Culture and the brain: Opportunities and obstacles

Authors


Haotian Zhou or John T. Cacioppo, Department of Psychology or Center for Cognitive and Social Neuroscience, University of Chicago, 5848 S. University Avenue, Chicago, IL 60637, USA. Email: zhouhaotian@uchicago.edu or jcacioppo@uchicago.edu

Abstract

A major evolutionary advance of humans is a mind that is capable of constructing, perpetuating, adapting to, and exploiting culture. The birth of cultural neuroscience reflects the growing realization that a full account of the human mind requires understanding of the multiple and reciprocal influences between the biological and the sociocultural. In the present paper, we illustrate how attention to the brain, as exemplified in functional magnetic resonance neuroimaging (fMRI) studies of sociocultural processes, contributes to a more comprehensive understanding of the human mind. We end by discussing a set of challenges facing researchers using fMRI and the possible means for dealing with these challenges.

Introduction

Social and cultural influences are essential to understanding what is fundamental about human nature. However, this was not always thought to be the case. From the perspective of many scientists during the 20th century, the contributions of the social world to behaviour were thought best to be considered later, if at all. This was because social factors were thought to be of minimal interest with respect to the basic development, structure, or processes of the brain and behaviour. To the extent that social factors were suspected of being relevant, their consideration was thought to be so complicated that they should be considered at some later date (e.g. Cacioppo, 2002). Consequently, single organisms have been the sole focus of many biological, behavioural, and cognitive scientists in the 20th century.

Further fuelling this focus on the solitary individual in scientific analyses was the dominant metaphor of the mind—the stand-alone desktop computer. Just like the computer whose functioning relies on the set of information-processing operations implemented exclusively inside the machine, the sources of mind and behaviour were once assumed to be located in the recesses of the individual brain. Sociocultural factors were thought to be not unlike the electrical systems of such an isolated computer (i.e. 220 V in China or 110 V in the USA), which had little effect on the computations inside the machine provided that a suitable adaptor was connected and, as a result were, at most, tangential to the understanding of the mind.

The understanding of computers as stand-alone devices is no longer appropriate as a result of the development of the computer network. Nowadays, computers operate as a connected collective resulting in unforeseen power, capacities and representations. If decades of connection were enough to render an asocial model of electronic computers essentially obsolete, models of the hominid brain and mind that ignore sociocultural levels of organization that have existed for millennia are likely to be, at least, equally inaccurate.

Humans are a quintessential social organism (Baumeister & Leary, 1995; Cacioppo et al., 2006). We are born to the most prolonged period of abject dependency of any mammal and dependent on conspecifics across the life span to survive and prosper. The result is selection pressure for construction and maintenance of superorganismal structures (i.e. emergent organizations beyond the individual) ranging from dyads, families and groups, to cities, civilizations, and cultures (Cacioppo et al., 2006; Dunbar & Shultz, 2007). Humans are most certainly not the only species characterized by superorganismal structures, but human superorganismal structures tend to be flexible rather than rigid, extend beyond the current moment and context, and organize around ideological as well as genetic similarity (Cacioppo & Patrick, 2008). One of the broadest and most surreptitiously influential of these structures is culture (Berry, 2003).

Culture is defined as patterns of representations (e.g. lay theories, scripts and worldviews) embodied in institutions, practices, artifacts and public narratives that exist simultaneously in people and the context in which people exist (Adams & Markus, 2004; Kitayama, 2002). These often-implicit patterns of ideas and practices have long been noted by cultural psychologists for their potential to regulate, express and transform the human mind. For instance, the Whorfian hypothesis postulates a systematic relationship between one's linguistic background and one's experience of and response to the environment. Consistent with this hypothesis, Kay and Kempton (1984) found that English speakers exhibited more differentiation among colours that straddle the English category boundary between ‘green’ and ‘blue’ than speakers of a Uto-Aztecan language in Mexico who uses a single word for these colours (see also Ames & Fiske, 2010).

Aside from language, other aspects of culture can also have profound effects on our thoughts, feelings and practices towards others. Scheper-Hughes (1992) conducted an ethnographic study in a remote Brazilian community to investigate the way maternal love was expressed in an area where infant mortality is high. Scheper-Hughes noticed that women in that community were emotionally detached from their babies and they rarely displayed signs of mourning when infants died. Scheper-Hughes regarded the maternal apathy in this culture as an adaptive response to its high infant mortality rate (about 50%). If nearly one out of two newborns is going to die before 1 year of age, it is adaptive for mothers to avoid forming strong emotional bonds with their newborn only to suffer the tormenting pain following an infant's death. This finding demonstrates clearly that mother–infant attachment is a biologically enabled rather than a biologically determined capacity. (For additional evidence of how culture can affect the activation of different neural systems see Ng, Han, Mao, & Lai, 2010.)

It should be noted that a strict sociocultural level analysis is not sufficient, either, if we seek comprehensive theories of human behaviour. For instance, the ‘missing women’ problem which refers to the terrible deficit of women in China has been attributed to a cultural preference for sons, resulting in both parental intervention to manipulate the gender composition of their family and a gender bias in the care of newborns (Sen, 2003). Many of the policies designed to reduce peoples' preference for sons have proven ineffectual in addressing the issue, which could imply that the policies were poorly designed—or that the problem the policies were designed to address had been specified incorrectly. A recent study (Oster, 2005) has identified a biological cause of this male-biased sex ratio. Hepatitis B is an endemic disease in many parts of China, and female carriers of hepatitis B have an elevated probability of having a son. Oster's investigations showed that nearly 75% of the gender imbalance observed in Chine in recent decades could be explained by hepatitis B. When cultural education and vaccination campaigns are used together, the ‘missing women’ problem may be eliminated. Thus, the ‘missing women’ problem in China illustrates the principle of multiple-determinism, which specifies that human behaviour can have multiple antecedents within or across levels of organization (Cacioppo & Berntson, 1992).

It is no longer a novel insight that all human behaviours at some level are biologically caused. However, the simple biological substitutionism (i.e. the belief that molar level analysis can be replaced by microscopic level analysis) is unlikely to yield a satisfactory explanation for complex behaviours (Berntson & Cacioppo, 2004). The biological substrates (i.e. neural, hormonal or genetic mechanism) of human behaviours have evolved to support various superorganismal structures that emerge to deal with physical and social environmental challenges and, consequently, helped humans to survive, reproduce, and care for offspring sufficiently long that they too survived to reproduce (Cacioppo & Patrick, 2008). Accordingly, cultural processes can both cause and be caused by the biological mechanisms underlying human behaviour, reflecting the principle of reciprocal determinism, which specifies that there can be mutual influences between microscopic (e.g. biological) and macroscopic (e.g. social or cultural) processes (Cacioppo & Berntson, 1992). For example, the handling of rat pups alters maternal behaviour towards the pups and affects the structure and reactivity of the stress-hormone system (Meaney, 2004). These early influences on this system, in turn, affect the pups' reactions to stressors and, in later life, promote similar maternal behaviour towards their pups (Meaney, 2004).

The inextricable links between sociocultural and biological levels of organization argue for the incorporation of the neurosciences into the ongoing integration of culture and psychology if we are to develop comprehensive theories. This view is echoed by the new paradigm, biocultural co-constructivism, introduced by Paul Baltes and colleagues, which argues that ‘brain and culture are in a continuous, inter-dependent, co-productive transaction and reciprocal determination’ (Baltes, Rösler & Reuter-Lorenz, 2006, p. 1). Social neuroscience is particularly well positioned to facilitate the development of contemporary theories of sociocultural processes and behaviours. As an interdisciplinary approach drawing on biological concepts and methods, social neuroscience is concerned with investigating various superorganismal structures and the underlying neural, hormonal, and genetic mechanisms that make them possible (Cacioppo & Berntson, 1992).

Indeed, a growing literature in social neuroscience has demonstrated the benefits of incorporating a biological level analysis in the study of social behaviour. The consideration of the principles and methods of the neurosciences can constrain and inspire new psychological hypotheses, foster empirical tests of otherwise indistinguishable theoretical accounts of behaviours, and increase the comprehensiveness and relevance of psychological theories (Cacioppo, 2002). Similar benefits are also emerging in the new area of cultural neuroscience (e.g. Han & Northoff, 2008). By investigating cultural variations in psychological, neural and genomic processes, cultural neuroscience seeks to uncover the biological mechanisms underlying culture and its instantiations in and influences on the brain and behaviours (Chiao & Ambady, 2007). A growing body of exciting findings in cultural neuroscience testifies to the fruitfulness of the marriage between neuroscience and sociocultural psychology (Ames & Fiske, 2010).

Instead of presenting a comprehensive review of the burgeoning field of cultural neuroscience, we focus here primarily on the exciting opportunity made possible by the rapid development of functional magnetic resonance imaging (fMRI). In particular, through a survey of a selection of representative cross-cultural neuroimaging studies, we illustrate how practitioners in cultural neuroscience can further our understanding of the mutual constitution of culture and mind by combining fMRI with well-established behavioural paradigms. Markus and Hamedani (2007) provided a succinct explanation of the meaning of mutual constitution: ‘as people actively construct their worlds, they are made up of or ‘constituted by’, relations with other people and by the ideas, practices, products and institutions that are prevalent in their social context'. We conclude by discussing the limitations of fMRI that warrant consideration. By spelling out the limitations as well as the contributions of fMRI, we hope to underscore the importance of converging evidence.

Assessing cultural– psychological interactions

After surveying the literature in the field of cultural psychology, Markus and Hamedani (2007) identified five major approaches adopted by researchers to investigate the dynamic interdependence between sociocultural context and mind: (i) the dimensional approach; (ii) the cognitive toolkit approach; (iii) the dynamic constructivist approach; (iv) the sociocultural model approach; and (v) the ecocultural approach. These five approaches are not mutually exclusive so that sometimes a study could easily be classified into more than one category (e.g. Kitayama, Ishii, Imada, Takemura, & Ramaswamy, 2006). Moreover, no single approach is necessarily superior over another. We demonstrate how fMRI can be integrated with each of the five major approaches to advance our understanding of the effects of the sociocultural factors on psychological processes.

Integrating fMRI with the dimensional approach

Researchers in the dimensional approach conceptualize culture as a multidimensional structure that can be evaluated along each of the constituting dimensions (e.g. Schwartz, 1990; Triandis, 1995). Their goal is to specify the quantifiable dimensions of culture that can account for psychological differences between people of different cultures. Such dimensions are called ‘cultural syndromes’ (Triandis, 1996). Some of the familiar dimensions identified by practitioners in this tradition include masculinity-femininity and collectivism-individualism. To dimensionalists, an individual's cultural affiliation is less important than an individual's ‘score’ on the set of dimensions underlying cultures, because the effects of sociocultural elements on psychology are mediated only through these dimensions. Therefore, it is of utmost importance to ensure that these dimensions are measured with a high degree of reliability and validity (Triandis, 1996). The dominant methodology for measuring dimensions is the questionnaire, which is susceptible to a number of artifacts (e.g. response set biases, social desirability biases) and is inherently limited in what it can reveal about processes about which people are unable to report. As LeDoux (2000) pointed out, most cognitive processes are taking place beneath the conscious awareness and are, thus, rarely accessible to self-reflection. Evidence of individuals having difficulty identifying reasons for their actions, reporting their preferences and predicting future behaviours abounds (e.g. Cacioppo & Sandman, 1981; Nisbett & Wilson, 1977).

One way to address this measurement conundrum is to develop instruments for assessing implicit processes. For instance, Kitayama, Duffy, and Uchida (2007) have advocated for the development of an implicit measure of the widely studied self-construal style (i.e. individualistic or collectivistic). Cacioppo, Lorig, Nusbaum, and Berntson (2004) noted that social neuroscience has the potential to extend investigations of underlying biological and psychological processes beyond the vistas accessible by self-report or chronometric measures. In particular, if a biological event (e.g. activation in certain brain areas) can be found to be reliably and exclusively related to a psychological construct (e.g. self-construal style), then we can draw inference about the nature of the latter from the measurement of the former (Cacioppo & Berntson, 1992). A recent study by Chiao et al. (2009) represents a promising first step towards this direction.

The focal dimension in the study by Chiao et al. (2009) is the collectivism-individualism dimension as measured by the Singelis Self-Construal Scale (Singelis, 1994). Past studies (e.g. Markus & Kitayama, 1991) have shown that individualists tend to view personal traits as stable across situations (e.g. I am an extrovert) but collectivists tend to view personal traits as contingent upon situations (e.g. I am extrovert with friends). Accordingly, Chiao and colleagues reasoned that individual differences on the individualism-collectivism dimension could be captured by the manner in which self-representation is constructed. In particular, for people with an individualistic propensity, their self-representations are more likely to be constructed in a general manner whereas for people with a collectivistic propensity, their self-representations are more likely to be constructed in a contextual manner. Therefore, when judging the degree of self-descriptiveness of a general description (e.g. In general, I am assertive), individualists are likely to search through their self-representations for answers, but collectivists may need to look for answers somewhere else. The situation reverses when judging the self-descriptiveness of a contextual description (e.g. When talking to my mother, I am casual). Drawing on prior studies which identified the medial prefrontal cortex (mPFC) as a neural region involved in tasks that activate self-knowledge (e.g. Heatherton et al., 2006; Macrae, Moran, Heatherton, Banfield, & Kelley, 2004), they reasoned that the process of browsing through self-representations would produce differential activation of the mPFC to general description versus contextual description, and the pattern of mPFC activation across these descriptions would indicate an individuals' position on the individualism-collectivism dimension.

Chiao et al. (2009) had participants undergo an fMRI scan while they read a number of short descriptive sentences and indicated whether or not the descriptions described them. The short descriptions fell into two major categories: general and contextual. Before the scanning session, researchers registered the apparent cultural affiliation of participants (i.e. American or Japanese) and assessed individual participants' individualistic-collectivistic propensity using the Singelis Self-Construal Scale. On the basis of participants' responses, they were split into an individualist group or a collectivist group.

The neuroimaging data revealed that, among individualists, processing general descriptions elicited greater activation in the mPFC than processing contextual descriptions. On the contrary, among collectivists, processing contextual descriptions led to stronger signals in the mPFC than processing general descriptions. However, there was no interaction between the types of descriptions and apparent cultural affiliations. This null result was interpreted as consistent with the dimensionalistic notion that sociocultural influence on psychological processes is mediated through the internalized cultural syndrome.

Specifically, Chiao et al. (2009) found that an individual's score on the individualism-collectivism dimension mapped nicely onto the differential activation of the mPFC in response to contextual versus general descriptions with a correlation coefficient of 0.7. The fMRI results in this study did not add anything to the survey responses they collected, but both proved more informative than chronometric measures, which did not discriminate either cultural affiliation or cultural syndrome. Additional research is needed to develop paradigms in which fMRI is shown to overcome limitations associated with the measures (e.g. verbal reports and questionnaires) typically found in cultural-psychological research.

Integrating fMRI with the cognitive toolkit approach

Researchers who adopt the toolkit approach focus primarily on how culture shapes the cognitive and perceptual systems (Markus & Hamedani, 2007). Nisbett, Peng, Choi, and Norenzayan (2001), for instance, have proposed that the influences of culture on psychological processes are mediated through its effects on cognitive systems. That is, culture can be conceptualized as a toolkit comprising specific cognitive tools acquired from the culture. According to Nisbett et al. (2001), the toolkits can vary in several respects as a function of cultural context: (i) the tool or the set of tools habitually used for a given problem; (ii) the proficiency with a given tool or a set of tools; and (iii) the accessibility of a given tool or a set of tools.

Evidence from cross-cultural studies demonstrated the cultural specificity of the cognitive toolkit. For instance, Kitayama, Duffy, Kawamura, and Larsen (2003) created a simple paradigm called the frame-line task (FLT) to examine cultural variation in performance on two types of cognitive problem. In the FLT, participants are presented with a square that encloses a vertical line and are then presented with a second square which could be the same or a different size but devoid of the embedded vertical line. The task of participants is to draw a line in the second square that is of either the same length as the line in the first square (i.e. absolute problem) or the same length-to-height of the surrounding square ratio as the first line-square combination (i.e. relative problem).

When Kitayama et al. (2003) presented both versions of the FLT to a sample of North Americans and a sample of East Asians, they found that the North Americans were more accurate on the absolute problem but less accurate on the relative problem than the East Asians, consistent with the notion that the cognitive toolkits of Easterners suit problems that require the incorporation of contextual information (i.e. the surrounding square) better, whereas the cognitive toolkits of Westerners suit problems that require the insulation of contextual information better.

This study is silent with regard to the deeper question about the exact nature of the inferred differences in toolkits. For instance, the behavioural study did not distinguish between two competing hypothesis: (i) participants of different cultures used different tools to address a given problem; or (ii) participants of different cultures deployed the same tool in response to a given problem but differed in their expertise in its use. Both hypotheses are capable of accounting for the observed cultural difference, but each implies distinctive mechanisms.

With its ability to identify brain regions engaged by a task in real time, fMRI may provide a means of testing these competing hypotheses. For instance, if people from different cultures use different tools (i.e. cognitive processes) to approach a given problem, they might exhibit activation in different sets of brain regions. However, if people from different cultures use the same tool or combination of tools to address a given problem and the performance discrepancy is driven primarily by differential proficiency with the tools, they should exhibit activation in similar networks, but the activation of regions within this network should be stronger in the group with inferior performance (Milham, Banich, Claus, & Cohen, 2003).

Hedden, Ketay, Aron, Rose Markus, and Gabrieli (2008) adopted this approach in an fMRI study of Asian and American participants on the absolute and relative versions of the FLT. They found that a set of regions (e.g. Brodmann's areas 6 and 40) in the association cortices was activated more strongly in Asians than in Americans during the absolute version of the FLT, and the same set of brain regions was activated more strongly in Americans than in Asians during the relative version of the task, but that, otherwise, the brain regions involved in performing each task were similar for Asians and Americans. Prior research has shown that the set of regions in the association cortices in which differential activation was found is involved in sustained attentional control (Wager & Smith, 2003), and activity in this network is greater when task difficulty increases, presumably due to demands for greater attentional control (Braver et al., 1997). These results, then, suggest that Asians and Americans performed similar cognitive operations (i.e. utilized the same tools) when carrying out the tasks, but the group with less experience with the task showed evidence of the greater attentional demands required to perform the task using tools with which they were less practiced. Such a conclusion suggests that in the context of FLT culture may not promote the use of different tools for given problems, but, rather, different cultural environments expose their respective inhabitants to different habitual tasks. As a result, people of varied cultural backgrounds become differentially practiced with the cognitive tools universally used for the same purpose.

When cultural idiosyncrasies fail to generate observable behavioural variations, the potential value of adding biological measures, in general, and fMRI, in particular, is increased (Cacioppo, Petty, & Tassinary, 1989; Chiao & Ambady, 2007). According to the principle of multiple determinism (Cacioppo et al., 2004), a target event at one level of organization (e.g. behavioural level) may have several very distinct antecedents at another level of organization (e.g. psychological level). As behavioural paradigms rely on the observable end product of mental processes to make inferences about the mind, a purely behaviouristic approach to cultural psychology may preclude researchers from uncovering those cultural-specific mental processes which do not yield a detectable difference in responses. However, neuroscientific methods may be well positioned to address this deficiency of behavioural paradigms, as evidenced by a recent fMRI investigation of cultural variation in numerical processing.

Tang et al. (2006) recorded participants' brain activation when they were performing two types of mathematical task (i.e. the addition task and the comparison task). In each trial, participants saw three numbers presented simultaneously. In the addition task, participants judged whether the third number was equal to the sum of the first two numbers, whereas in the comparison task, participants judged whether the third number was larger than both of the first two numbers (Tang et al., 2006).

Results indicated that both Chinese and Westerners performed equally well in terms of response time and accuracy. Significant group differences emerged in regional brain activation, as measured by fMRI. Specifically, Western participants showed greater activation in the perisylvian regions, including Broca and Wernicke areas, in both tasks, whereas Chinese participants showed greater activation in the visuo-premotor association areas, especially during the comparison task. Additionally, a connectivity analysis was carried out to investigate possible cultural differences in the functional neural networks involved in the comparison task. Results suggested that different neural networks were activated in Western and Chinese participants. On the basis of these results, Tang and colleagues suggested that the culturally specific brain activation during numerical processing reflected the deployment of different cognitive tools that had been acquired from the culture.

Integrating fMRI with the dynamic constructivist approach

The dynamic constructivist tradition regards culture as a network of distributed knowledge shared by a collection of interconnected individuals (Hong, Morris, Chiu, & Benet-Martinez, 2000). Therefore, similar to other types of knowledge (e.g. social stereotypes), culture exerts its influence on psychological processes only when these knowledge structures are activated; and the activation of cultural knowledge and the subsequent application of this knowledge in a given situation follow the principles of availability, accessibility and applicability (Hong, Wan, No, & Ciu, 2007). According to Hong et al. (2007), the essence of the dynamic constructivist approach is the notion that ‘all individuals are capable of representing multiple cultures in their minds and switching between representations of cultures’ (Hong et al., 2007, p. 340). For instance, Briley, Morris, and Simonson (2000) reported that Easterners were no more likely to compromise than their Western counterparts in a decision-making task, but cultural differences emerged when participants were required to justify their decision, a process presumed to activate cultural norms.

A unique aspect of the dynamic constructivist approach is its use of bicultural individuals to model the dynamic application of cultural knowledge. The priming procedure is used to experimentally manipulate the temporary accessibility of certain cultural knowledge items and to observe the behavioural consequences of such manipulations. For instance, in a highly cited study (Hong, Chiu, & Kung, 1997), bicultural participants (i.e. Chinese Americans) were randomly assigned to two conditions where they were primed with either Chinese cultural icons (e.g. Tiananmen, Confucius etc.) or American cultural icons (e.g. the White House, Abraham Lincoln etc.). The researchers reasoned that exposure to American icons should activate the internal attribution schema in these bicultural participants' American cultural knowledge network, whereas exposure to Chinese icons, instead, should activate the external attribution schema in their Chinese cultural knowledge network. Consistent with their hypothesis, participants primed with American icons were more likely to view a figure's action (i.e. a fish swimming in front of a group of fishes) as self-determined, whereas participants primed with Chinese icons tended to interpret the same figure's action as caused by others.

By re-conceptualizing culture as knowledge structure that can be experimentally primed, research in the dynamic constructivist tradition makes it possible to test causal inferences about cultural influences on psychological processes (Cohen, 2007). Nonetheless, the mechanism proposed to account for the psychological consequences of cultural priming remains speculative. In particular, dynamic constructivists conjecture that incoming social information is channelled through the situationally activated cultural knowledge which biases the way such information is construed. Fortunately, with the advent of fMRI, one can investigate the effects of culture on differences in which the incoming information is construed.

For instance, Chiao et al. (2010) primed a group of bicultural participants (i.e. Asian Americans) with either the independent self schema of their American cultural heritage or the interdependent self schema of their Asian cultural heritage with two commonly used priming procedures, the Sumerian Warrior Story task and the Similarities and Difference with Family and Friends task (Trafimow, Triandis, & Goto, 1991). After the cultural prime, while in the fMRI scanner, participants were asked to judge whether or not descriptive sentences described themselves. The sentences were of two types: contextual description (e.g. When talking to my mother, I am modest) and general description (e.g. In general, I am humble).

The results revealed that the activation of a neural circuit for processing self-relevant verbal information, which includes the mPFC and the posterior cingulated cortex (PCC) (Amodio & Frith, 2006; Northoff & Bermpohl, 2004), varied as a function of cultural primes. In particular, bicultural participants primed with an independent self schema showed greater activation in the mPFC and PCC to the general description, whereas participants primed with an interdependent self schema showed greater activation in the same regions to the contextual description. Moreover, the efficacy of the priming manipulation was correlated with the differential activation of the mPFC and PCC to the general description versus the contextual description. Such findings suggest that depending on which part of the cultural knowledge network was made salient, the degree to which the same description is construed as self-relevant differed. Specifically, people with activated independent self concepts viewed general descriptions as more self-relevant, whereas those with activated interdependent self concepts viewed contextual descriptions as more self-relevant, a pattern consistent with the claims of dynamic constructivists that when particular pieces of cultural knowledge are activated they will guide individuals' construction of the meanings (Hong et al., 2000).

Integrating fMRI with the sociocultural model approach

A sociocultural model is a cognitive schema that includes ideas and practices about how the material and social world operates and that is widely shared by members of a society or cultural context (Markus & Hamedani, 2007). Markus and Hamedani (2007) noted that a ‘sociocultural model gives form and direction to individual experience, for example, perception, cognition, emotion, motivation, action’ (p. 15). The sociocultural modelling approach has been used to study a wide array of phenomena, ranging from agency, motivation to morality and well-being. For instance, Kitayama et al. (2007) proposed two major sociocultural models of action regulation: (i) independent, which refers to actions that are directed towards achieving one or more task goals; and (ii) interdependent, which refers to actions that are responsive to the social context. They further suggested that independent sociocultural models are more prevalent in Western cultures, whereas interdependent sociocultural models are more prevalent in Eastern cultures.

Kitayama et al. (2007) proposed that a central component of these models is the representation of self and other relevant people. For instance, individuals whose dominant sociocultural model is independent (i.e. typical Westerners) tend to view others as important only to the extent that they are seen as instrumental in the realization of one's own goal and, as a result, the self representation is more elaborated than is the representation of (close) others. In contrast, individuals whose dominant sociocultural model is interdependent (i.e. typical Easterners) tend to view the self as defined and made meaningful in reference to others and, consequently, the representations of self and (close) others are equally elaborated. This distinction had been found previously in behavioural studies (reviewed in Zhu & Han, 2008). Utilizing the self-reference task, for instance, Western participants remembered information presented in the self-reference task better than information presented in the other-reference task even when the other was closely related (e.g. mother; Klein, Loftus, & Burton, 1989). In contrast, Chinese participants remembered information equally well when it was associated with the self-reference and other-reference tasks (Zhu & Zhang, 2002). The behavioural response represents the final outcome of a series of component processes, however, which makes it difficult to discern unequivocally the specific cause of the observed cultural differences. To address this question, fMRI has been used.

Specifically, Zhu, Zhang, Fan, and Han (2007) scanned both Chinese and Western participants while they carried out a simple trait-judgment task that required they decide whether a personal trait described themselves (self-reference task), their mother (close other-reference task), or a public figure (public other-reference task). The mPFC, related to the representation of self-knowledge, served as the region of interest (ROI). Results indicated that Western participants showed stronger activation of the mPFC in the self-reference condition than in either other-reference conditions, with comparable mPFC activation in the latter two conditions. The Chinese participants, in contrast, showed stronger activation of the mPFC in the self-reference and close-other reference conditions than in the public-other condition, and mPFC activation did not differ in the former two conditions. Additional contrasts indicated that among Chinese participants there was no brain region that differentiated the self-reference condition from the close-other reference condition. These findings conceptually replicate the behavioural findings and provide additional evidence that the distinction exists in the cognitive representations of self and others that are acquired from one's culture (Zhu et al., 2007).

In a follow-up study, Ng et al. (2010) demonstrated that in the same bicultural individuals, the neural substrates underlying the representation of close-other can either overlap with or distinguish from those underlying the self-representation depending on whether participants were exposed to primes associated with Chinese culture or Western culture prior to the fMRI scan. Given its experimental nature, this study provides causal evidence of the influence of culture on the cognitive structure of the sociocultural model.

Integrating fMRI with the ecocultural approach

The ecocultural approach is concerned with how the ecological and sociopolitical contexts, as well as the population-level biological and cultural adaptations to these two contexts, exert influence on individual-level mental and behavioural processes via mediating processes, such as genetic and cultural transmissions (Berry, 2003). Unlike the other four approaches, the ecocultural framework explicitly emphasizes the integration of biological and cultural factors by including both population and individual level biological processes, such as biological adaptation and genetic transmission (Berry, 2002). In addition, proponents of the ecocultural perspective have tended to emphasize the effects of particular features of the ecological or sociopolitical context (Markus & Hamedani, 2007). For instance, Mishra, Dasen, and Niraula (2003) have found that geological conditions (e.g. flatland vs mountainous areas) induce linguistic variations in spatial orientation systems (e.g. relative directions vs cardinal directions) among residents and that these differences affect performance on spatial tasks. Furthermore, a series of studies by Berry (2002) indicate that the dominant form of economic activity (i.e. hunting-gathering vs agriculture) is related to the cognitive style (field independence vs field dependence) that is typical of the members in a culture. To explain these differences, Berry (2002) theorized that the relatively loose social structure of nomadic hunter society leads to the socialization of assertiveness and, hence, promotes field independence, whereas the relatively tight and stable social structure of agriculture society leads to the socialization of compliance and, hence, promotes field dependence.

Proponents of the ecocultural perspective have also tended to emphasize cultural transmissions (e.g. educational systems) that link ecological and sociopolitical contexts to psychological processes (Markus & Hamedani, 2007). Two major types of cultural transmission have been distinguished, namely enculturation and acculturation (Berry, 2002). If the transmission process takes place within one's own culture, then it is termed enculturation. However, if the process is brought into play during contact with a different culture (e.g. immigrating to a new country), then it is termed acculturation. Although more is known about enculturation than acculturation processes, the latter has been gaining research attention (Berry, 2002). In a recent review article, Han and Northoff (2008) note that fMRI permits the investigation of neural differences between native and immigrant individuals to address questions regarding the effects of acculturation on neural and psychological processes.

To our knowledge, there is currently no published fMRI study that falls under the ecocultural approach. However, incorporating fMRI into research has the potential to illuminate some of the component processes underlying cultural transmission. For instance, Henrich et al. (2005) carried out a cross-cultural study in 15 small-scale societies using the ultimatum game. In this game, two players are paired and the first player is endowed with a sum of money to divide between the dyad. The first can offer a portion of the total money to the second player and the second player can either accept or reject this offer. If the second player rejects, neither player receives anything. If the second player accepts, the money is split according to the offer. Results indicated that the offers made by the first player differed predictably across societies based on two facets of their society's economic system: the degree of market integration, and the prevalence of extrafamilial cooperative economic institutions. In particular, participants from societies where market transactions are relatively commonplace and economic life is more dependent on cooperation with non-kin members (e.g. whale hunting) were more likely to offer equal distributions (i.e. 50:50 split) of the endowment. The authors interpreted these data as pointing to the important role played by the economic organization of the given groups in producing between-group differences in psychological constructs, such as fairness, moral empathy and equity, which, in turn, influence social preferences.

The data from an experimental game reflects the outcome of a decision, not the psychological processes underlying the decision, and alternative interpretations are possible. For instance, Krupp et al. (2005) suggested that the ultimatum game may not measure the same construct across cultures. It is conceivable that brain imaging could address these concerns. For instance, Shamay-Tsoory, Tomer, Berger, Goldsher, and Aharon-Peretz (2005) found that the ventromedial prefrontal cortex was related to affective empathic ability (related to moral empathy), whereas Sanfey, Rilling, Aronson, Nystrom, and Cohen (2003) found unfairness to be associated with activation in the insula. If the ultimatum game is tapping culturally shaped perceptions of moral empathy and fairness, then making the participants the recipient of the offers should reveal differential activation of these brain regions to the same offer (e.g. a $3/$7 split should lead to greater insula activation in participants from societies whose economies are based on joint cooperation than in participants from societies whose economies are not).

Caveats

The fMRI technology, which has been termed by some as the ‘tool for the future’ (Ramsoy, Balslev, & Paulson, 2007), has the potential to contribute to cross-cultural research. Such progress will be steadier, however, if the current limitations of fMRI are also recognized and alternative methodologies are used where appropriate. We therefore review briefly some of the issues that warrant consideration.

Time scale

The fMRI provides spatial resolution that exceeds what had been possible previously in research on healthy humans, but the same is not true of its temporal resolution. The temporal resolution of fMRI is limited by the haemodynamic response function upon which the signal change is based. As a result, if the interval separating the brain events is in the order of tens of milliseconds, these events may not be distinguishable by fMRI. In addition, the order in which brain regions are activated can also be difficult to determine using fMRI. Current practice is to collect fMRI and event-related potentials (ERP) because the latter can provide information of superior temporal resolution.

Granularity issues

The description of fMRI as having high spatial resolution is relative. fMRI is capable of identifying general areas of differential haemodynamic activation, but it is limited when one wishes to draw conclusions about subregions. Many well-defined brain structures are far from functionally homogeneous, but, instead, consist of subnuclei each of which makes a unique contribution to function. For example, the lateral nucleus of the amygdala plays a major role in the integration of information from multiple sensory channels, whereas the medial nucleus of the amygdala is involved in the expression of those innate behaviours controlled by the brainstem (LeDoux, 2008). With the spatial resolution of current fMRI (typically in the range of 3 mm), researchers could locate changes in brain activity to the level of the amygdala but not at the level of the subnuclei.

Susceptibility artifacts

The images of the working of the brain created by blood oxygen level-dependent fMRI (BOLD fMRI) are liable to be distorted by magnetic field inhomogeneities, especially in regions adjacent to air-tissue interfaces (Ramsoy et al., 2007). These distortions are termed susceptibility artifacts, which are often found in the medial temporal lobe, ventral prefrontal cortex, and orbitofrontal cortex—regions that are involved in many social and emotional processes. For instance, the hippocampus, the crucial component in memory network and the orbitofrontal cortex, part of the punishment-reward network, are all vulnerable to the corruption of susceptibility artifacts. Researchers should be aware of the possibility of signal losses when interpreting null findings in regions affected by susceptibility artifacts.

Do not equate effort with work

fMRI, as it is commonly used, measures changes in haemodynamic events which index metabolic demands of active neurons. However, the amount of energy consumed by neurons is not always positively correlated with the amount of work accomplished by them. Several studies have demonstrated that task fluency is typically associated with reduced activation (e.g. Milham et al., 2003). When interpreting fMRI data in cross-cultural studies, researchers should keep this distinction in mind. For instance, suppose you scanned participants from cultures A and B while they carry out a decision-making task that could but need not involve arithmetic calculations. If the brain regions underlying numerical processing were found to be more active in participants from culture A than from culture B, it would be tempting to conclude that the participants from culture A adopted a more quantitative approach in their decision-making. However, it is equally plausible that participants from both cultures adopted the same quantitative approach to the problem with the greater activation in culture A participants reflecting the fact that they were not as adept at carrying out the numerical calculations as their culture B counterparts.

Cerebra are only part of the story

The functional organization of the central nervous system, which includes the brain and spinal cord, constitutes a multi-levelled structure (Berntson & Cacioppo, 2009). A given behavioural function that is represented in the lowest level (e.g. behavioural withdrawal mediated by a spinal cord withdrawal reflex) may be re-represented at successively higher levels of the neuraxis (e.g. behavioural withdrawal mediated by a cost–benefit analysis implemented in the neocortex). Both low-level and high-level representations contribute to behavioural manifestations. However, fMRI which is more sensitive to higher than lower levels of the neuraxis, may mislead one to believe that lower-level mechanisms are unimportant or inoperative.

fMRI is not scientific phrenology

Those unfamiliar with the neurosciences tend to hold an oversimplified idea of functional localization which is promulgated by misrepresentations of scientific work in the popular media (Huettel, Song, & McCarthy, 2004). Yet, the empirical evidence is more consistent with the view that psychological states and processes are mediated by a network of distributed, often recursively connected, interacting brain regions, with the different areas making specific, often task-modulated contributions (see Cacioppo, Berntson, & Nusbaum, 2008; Poeppel & Monahan, 2008). If a single brain area is found to be activated by a task, it does not necessarily imply that this region is the seat of the information process supposedly tapped by the task; instead, it suggests only that this region may be part of a widely distributed network jointly responsible for the observed process. More and more fMRI studies are moving beyond simply identifying localized brain regions that are differentially activated by a task to include structural and functional connectivity mapping (Huettel et al., 2004). As we have seen, such analyses can make important contributions to our understanding of cultural differences.

Small sample size problem

Due to the costs of scanning, fMRI studies tend to use small samples. One major problem with small sample size is low statistical power, which means only large effect sizes are likely to be detectable. For instance, the power to detect a conventionally large effect size of d = 0.8 using a paired t-test in a sample of 20 subjects tested at an alpha level of 0.001 is 40% (Yarkoni & Braver, in press).

Another related problem is that the small sample may not be particularly representative of the population where it is drawn due to sampling error (see also Yarkoni & Braver, in press). Concerns about generalizability from a study sample to the cultural population are heightened further when convenience rather than population-based samples are used. Research in cultural neuroscience using larger, population-based samples would be helpful to address these concerns.

Two biasing assumptions

Interpretations of fMRI data typically rest on two seemingly axiomatic assumptions: (i) if a brain region is differentially activated during a specific information-processing operation, this region is involved in this information-processing operation; and (ii) if a region is not active during a specific information-processing operation, this region is irrelevant for this information-processing operation. Cacioppo et al. (2003) noted that the uncritical acceptance of these assumptions can lead to incorrect interpretations of fMRI data.

Regarding the first assumption, consider the work by Phelps et al. (2000), who reported that amygdala activity covaried with relatively negative evaluations of black people versus white people. This led them to interpret the activity of the amygdala as important in the implicit affective response to African Americans. In a subsequent study, however, Phelps, Cannistraci, and Cunningham (2003) tested patients suffering from bilateral amygdala lesions and found that these patients demonstrated the same implicit affective responses to African Americans as control participants.

A simple computer metaphor can be used to illustrate why the second assumption should not be accepted uncritically, either. The magnetic disks housed in a computer's hard drive revolve quickly when the user retrieves data. However, data can also be read from other storage devices, in which case the hard drive may be inactive. This inactivity only implies that hard drives are not involved in data retrieval in this particular context. The same can be said of a given brain region.

Affirming the consequence

A potential inferential error in fMRI studies is affirming the consequence, also called the converse error (Sarter, Berntson, & Cacioppo, 1996). Briefly, given that a particular information-processing operation produces regional brain activation, it does not follow that activation of this region means it was elicited by this information-processing operation. To see why this type of inference is error-bound, suppose that 90% of previous studies involving manipulation of pleasure found that ventral caudate (VC) activity is heightened when pleasure is induced (i.e. P(VC activity | pleasure) = 0.9). However, among all the other studies concerned with psychological states other than pleasure, 40% of them found activity in the VC (i.e. P(VC activity | non-pleasure) = 0.4). Now, assume that across all possible psychological states, the base rate for a given person experiencing pleasure at any given time is 0.1 (i.e. P(pleasure) = 0.1). If a study was carried out in which contrasting participants' brain response to pictures of ingroup versus outgroup members resulted in a strong signal in the VC, would this response mean that ingroup members evoked more pleasure than outgroup members? The answer is probably not, as calculations based on Bayes' theorem show that the likelihood the ingroup elicited pleasure is only 0.2 (i.e. P(pleasure | VC activity) = 0.2).

Converging methods

The purpose of elaborating on the limitations circumscribing fMRI is not to portend a pessimistic future of fMRI investigation. Instead, we believe that to take full advantage of fMRI, we need to understand and address these weaknesses, including through the use of complementary methods that help address these weaknesses. For instance, the weaknesses of fMRI and other neuroscientific tools provide the justification for advocating a convergent approach to cultural neuroscience (Cacioppo & Decety, 2009). Because different measures have their own strengths and shortcomings, it can be important to consider converging evidence from different methodologies, such as fMRI, animal studies and electrophysiology (Chiao & Ambady, 2007; Ramsoy et al., 2007).

Among some of the neuroscience methods that have been helpful are animal models, ERP, magnetoencephalography (MEG), positron emission tomography (PET), transcranial magnetic stimulation (TMS), and lesion studies. Detailed accounts for each of the techniques are beyond the scope of this article and interested readers can refer to several excellent sources for further information (e.g. Berntson & Cacioppo, 2009; Cacioppo, Tassinary, & Berntson, 2007; Huettel et al., 2004; Ramsoy et al., 2007). Here, we consider two related studies that illustrate how different tools complement each other and, when jointly deployed, enrich our understanding of important questions in cultural neuroscience.

A group led by Nan et al. (2006) used ERP to investigate cross-cultural music phrase perception. A group of German musicians listened to bi-phrase excerpts from previously unknown musical pieces of either Chinese or Western origin. A control group of ‘bicultural’ musicians (i.e. Chinese musicians who received formal training in Western classical music) were also included to isolate differences in cultural familiarity from the differences in basic acoustic properties between the two types of music. Although the neural activities underlying phrase processing were generally comparable when listening to native music (i.e. Western music) and alien music (i.e. Chinese music), a modulating effect of culture did emerge in a relatively early time window (i.e. 100–450 ms after the offset of inter-phrasal pauses). Given the high temporal resolution of ERP, researchers were able to pinpoint when cultural factors were making an impact. But, what is the nature of this putative cultural influence? The author posited that during this early processing stage, neural activities were partially driven by a top-down process which involves the activation of cultural-specific knowledge. However, this was only a speculation because the lack of spatial precision of ERP obscured the neural sources of this culturally dependent activity which may hold the answer to the question.

In a follow-up study, with the aid of fMRI, Nan, Knosche, Zysset, and Friederici (2008) found evidence consistent with this hypothesis. Again, a sample of German musicians were scanned while listening to either unknown Western music or unknown Chinese music. In comparison with Chinese music, Western music activated a brain area which has been attributed a function regarding the subjective sense of familiarity (Schnyer et al., 2004). However, given that all the melodies used in the study were unknown to participants, the authors reasoned that the subjective sense of familiarity was derived from the activation of the ‘general Western musical schema,’ consistent with the hypothesis they derived from their ERP data. Although more research is needed to confirm their hypothesis, these studies exemplify the value of using convergence to investigate the complexities underlying the effects of culture on brain and behaviour.

Concluding comments

By emphasizing the interdependence between mind and cultural context, sociocultural psychologists advocate thinking beyond the person and attending to superorganismal structures, such as culture. Yet, a strict social-level analysis will be, at best, incomplete. It is a well-grounded reality that engagement with the meanings and practices of the social world makes us uniquely human beings, yet there is no denial that we are also biological beings. Key principles from social neuroscience (i.e. multiple determinism, non-additive determinism and reciprocal determinism) suggest that a more comprehensive explanation of the mind and behaviour will be promoted by the integration of biological and social-cultural approaches (Cacioppo & Berntson, 1992). Such a view is borne out by the strides made in the nascent field of cultural neuroscience, which has already produced important insights into the nature of the contingency between the psychological and the sociocultural by bringing together biological, cognitive and social levels of analysis (Chiao, Zhang, & Harada, 2008).

Our discussion of the deployment of fMRI in the study of the cultural mind was intended to demonstrate the potential of neuroscientific methods and theories to extend the investigation of culturally bounded information processing operations beyond the vistas accessible through traditional means. Although we couched our discussion in the five theoretical frameworks currently dominant in sociocultural psychology, we did so to illustrate opportunities for advances and to encourage ground-breaking theoretical insights. Moreover, we noted that fMRI, like any measurement approach, has limitations as well as strengths, and we illustrated both to encourage progress in the field. The complexities associated with the design of fMRI experiments and the analysis and interpretation of fMRI data are no reason to discount the value of this tool, but it would be equally naïve to treat fMRI as a panacea for the various limitations associated with self-report and behavioural measures (Cacioppo et al., 2004). Neuroscience proffers a plethora of techniques to probe the mind, which provide information complementary, rather than redundant, with fMRI data. It is the combination of these converging methods in combination with traditional methods and measures that may provide the best opportunity for advances in the burgeoning field of cultural neuroscience.

Acknowledgements

Preparation of this article was supported by grants from the National Institute of Mental Health (Grant No. P50 MH72850) and the John Templeton Foundation.

Ancillary