated. Thus far, cross-cultural research on RSFC has largely focused on the impact of independent versus interdependent self-construal (i.e., an individual’s subjective definition of the self in relation to others, or as an independent entity). Across individuals, higher levels of independence are positively correlated with within-network connectivity of the default mode network (DMN) and the executive control network, but negatively correlated with the connectivity between the two networks (Li et al., 2018). Using a machine learning approach, independence was found to be primarily associated with right hemisphere functional connectivity whereas interdependence was primarily associated with left hemisphere functional connectivity (Zhu et al., 2021). In a study by Wang et al. (2013), chronic self-construal inclinations correlated with RSFC in the DMN. Specifically, participants who scored higher in interdependent self-construal showed greater RSFC between the ventral medial prefrontal cortex (vMPFC) and dorsal medial prefrontal cortex (dMPFC) compared with those who scored higher in independent self-construal. In turn, the independent self-construal group showed greater RSFC between the vMPFC and PCC.

Taken together, these findings serve as evidence that culture, defined through self-construal in these studies, could modulate functional connectivity. However, potential differences in functional connectivity related to memory processes are unclear. By examining functional connectivity in brain regions associated with memory specificity in American and East Asian participants, the current study will explore potential ways in which functional connectivity could contribute to cultural differences, including in memory specificity.

Cultural differences in analytic-holistic processing and memory

Cultural differences in analytic versus holistic approaches to information processing are robust. That is, whereas people from Western cultures tend to understand a scenario by focusing on the central objects and its details and processing each component on its own, Easterners are inclined to pay more attention to contextual information and the configural relationship between different parts of the scene (Masuda & Nisbett, 2001; Miyamoto et al., 2011; Nisbett et al., 2001; Wang et al., 2012).

In terms of memory, preference for analytic versus holistic information processing may shape the allocation of attention to different information in the environment (Boduroglu et al., 2009) and eventually affect subsequent memory for the information (Chiao, 2009; Chiao & Harada, 2008; Gutchess & Indeck, 2009; Han & Northoff, 2008), particularly in terms of the details contained in memory (Schacter et al., 2009). Westerners, adopting an analytic approach, typically prioritize object processing, making them more likely to remember visual details of an object. In contrast, Easterners, who are more holistic, pay more attention to the context and relations, and therefore remember less about the specific features of the focal object (Millar et al., 2013).

One aspect of memory that differs across cultures is in the amount of detail, or memory specificity. Memories can differ in the amount of specific detail or general “gist” that they contain (Garoff et al., 2005). For example, one could remember the general theme of a movie or recall the specific details of the scenery or a character’s facial expressions and manner of speech. Memory specificity has been shown to be modulated by culture, with Americans remembering more details than East Asians of visual objects (Leger & Gutchess, 2021; Millar et al., 2013) and in autobiographical memory (Wang, 2006, 2009), perhaps reflecting prioritization of analytic versus holistic processing.

Cross-cultural investigations on the neural underpinnings of memory specificity provide clues for the regions differently engaged across cultures during memory encoding (Gutchess et al., 2006; Paige et al., 2017b). Paige et al. (2017b) found greater activations in the left fusiform, left anterior and posterior parahippocampal gyrus (PHG), and the left hippocampus in East Asians when encoding specific versus general memories (e.g., supporting the ability to later recognize that one studied this specific exemplar of an apple, as opposed to generally recognizing that one studied an apple, regardless of the specific details), which lays the groundwork for the current investigation of RSFC of these regions after encoding. Although behavioral differences in memory performance did not emerge in the Paige et al. (2017b) study, which may reflect the demands of the scanner environment, as cultural differences in performance on this task, and variants of it, have been demonstrated in multiple other studies (Millar et al., 2013; Paige et al., 2017a; Leger & Gutchess, 2021; see also Gutchess & Sekuler , 2019 for a discussion). The authors speculated that cultural differences could reflect the additional processing needed for East Asians to form detailed memories at the same level of memory specificity as Americans.

Post-task rest and memory

RSFC following a task (referred to in this manuscript as post-task RSFC) reflects cognitive demands of the task and usually engages regions mobilized during the task. Functional connectivity can reflect how the brain “winds down” after effortful cognitive tasks. As demonstrated by Gordon et al. (2014), speed of performance in an N-back task as well as the connectivity involving the task positive network (TPN) was able to predict functional connectivity during the subsequent resting state phase. This pattern implies a positive relationship between task demands and the stabilizing procedures it takes to recover from the task afterwards. Similarly, many other studies also reported modulated RSFC in regions corresponding to the task or training such as motor or visual learning that took place in recent past (Albert et al., 2009; Grigg & Grady, 2010; Hasson et al., 2009; Lewis et al., 2009; Stevens et al., 2010; Tambini et al., 2010; Tung et al., 2013; Waites et al., 2005). Some of them showed behavioral outcomes correlated with the strength of RSFC (Lewis et al., 2009; Schlichting & Preston, 2014; Stevens et al., 2010; Tompary et al., 2015).

In terms of memory, spontaneous reactivation can occur during periods of rest, which have been interpreted as reflecting replay of events (Schapiro et al., 2018; Wilson & McNaughton, 1994). Tambini et al. (2010) supported the interpretation that spontaneous activity at rest can support memory. They demonstrated that stronger coupling between hippocampus and lateral occipital complex at rest predicted better subsequent memory at retrieval, even when the later memory test was a surprise and when controlling for preencoding connectivity as a baseline. Likewise, connectivity between regions of the default network and temporoparietal junction predicts higher levels of social recognition memory (Meyer et al., 2019).

Thus far, evidence suggests that RSFC can be modulated by recent tasks, especially in functionally relevant regions, and often is predictive of subsequent memory outcomes. To focus on processes during rest that potentially relate to memory, we selected regions that were differentially engaged across cultures during a memory encoding task (left hippocampus and left PHG, based on the results of Paige et al., 2017b) and assessed their functional connectivity to the rest of the brain during a resting period immediately following encoding. Without a baseline resting state scan, results from a single resting state scan cannot be conclusively associated with post-encoding processes rather than general cognitive processes. Nevertheless, our focus on memory regions increases the likelihood that any cultural differences that emerge during rest could impact memory. Moreover, revealing differences across cultures in functional connectivity that generally occur across tasks has indirect implications for memory, through the potential for cultural differences in what is attended to and the cognitive processes (e.g., analytic vs. holistic information processing styles) applied to that information.

The dearth of research linking cross-cultural differences in memory and neural activity (FC or task-based fMRI) limited our ability to make specific predictions about which regions would show functional connectivity differences across cultures. However, the use of visual stimuli and past studies of periods of rest following the encoding such stimuli (Tambini et al., 2010) implicate visual regions. Furthermore, considering the selected regions were engaged in East Asians more than Americans, East Asian participants might be expected to exhibit stronger FC between the target regions and other brain areas compared to American participants, reflecting the recapitulation of processes during encoding. Conversely, it was possible that Americans could exhibit stronger connectivity, perhaps reflecting consolidation that supports their typically higher recognition memory performance for detailed objects (Leger & Gutchess, 2021; Millar et al., 2013).

Methods

Participants

The initial dataset consisted of 51 Americans, recruited from Brandeis University and the greater Boston area, and 59 East Asians, recruited from National Taiwan University and the surrounding area. This is an independent dataset from Paige et al. (2017b), which was used as the basis for selecting seed regions. American participants were native English speakers, who were not of Asian descent, and who lived entirely in the United States their whole lives or lived outside of the United States for no more than 5 years. East Asian participants were native Mandarin speakers who lived entirely in Taiwan their whole lives or lived outside of Taiwan for no more than 5 years. Three American participants were excluded due to excessive motion (more details in Image Acquisition and Preprocessing section). Two more American participants were excluded for not fitting the inclusion criteria (not native to the United States). After exclusion, the sample consisted of 46 Americans (age 18-30 years; M_age = 21.67, SD_age = 3.46; M_education = 15.50, SD_education = 2.07; 24 males) and 59 East Asians (age 19-32 years; M_age = 23.31, SD_age = 2.35; M_education = 16.81, SD_education = 1.89; 30 males). East Asian participants were slightly older in age, t (103) = 2.88, p = 0.005, and higher in years of education, t (103) = 3.38, p = 0.001. All participants were right-handed, free of recent history of neuroactive medications, and neurological/psychological conditions that might impact MRI data.

Procedure

In the MRI scanner, participants initially completed an incidental encoding task during the acquisition of the high-resolution T1 image. In this task, they viewed 128 photos of familiar everyday objects (adapted from Mnemonic Similarity Task; Kirwan et al., 2007; Stark et al., 2015). Items were normed on familiarity across cultures. For the norming, 51 American and 57 Taiwanese older adults¹ each rated a subset of items. Each item was rated for familiarity on a 1-5 scale, and participants generated a name for the item. Items selected for the study were each rated an average of 3 of 5 by each group and had acceptable names generated, as determined by a bilingual coder.

In order to encourage participants to attend to the task, they decided whether each item belongs indoors or outdoors. The resting-state fMRI scan immediately followed the encoding scans; participants were instructed to keep their eyes open and try not to fall asleep while viewing a white fixation cross against a black background. After the resting-state scan, memory was tested using the same stimuli as during encoding, similar exemplars to those studied during encoding (e.g., a different bicycle than the one studied originally), and new unstudied pictures (64 items in each category). This design allowed for the assessment of neural regions involved in pattern separation at the time of retrieval (an ongoing study, separate from the current study).

In a separate session prior to the scan, participants completed demographic, neuropsychological, and individual difference questionnaires, administered in their native language. Measures included the Singelis Self-Construal Scale (SCS; Singelis, 1994) and the Schwartz Value Survey (SVS; Schwartz, 1992), which will be used for exploratory analyses. For the SCS, participants rated their endorsement of items related to the independent (“I enjoy being unique and different from others in many respects”) or interdependent (“I will sacrifice my self-interest for the benefit of the group I am in”) self on a 7-point scale from “strongly disagree” to “strongly agree.” For SVS, participated rated their endorsement of items related to 10 cultural values (e.g., tradition, hedonism) on a 9-point scale, ranging from “opposite of what I value” to “extremely important.” Participants’ responses to items are averaged for each subscale.

Image acquisition and preprocessing

In the United States, imaging data were acquired at the Center for Brain Science at Harvard University in Cambridge, Massachusetts. In Taiwan, imaging data were acquired at the Imaging Center for Integrated Body, Mind, and Culture Research, National Taiwan University, Taipei. Both sites used 3.0 T Siemens MAGNETOM Prisma whole-body MRI systems (Siemens Medical Solutions, Erlangen, Germany) with 64-channel head coils. Because each culture is scanned at a separate location, we first ensured the comparability of scanners across sites. Four participants (who did not participate in the present study) completed a calibration procedure in which each participant was scanned in multiple sessions on each of the two scanners. Calibration analyses indicated that global signal did not meaningfully differ across scanners and comparisons of activations were comparable across scanners. Differences in activations across sites were limited to primary visual cortex, likely reflecting differences in screen luminance that could not be further equated across sites (Chen et al., 2020; see also Lee et al. 2021 for a comparison of cultural groups using these scanners). Our comparison of these two scanners is in line with prior work indicating that site can play a small role in accounting for differences and that between-subject variance can be much larger than between-site variance (Sutton et al., 2008). Although prior work has focused on comparisons of activations across scanners and sites, there is no reason to believe that covariances, which functional connectivity relies on, differ across sites. Indeed, cross-site studies has become more prevalent nowadays (e.g., ADNI; ABCD; HCP projects) and studies have published functional connectivity data collected from multiple sites (Jann et al., 2015; Noble et al., 2017; Van Horn & Toga, 2009).

The resting state run used an echo-planar sequence that employed multiband RF pulses and Simultaneous Multi-Slice (SMS) acquisition (Feinberg et al., 2010; Moeller et al., 2010; Setsompop et al., 2012; Xu et al., 2013). The SMS technique allowed the acquisition of 8 slices simultaneously, increasing temporal resolution considerably over conventional techniques. 64 total slices were acquired at repetition time (TR) = 650 ms, echo time (TE) = 34.80 ms, flip-angle (FA) = 50°, field of view (FOV) = 220 × 220 mm, and 2.3- × 2.3- × 2.3-mm voxel resolution for a total of 626 scans (6 minutes and 46.9 seconds). Slices were aligned with a 20° tilt (upward at the anterior location) relative to the anterior-posterior commissural plane and positioned for whole-brain coverage. The SMS-EPI acquisition used the CMRR-MB pulse sequence from the University of Minnesota. The resting-state fMRI sequence was collected immediately after the encoding task described earlier. A high-resolution, T1-weighted, magnetization-prepared rapid gradient echo image (multi-echo MPRAGE: van der Kouwe et al., 2008) was acquired for normalization to standard space with 176 sagittal slices, voxel size 1.0 × 1.0 × 1.0 mm, FOV = 256 × 256 mm, TR = 2,530.0 ms, short TE = 1.69 ms, long TE = 7.27 ms, and FA = 7°.

The resting-state fMRI data were preprocessed using the CONN Toolbox (Whitfield-Gabrieli & Nieto-Castanon, 2012; http://www.nitrc.org/projects/conn; RRID:SCR_009550), implemented using MATLAB 2019a and Statistical Parametric Mapping 12 (SPM12; Wellcome Trust Centre for Neuroimaging, London; http://www.fil.ion.ucl.ac.uk/spm12/). To avoid including unstable images, the first 12 resting-state scans were removed. The data then went through realignment, unwarping, segmentation, normalization (to MNI space), smoothing (Gaussian smoothing kernel = 8 mm), and outlier detection for overall intensity and motion (with a conservative setting: global-signal z-value threshold = 3; framewise displacement motion threshold = 0.3 mm). We used Diffeomorphic Anatomical Registration Through Exponentiated Lie algebra (DARTEL) to create an anatomical template appropriate for our two samples (Ashburner, 2007). DARTEL is an optimized registration toolbox used to achieve sharper nonlinear registration than the native SPM registration method. Using a sample of 46 American and 59 Taiwanese brains, we created a template for normalization that should be more appropriate across the two cultural groups than the standard template brains. During normalization, functional images of the resting state were then normalized according to the DARTEL template.

Global signals were regressed out of the resting state functional images. Motion outliers also were included as covariates in the overall GLM. Nonneural noise was estimated by applying the aCompCor method in the CONN toolbox to estimate and regress out signal in subject-specific white matter and CSF regions of interest, after which a linear detrending and a BOLD signal band-pass filter of (0.008-0.09 Hz) was applied to isolate low-frequency fluctuations that characterize resting-state BOLD signals (Fox et al., 2005, 2006; Waheed et al., 2016). The Quality Assurance (QA) plots after each step were visually inspected for quality control purposes. Three American and 4 East Asian participants were excluded for having too few valid timepoints (less than 1stQ - 1.5*IQR valid scans based on 0.3 mm framewise displacement threshold; adapted from Kark et al., 2021; IQR: interquartile range) or too much motion (greater than 3rdQ + 1.5*IQR mean of motion). After preprocessing, each participant’s data contained at least 5 minutes 50 seconds of valid scan (M = 6 minutes 26 seconds). The number of excluded scans due to excessive framewise displacement did not differ significantly across cultures, t (82) = 0.49, p = 0.62.

Regions of interest

Resting-state fMRI analyses were performed using the CONN Toolbox (Whitfield-Gabrieli & Nieto-Castanon, 2012). Regions were selected as the focus of connectivity analyses based on the medial temporal regions that showed cultural differences during the formation of specific, as opposed to general, memories in a previous encoding study (Paige et al., 2017b). In Paige et al. (2017b), neural activity associated with the encoding of items subsequently recognized as “same” versus those subsequently recognized as “similar” was compared across American and East Asian participants. Those analyses yielded cultural differences in left hippocampus, left PHG, and left fusiform whereby East Asians were found to activate these regions more than American participants when correctly encoding details of objects; no regions were activated more in American participants for this contrast. For the present investigation, seed-to-voxel analyses were conducted using the left hippocampus and left PHG as seed ROIs, as defined in Harvard-Oxford subcortical atlas (Desikan et al., 2006). After close examination, the left fusiform (peak at [−36,−10,−20]) was not included, because upon inspection, this peak was close to PHG (moving only one unit to the left, the coordinate [−37,−10,−20], is labeled as left anterior PHG), leading us to question whether the activation truly reflected fusiform activity. Therefore, we focused on the PHG, because two activations were more squarely located in the region, with one peak in anterior left PHG (−36,−10,−20) and a second in posterior left PHG (−14,−26,−16). Thus, the two portions of the PHG were used as two separate seed regions. The hippocampus served as a third region, based on the peak of (−24,−14,−16). The locations of the seed ROIs are shown in Fig. 1.

Fig. 1.

Three ROIs used for functional connectivity analysis, based on the Harvard-Oxford Atlas

To conduct the analysis, the mean BOLD signal was computed across all voxels within each of the three ROIs. A general linear model was adopted to assess the temporal correlation between the seed ROIs and each voxel in the rest of the brain. Next, the map of correlation coefficients underwent a Fisher r-to-Z transformation for the subsequent group-level t tests. We adopted a peak threshold of p < 0.001 and a cluster threshold p < 0.05 (FDR-corrected).

Correlation between FC and memory performance

To further probe the relationship of FC to memory specificity performance, we conducted correlation analyses for regions in which cultural differences emerged. These analyses examined the relationship between the strength of FC with memory specificity performance in the behavioral task. Memory specificity performance was measured using d’ scores for the same versus similar items. According to signal detection theory, d’ measures the discriminability of an individual between signal and noise (Stanislaw & Todorov, 1999). The d’ for same versus similar items hence represents the sensitivity to same items as correct signals from the interfering signal of similar ones. Therefore, the better one is able to distinguish same items from similar ones, the more specific and detailed the memory is.

Results

Resting-state FC

We observed cross-cultural differences in FC in two of the three regions of interest (left anterior and posterior PHG but not left hippocampus; see Table 1 for a full list of the peak coordinates in regions exhibiting cultural differences). East Asian participants, who activated left hippocampus and left PHG during encoding of specific information more than American participants in the previous study (Paige et al., 2017b), exhibited stronger FC between left posterior PHG and a wide range of posterior brain regions focused on bilateral cuneus and lingual gyrus, extending along the ventral stream into the temporal fusiform area. See Fig. 2 for visual depiction of the regions and graphs of connectivity strength. Each of the three clusters demonstrated positive connectivity with left posterior PHG in East Asian participants to a greater extent than in American participants; even though the connectivity between these regions was positive in Americans, the strength of the connectivity was not as large as in East Asians (as depicted in Fig. 2). In terms of regions that had stronger connectivity for American participants compared to East Asians participants, there was stronger FC between left posterior PHG and a number of parieto-occipital regions spanning lateral occipital cortex and superior parietal lobule. See Fig. 3 for an illustration of the regions and graphs of connectivity strength. Functional connectivity with the posterior PHG of East Asians was close to, but significantly different from, zero, t (58) = 2.22, p = 0.03, and numerically negative.

Table 1.

Full list of brain regions functionally connected to left hippocampus and left PHG during rest post-encoding that survived the threshold of p_peak < 0.001 and p_cluster < 0.05

	k	t	BA	(x, y, z)
Region (connected to left PHG, posterior)
East Asians > Americans
Bilateral cuneal cortex; right intracalcarine cortex; precuneous cortex; lateral occipital cortex; right lingual gyrus; right supracalcarine cortex (cluster A in Fig. 2)	988	4.80	18	[8;−74;20]
Right lingual gyrus; right temporal occipital fusiform cortex (cluster C in Fig. 2)	521	5.18	37	[22;−48;−6]
Left temporal occipital fusiform cortex, left lingual gyrus; left intracalcarine cortex (cluster B in Fig. 2)	376	4.99	37	[−22;−54;−14]
Americans > East Asians
Right superior lateral occipital cortex; right superior parietal lobule (Fig. 3)	286	5.48	19	[28;−60;34]
Region (connected to left PHG, anterior)
East Asians > Americans
Right temporal pole; right anterior inferior temporal gyrus, temporal fusiform cortex, middle temporal gyrus, and PHG (cluster A in Fig. 4)	1031	6.59	36	[36;6;−38]^
Left temporal pole; left anterior middle and inferior temporal gyrus (cluster B in Fig. 4)	369	5.55	20	[−50;2;−32] ^
Americans > East Asians
Left middle frontal gyrus; left frontal pole; left pars triangularis inferior frontal gyrus (Fig. 5)	644	5.76	45	[−46;38;20] #
Region (connected to left hippocampus)
East Asians > Americans: None
Americans > East Asians: None

Notes. PHG = parahippocampal gyrus; BA = Brodmann area. The (x, y, z) coordinates are in MNI space.

^{^}Regions were significantly associated with adjusted independence scores (independence – interdependence). See the section “SCS” under “Correlation between FC and Cultural Values and Self-Construal.”

^#Region was significantly associated with adjusted independence scores (independence – interdependence) and SVS-Tradition. See the section “SVS” under “Correlation between FC and Cultural Values and Self-Construal.”

Fig. 2.

Functional connectivity analyses of the left posterior PHG region identified three clusters with stronger connectivity in East Asian versus American participants. These clusters—(A) spanning cuneus, precuneus, and right intra- and supracalcarine cortex; (B) spanning left temporal occipital fusiform cortex and lingual gyrus; and (C) spanning right temporal occipital fusiform cortex and lingual gyrus—are located primarily in occipital regions and depicted in the top row of the figure (see Table 1 for coordinates and additional details). Cross-cultural comparisons of connectivity strengths (bottom row) indicated that the connectivity between the left posterior PHG and each of these regions was greater for East Asians than Americans. All FCs are positive

Fig. 3.

Functional connectivity analyses of the left posterior PHG identified stronger connectivity for Americans than East Asians in right superior lateral occipital cortex and superior parietal lobule (depicted in the left panel). Cross-cultural comparisons of connectivity strengths (right panel) indicated that the connectivity between the left posterior PHG and this cluster was greater for Americans than East Asians. The FC of Americans is greater than zero and that of East Asians is close to zero

Analysis of FC to the left anterior PHG revealed that East Asians demonstrated stronger connectivity between left anterior PHG and a number of temporal regions, such as bilateral temporal pole, bilateral inferior temporal gyrus, and right temporal fusiform cortex (Fig. 4). Mean FC are greater than zero in both cultures. Figure 4 illustrates the two clusters along with the connectivity strength by culture. Americans showed stronger FC between the left anterior PHG and a series of frontal regions in the left hemisphere, including the left middle frontal gyrus, which is depicted in Fig. 5. The connectivity between the left anterior PHG and this cluster was positive for Americans and negative for East Asians.

Fig. 4.

Functional connectivity analyses of the left anterior PHG (L aPHG) region identified two clusters with stronger connectivity in East Asian versus American participants. These clusters—(A) spanning right temporal pole, temporal gyrus, and fusiform, (B) spanning left temporal pole and temporal gyrus—are located in temporal regions and depicted in the top left panel. Cross-cultural comparisons of connectivity strengths (top middle and top right panels) indicated that the connectivity between the left anterior PHG and each of these regions was greater for East Asians than Americans. Both FCs involving cluster A and B showed a negative correlation with the SCS (adjusted independence score; bottom panels)

Fig. 5.

Functional connectivity analyses of the left anterior PHG (L aPHG) identified greater connectivity with left middle frontal gyrus for Americans than East Asians (top left panel). Cross-cultural comparisons of connectivity strengths (top right panel) indicated that the connectivity between the left anterior PHG and this cluster was above zero for Americans and below zero for East Asians. This FC showed a negative correlation with Schwartz Value Scale scores for tradition (SVS-tradition; bottom left panel) and a positive correlation with the self-construal scores (SCS-adjusted independence score; bottom right panel)

No regions emerged as being more strongly connected to left hippocampus in one culture than the other. To aid interpretation of the cultural differences reported here, functional connectivity maps are presented separately for each culture for each of the three regions of interest in the Supplemental Materials.

Correlation between FC and memory performance

We conducted exploratory tests of correlations between memory performance (measured by d’ of Target and Lure, namely the discriminability between old items and similar lures) and the FC of the seven regions showing cultural differences (i.e., with each of the clusters listed in Table 1), setting a p value threshold at 0.007 after Bonferroni correction. No correlations emerged as significant, ps > 0.012.

Correlation between FC and cultural values and self-construal

We further examined the influence of culture on patterns of connectivity using participants’ scores on the Singelis Self-Construal Scale (SCS; Singelis, 1994) and the Schwartz Value Survey (SVS; Schwartz, 1992). Although some research conceptualize culture as encompassed by measures of these types of values and orientations, other work demonstrates that results based on individual difference measures may not converge with comparisons based on group membership (Chua et al., 2022; Na et al., 2010). We first examined the correlations across the entire sample of Americans and East Asians, in keeping with prior research (Yu et al., 2019; Yu et al., 2021 Exp 2), and then tested each cultural group separately (see Chuang et al., 2020 for within-group use of values scales). For these analyses, two American participants and one East Asian participant were excluded due to missing SVS and SCS data.

Singelis SCS:

SCS measures the degree of independence and interdependence in an individual. In order to cancel out the within-individual noise and reduce the number of variables (hence, the likelihood of Type I error) in subsequent correlation analyses, we subtracted the interdependence score from the independence score to generate an adjusted independence score. This approach has been used in prior work (Yu et al., 2021) and also helps to adjust for potential cultural or individual differences in response bias (e.g., should one individual tend endorse items as “strongly agree” whereas another individual endorses items more moderately, subtracting the scores provides us with a measure of each subject’s relative independence or interdependence, adjusting for response bias). American and East Asian participants differed significantly in the adjusted independence measure, with Americans (M = 0.28) demonstrating higher independence than East Asians, whose scores on average tended to be more interdependent (M = −0.41), t (100) = 3.74, p < 0.001.

To further explore how self-construal relates to cultural differences, we tested the relationship between self-construal and FC strengths. This would allow us to examine whether the strength of FC in an individual is affected by how strongly one endorses the view of the self that is associated with a given culture (in this case, independence vs. interdependence). To do this, analyses were conducted correlating the adjusted independence score from Singelis SCS with the FC values for each of the seven clusters in which we identified cultural differences (Table 1). The p value threshold was set at 0.007 after Bonferroni correction. Degree of independent self-construal was significantly associated with functional connectivity for left anterior PHG and all three clusters. That is, connectivity between left anterior PHG and left middle frontal gyrus (peak [−46;38;20]), which was seen to be stronger in Americans in the comparisons across cultures, was positively associated with independence, r (100) = 0.26, p = 0.007 (Fig. 5). Connectivity between left anterior PHG and the bilateral temporal pole (peak [36;6;−38] and [−50;2;−32]), which was seen to be stronger in East Asians in the culture comparisons, was negatively associated with independence, r (100) = −0.38, p < 0.001 for right, and r (100) = −0.363, p < 0.001 for left. These regions correspond to clusters A and B shown in Fig. 4.

Because groups differed in overall SCS scores and this could have contributed to our findings, we also examined the relationship between self-construal and FC separately for the two cultural groups. The patterns generally tended to persist and correlations went in the same direction for both cultural groups, although only a subset of the correlations reached significance. See Table 3 for a complete breakdown of the correlations in each culture between the FC values and the SCS scores.

Table 3.

Breakdown of the correlations between FCs and SCS or SVS-Tradition separately in Americans and East Asians

	SCS East Asians (df = 56)	Americans (df = 42)
[−36,6,−38]	r = −0.14, p = 0.29	r = −0.35, p = 0.02
[−50,2,−32]	r = −0.29, p = 0.03	r = −0.16, p = 0.30
[−46,38,20]	r = 0.14, p = 0.14	r = 0.004, p = 0.98
	SVS-Tradition East Asians (df = 56)	Americans (df = 42)
[−46,38,20]	r = −0.23, p = 0.09	r = −0.18, p = 0.24

SVS:

East Asians and Americans differ in two out of the ten values in SVS (all mean values and statistics listed in Table 2), which provided further evidence for the existence of systematic discrepancies between cultures on personal values. East Asians are higher than Americans in their value rating for Tradition (M_{East Asian} = 3.63 vs. M_American = 2..91) t (100) = 2.98, p = 0.004, and Power (M_{East Asian} = 3.63 vs. M_American = 2.69), t (100) = 3.64, p < 0.001. That is, compared with Americans, East Asian participants place a higher value on subordination to socially imposed expectations, social power, authority, and wealth. We conducted correlation analyses involving SVS-Tradition, SVS-Power, and the FC values for each of the seven clusters in which we identified cultural differences (Table 1). The p value threshold was set at 0.0036 after Bonferroni correction. Connectivity between left anterior PHG and left middle frontal gyrus (peak [−46;38;20]), which was seen to be stronger in Americans, was negatively associated with individual’s value of tradition, r (100) = −0.31, p = 0.001 (Fig. 5).

Table 2.

Cross-cultural comparison of scores for each culture on the Schwartz Value Scale

	MAmerican (SD)	MEast Asian (SD)	t (100)	p
Conformity	4.36 (1.27)	4.83 (1.29)	1.85	0.067
Tradition*	2.91 (1.27)	3.63 (1.17)	2.98	0.004
Benevolence	5.27 (.78)	5.24 (1.18)	0.14	0.892
Universalism	4.92 (.73)	5.09 (.97)	0.93	0.353
Self-direction	5.20 (.93)	5.71 (.99)	2.66	0.009
Stimulation	4.21 (1.44)	4.92 (1.30)	2.46	0.016
Hedonism	4.75 (1.30)	4.70 (1.46)	0.17	0.863
Achievement	5.03 (1.04)	5.28 (1.14)	1.12	0.266
Power*	2.69 (1.28)	3.63 (1.32)	3.64	<0.001
Security	4.52 (1.07)	5.00 (1.03)	2.32	0.023

^*Values that showed significant cross-cultural difference on a threshold of p = 0.005 (Bonferroni correction for multiple comparisons).

Similar to the SCS findings, when examining these FC and SVS relationships in each cultural separately, the results trended in the same direction although the correlations did not reach significance in either group. See Table 3 for a breakdown of the correlations in each culture between the FC values and the SVS-Tradition scores.

Discussion

Previous work has illustrated the ways in which differences in goals and tasks across individuals or motivational states can influence the recruitment of networks during rest, with some patterns of functional connectivity predicting higher levels of memory performance. In the present study, we extend those lines of research to investigate to the influence of culture on resting state functional connectivity, with a focus on memory regions. These memory regions were selected from a previous study that identified regions more engaged by East Asians than Americans during the successful encoding of detailed memories, compared to more general memories (e.g., supporting successful recognition of the identical exemplar, compared with a different exemplar of the same item type, such as distinguishing between two different exemplars of cats). Results indicated that cultural background shapes functional connectivity even during non-task-related periods.

Our results indicated that in East Asian participants, targeted medial temporal regions previously associated with cultural differences showed stronger FC to a variety of areas, primarily in the occipital regions (including the cuneus and lingual gyrus on the medial side, lateral occipital cortex, and temporal occipital fusiform cortex) and the temporal pole. These regions span the primary visual cortex (e.g., calcarine cortex) and areas associated with higher-order visual processing such as object recognition (lateral occipital cortex; Malach et al., 1995; Kanwisher et al., 1996; Grill-Spector et al., 2001), extending into regions with higher-order functions, such as encoding of visual memory (temporal occipital fusiform and temporal pole; Weiner & Zilles, 2016; Henson et al., 2000; Wagner et al., 1999; Herlin et al., 2021) and roughly tracing the “what” pathway (two-streams hypothesis; Goodale & Milner, 1992). Interestingly, American participants demonstrated stronger connectivity between the left posterior PHG and a few regions in the dorsal pathway such as the superior lateral occipital cortex and the superior parietal lobule, potentially implying processing of information about spatial orientation (the “where” pathway; Goodale & Milner, 1992).

We speculate that the pattern of results could reflect cultural differences in the prioritization of different types of information and its effects on memory for that information. Because the participants had just experienced an object encoding task, albeit an incidental one, communication between medial temporal and ventral stream regions could suggest that after the task, East Asian participants tended to continue to mobilize object-processing regions. Likewise, Americans could have continued to mobilize spatial-processing regions. Such an interpretation could be consistent with past work illustrating that weaker memories are prioritized in replay during rest periods (Schapiro et al., 2018). That is, based on previous cross-cultural work arguing that Americans focus more on the properties of objects (Leger & Gutchess, 2021; Millar et al., 2013), in line with analytic processing, and that East Asians focused more on relationships and context (Masuda & Nisbett, 2001; Miyamoto et al., 2011; Nisbett et al., 2001; Wang et al., 2012), in line with holistic processing, it could be the case that increased activity during a rest period reflects spontaneous ongoing processing and consolidation of information that is less emphasized in each culture (i.e., ventral stream for object processing for East Asians; dorsal stream for spatial processing for Americans). Future studies manipulating the object versus spatial demands of the task would be necessary to test this interpretation, including explicitly manipulating the spatial elements of the task beyond the randomly occurring differences in the orientation of objects (which has implications for the grasping of objects; Verhagen et al., 2012) that occurred for these stimuli. Relatedly, frameworks distinguishing between an anterior memory system invoked in retrieval of conceptual knowledge (Ranganath & Ritchey, 2012) or abstract, gist-like processing (Burke et al., 2018) as opposed to a posterior system emphasizing contextual remembering (Ranganath & Ritchey, 2012) or detail (Burke et al., 2018) could be relevant to interpreting patterns of activity during post-encoding rest. Evaluating this interpretation also would require additional research explicitly manipulating memory and task demands, with the potential to support cultural differences in analytic versu holistic processing.

Moreover, relationships between functional connectivity and measures of cultural values converge with the patterns of group differences across cultures and the candidate explanations for these patterns. Namely, higher endorsement of independent self-construal is associated with lower levels of functional connectivity between left anterior parahippocampal gyrus and the temporal poles, bilaterally. This is consistent with Americans, a more independent culture, tending to have lower levels of connectivity between these regions than East Asians, a more interdependent culture. Additionally, both independent self-construal and rejection (i.e., lower levels of endorsement) of the value of tradition, both of which would be considered more American patterns of responding, were association with functional connectivity of left anterior parahippocampal gyrus and left middle frontal gyrus; this network had stronger connectivity in Americans than East Asians in the analysis of group differences. We speculate that this pattern could suggest that Americans more readily engage more top-down processing than East Asians, but further tests of the connectivity of this pathway under different conditions, including more cognitively demanding ones, are necessary to understand this pattern. There also is the potential for group differences to contribute to these effects. Given that in this study, some of the correlations between FCs and cultural values were not significant when analyzed separately for each cultural group, further research is needed. By testing relationships within each culture group with large samples, we can better understand the extent to which these associations exist across individuals within one culture or if they are more useful for characterizing differences across multiple cultural groups.

Although we suggest that the cultural differences in connectivity during rest reflect memory-related processes, such as spontaneous reactivation of information to support memory formation, it is possible that the patterns we identify reflect more stable trait-based differences across cultures. Due to the current design with only one resting state scan, it is impossible to discern whether these findings reflect post-encoding processes or reflect more general cultural differences in the operation of networks. That is, the patterns of co-activation reported here may occur spontaneously during any resting period or other aspects of the task (e.g., visual information processing). The relationships between functional connectivity and scales of cultural values and self-construals, but not memory performance, could be seen as support for the idea that cultural differences reflect traits applying across tasks. Such an interpretation would be in line with previous culturally-relevant work reporting differences in functional connectivity at rest as indicating chronic differences in independent or interdependent self-construals (Li et al., 2018; Wang et al., 2013; Zhu et al., 2021). However, it is important to acknowledge that neural differences can occur in the absence of behavioral differences (Kitayama & Uskul, 2011), perhaps reflecting different strategies or sensitivity of the measures, and that meaningful cultural differences can emerge even in the absence of a task (Goh et al., 2007). One might argue from a memory perspective that traits are closely associated with styles of memory performance, including spontaneous differences in the ways in which information in the world is attended to and selected for encoding into memory (Gutchess & Indeck, 2009). Different styles of remembering—more episodic or more semantic—are associated with patterns of connectivity that are, respectively, more posterior or anterior (Sheldon et al., 2016). Our use of medial temporal lobe regions implicated in successfully encoding detailed memories in prior cross-cultural research (Paige et al., 2017b) provides some support that the present findings have implications for memory, although further research is needed.

Future studies should test for cultural differences across multiple periods of rest, following different tasks. In addition, cultural differences in functional connectivity should be assessed during tasks as well as at rest. Finn (2021) highlights shortcomings of RSFC research, including the interpretability of differences. She argues that testing functional connectivity during tasks (e.g., during encoding or retrieval) is a superior approach, because it may be more sensitive to revealing relationships with behavior, such as memory performance. We agree that directly manipulating cognitive processes or testing whether the patterns revealed in this study are associated with particular types of perceptual and mnemonic experiences (e.g., attending holistically vs. analytically; processing object or location information) is necessary for interpreting effects; this can be achieved during rest, by manipulating preceding events, as well as by comparing task-based fMRI data. Ultimately, data from multiple different tasks is needed to fully characterize the nature of cross-cultural differences. Longitudinal research also can contribute by characterizing the stability of cultural-specific patterns over time, including comparisons within an individual.

Overall, the current study illustrates cultural differences during post-encoding rest. The results demonstrated the importance of FC with medial temporal regions, particularly anterior and posterior PHG, which, respectively, showed stronger connectivity with dorsal brain networks in Americans and ventral brain networks in East Asians. These networks may have implications for post-encoding consolidation of information in memory or reflect more general cultural differences in cognitive processes (e.g., holistic vs. analytic processing) or network connectivity. Future research would benefit from systematically examining how cultural differences emerge across different stages of memory (encoding, consolidation, and retrieval), as well as under different task demands (e.g., pre- and post-encoding; tasks emphasizing relatively more analytic vs. holistic processing).

Open Practices Statement.

None of the data or materials for the experiment reported here are available to be shared, and the experiment was not preregistered.