Cultural conceptions of morality: Examining laypeople’s associations of moral character

Whether moral conceptions are universal or culture-specific is controversial in moral psychology. One option is to refrain from imposing theoretical constraints and to ask laypeople from different cultures how they conceptualize morality. Our article adopts this approach by examining laypeople’s associations of moral character in individualistic- and collectivistic-oriented cultures. Using correspondence analysis we found that the concept of moral character yielded widely shared associations with justice and welfare concerns. Yet, there were also clear cultural differences with individualistic-oriented samples associating more frequently rights-based features and collectivistic-oriented samples more frequently associating duty-based attributes. When matching freelisted trait categories with Schwartz’s value types, moral value hierarchies were similar across cultures and correlated significantly with explicit moral value ratings. We conclude that imposing constraints through an expert-designed category system can narrow the scope of inquiry to common moral aspects related to problem-solving, promotion of prosocial actions and control of antisocial behaviour.


Introduction
Morality is most controversial when it comes to the question of whether it is universal or culturally relative (cf. Frimer & Walker, 2008). Moral psychology has a long history of advocating a universalism position (Kohlberg, 1984;Turiel, 1983), yet more recent theorizing suggests that there are both universal and culturally specific aspects (e.g., Haidt, 2008). Despite the recent theoretical advancements in this area, there is still a great deal of ambiguity as to which elements of the human moral system are universal and which ones are culture-specific. Divergent findings that lend support to either a moral universalism or relativism position may be due to different definitions of morality and divergent methodological approaches in studying a moral domain. In this article, we aim to bring some clarity to the matter by carefully reviewing the literature and presenting a synthesis from which we derive our hypothesis on moral universals and specifics. We then empirically compare how individuals from four different cultures conceptualize morality and whether similarities and differences are in line with our theoretical proposition.
More specifically, we are interested in how laypeople conceive of morality when they are not given external cues. Most studies use expert-designed taxonomies and self-report measures to assess individuals' moral conceptions (Chiu, Dweck, Tong, & Fu, 1997;Colby & Kohlberg, 1987;Graham, Haidt, & Nosek, 2009;Nucci, Turiel, & Encarnacion-Gawrych, 1983). Yet, this may narrow the focus of study to what the experts define as moral. By using a layperson's perspective we circumvent this problem and leave it to the respondents to determine what does and does not qualify as moral. We also use the opportunity to compare our results to data from the Schwartz Value Survey (SVS; Schwartz, 1992), an expert-designed value taxonomy that contains value types theorized to belong to the moral domain (Helkama, 2004). We aim to identify the methodological implications of adopting an expert-versus lay-guided approach.

Moral universalism
The dominant moral universalism perspective posits that there is a universal morality that can be attained through rational reflection (Kohlberg, 1984;Turiel, 1983). Justice, rights and welfare are seen as the pinnacle of moral maturity and used for definitions of what belongs to the moral domain and what does not. The moral domain has usually been assessed through moral reasoning. A very different line of research ( i.e. values research based on Schwartz's, 1992 values theory) has also come to the conclusion that there is a set of universal moral values; however these are not limited to justice, rights or welfare concerns. Schwartz (1992) developed an expert-designed category system and value survey. He empirically corroborated the existence of 10 value types in more than 60 countries: power, achievement, hedonism, stimulation, self-direction, universalism, benevolence, conformity, tradition and security (e.g., Schwartz & Boehnke, 2004). Helkama (2004) proposed that the most important moral values that may be widely shared across cultures should be universalism, benevolence and conformity/tradition because they serve important moral functions in every culture: universalism serves moral problem-solving (resolving value conflicts and issues of justice), benevolence serves the promotion of prosocial actions (being kind and considerate towards others) and conformity/tradition serves the control of antisocial action (resistance to temptation).

Culture and morality
Findings from qualitative cross-cultural studies raised the question whether pluralistic conceptions of morality exist (e.g., Nisan, 1987). Consequently, Shweder, Much, Mahapatra, and Park (1997) developed the 'big three' theory of morality: the ethics of Autonomy, Community and Divinity. In the ethic of Autonomy moral agents are defined as independent people who are free to make their own choices, restricted only by concerns whether their behaviour would inflict harm on others or restrict their rights. The ethic of Community defines moral agents in terms of their interdependence or membership in social groups and the obligations that go along with membership. The ethic of Divinity defines moral agents as spiritual entities who aim to follow divine laws in an attempt to achieve moral purity. Although the ethic of Autonomy is apparent in both Western and non-Western moral discourse, the ethics of Community and Divinity are also important in collectivistic-oriented cultures (e.g., Haidt, Koller, & Dias, 1993;cf. Miller, Bersoff, & Harwood, 1990;Shweder et al., 1997).
Further developing this line of theory, Graham and colleagues (Graham et al., 2009) identified two major 'moral foundations': individualizing and binding. The individualizing foundation emphasizes justice, rights and welfare concerns of individuals and corresponds to Kohlberg's ethic of justice, Shweder and colleagues' (Shweder et al., 1997) ethic of autonomy and Schwartz's (1992) values of universalism and benevolence. The binding foundation emphasizes group concerns, such as duty, and self-control, and corresponds to Shweder's ethic of Community/ Divinity and Schwartz's values of conformity and tradition. The greatest differences between Eastern and Western cultures were found for the binding foundation (Graham et al., 2011).
We think that one important issue with this conceptualization of morality is that it conflates two different moral concerns, i.e. the concern about justice/welfare and the concern about rights. Whereas the former is likely to be widely shared, because the universal rules of social cooperation dictate to be fair, not to cheat, deceive or intentionally harm someone (Richerson & Boyd, 2005), the latter is likely to show some variation across cultures.

Rights and duties
Rights-and duty-based moral orientations define conceptions of an individual's choices and conduct in a society (Chiu et al., 1997;Dworkin, 1978). Duty-based cultures are more restrictive and individuals believe in a fixed socio-moral reality, (i.e., a rigid moral order prescribing what is right or wrong). The dominant moral orientation is toward supporting the status quo and therefore it is a system-oriented morality. Rights-based cultures are more flexible and believe in a malleable socio-moral reality, i.e. the authority of the existing moral order is no longer absolute and primary concern is to uphold fundamental human rights. The dominant moral orientation is towards promoting social change and it is therefore a personcentered morality. Chiu and colleagues (Chiu et al., 1997) corroborated that there is cultural variation in these moral orientations with individualistic-oriented samples prioritizing a rights-based morality and collectivistic-oriented samples a duty-based morality. Yet, contemporary theories on morality usually blend universal justice and welfare concerns with rights concerns which might in fact vary in importance depending on the culture.

Laypeople's conception of morality
One limitation of past research, which may have led to the blending of rights and justice/welfare concerns, is that researchers have mostly employed expert-designed measures or taxonomies to study individuals' moral orientation and possible cultural variations (e.g., Colby & Kohlberg, 1987;Graham et al., 2009;Nucci et al., 1983). Hence, researchers imposed their definition of morality onto their participants which may have narrowed the focus of study to what the experts regard as relevant. An alternative is to leave it entirely to the respondents as to what qualifies as moral. One way of doing so is to ask laypeople to freelist the attributes they associate with a concept's meaning (cf. Rosch, 1975). Group-generated frequencies of attributes can be regarded as an index of everyday accessibility in the particular sample in which it has been produced (see e.g., Gregg, Hart, Sedikides, & Kumashiro, 2008;Smith, Tü rk Smith, & Christopher, 2007). Moreover, since the analysis is not based at the individual-level, but group-level, the attributes describe the normative associative meaning of a term and reflect a form of 'collective wisdom' (Horowitz & Turan, 2008, p. 1059 and are therefore especially intriguing for cross-cultural comparisons. There are a few studies that examined laypeople's associations with moral concepts (see Hardy, Walker, Olsen, Skalski, & Basinger, 2011;Lapsley & Lasky, 2001;Walker & Pitts, 1998;Walker & Hennig, 2004). Yet, there are a number of shortcomings that motivated us to conduct the present study. First, the possibility that people from different cultures could have different conceptions of morality was not examined. Second, examining what laypeople associate with morality was not the focus in past cross-cultural studies. Smith et al. (2007) examined associations with the 'good person' across seven cultures, but not associations with a person of moral character. This is problematic as target concepts may not be interchangeable (Lapsley & Lasky, 2001). The authors also categorized the freelisted responses into established category systems, using Schwartz (1992) 10 value types, in order to content analyze and compare the freelistings across cultures. They found that freelisted attributes related to benevolence, conformity, tradition and universalism were highly accessible descriptors of a good person across cultures. On the other hand, attributes related to self-direction and achievement varied widely in terms of their accessibility. This differs somewhat from Schwartz and Bardi (2001) who found that group-based value hierarchies based on explicit personal value ratings were strikingly similar across cultures. Given the normative relevance of morality, there might be more consensus in the moral value hierarchy if it were based on freelistings referring to a moral concept. It is also conceivable that this moral value hierarchy would correlate more highly with explicit value ratings that assess the normative and not the personal relevance of values.

Summary of research goals and hypotheses
Our aim was to study freelisted associations with moral character across cultures. We first examined cultural similarities and differences in moral attributes and whether they fit recent theorizing on moral universalism and relativism. We expected that attributes related to justice and welfare would be widely shared, but that attributes related to duties would be mentioned more often in the collectivistic-oriented samples and that attributes related to rights would be more prominent in the individualistic-oriented samples.
We also compared moral value hierarchies obtained implicitly through the freelisting procedure and explicitly through value ratings. We expected moral value hierarchies based on freelistings to be correlated with those based on explicit normative ratings of values. We also expected that the moral value hierarchies would stress universalism, benevolence and conformity, since these values serve universal functions of morality (see Helkama, 2004).

Participants
A total of 458 undergraduate university students from the social sciences participated in this study. Participants were recruited from two individualistic-(New Zealand, Germany) and two collectivistic-oriented cultures (The Philippines, Brazil) according to past research (see Hofstede, 2001;Schwartz, 2006). Respondents were only included in the analysis if they identified with the country of residence as assessed in the survey. Hence, responses were analyzed from 98 Brazilians, 119 Germans, 108 Filipinos and 90 students from New Zealand (see Table 1).
There were significant age differences across the four samples, F(3, 425) = 72.93, p < .001, and differences in the proportion of females, χ²(3) = 16.39, p < .001. Age-group differences correspond roughly to differences in student profiles as a result of different university entry requirements in each country. The samples reflect the typical gender composition in the social sciences with females being often in the majority.

Procedure
Participants were first asked to provide basic demographic information. They were then asked to write down what kinds of characteristics they associate with a 'moral character' after reading an instruction which was adapted from Lapsley and Lasky (2001). They were given five minutes and a page with six lines and bullet points and were instructed to freelist one attribute per line. This was done in order to encourage participants to write down single attributes and not to engage in essaylike explanations about moral character. We chose six bullet points as this corresponds to the average number of attributes that were reported in one of our unpublished freelisting studies on the good person. We gave participants only 5 minutes because we wanted to examine which attributes are readily accessible and not the ones that come to mind when participants have some time for contemplation.
Participants from Germany received the instruction in German and those from Brazil in Portuguese. Freelisted attributes were translated into English using a committee approach (Harkness, 2003) with two bilingual translators for each language.
Analytical strategy Content analysis. The number of elicited traits was 590 for the Filipino moral target sample, 496 for the New Zealand sample, 608 for the German and 512 for the Brazilian sample. Judgment rules (see also Fehr, 1988;Walker & Pitts, 1998) were developed in order to facilitate the final cross-cultural frequency analysis of moral attributes. These rules were used in two different stages: the first stage consisted of an intra-cultural analysis in which frequencies of moral attributes were analyzed separately in each culture; and the second stage of a pan-cultural analysis of freelistings in which all freelisted features were pooled. The aim during the whole procedure was to: (1) reduce the list of attributes as much as possible, while at the same time remaining as conservative as necessary to preserve culture-specific features; and (2) to categorize specific traits into more general attributes, so that a computerized frequency analysis could detect and count these attributes.
The intra-cultural analysis consisted of the following steps: dividing whole freelisted sentences into linguistic units; separating compound phrases into separate descriptors if they could stand alone (e.g. just and open-minded were separated into

Cultural conceptions of morality 59
just and open-minded ); dropping modifiers (e.g. very loving became loving); converting nouns into adjectives whenever possible (e.g. courage became courageous); and summarizing direct antonym pairs (e.g. fair and unfair were both counted under the attribute of fair since they belong to the same underlying meaning of fairness; decision about the polarity of this attribute was determined by the most frequently mentioned pole). The remaining attributes were then analyzed with the software program MAXQDA 2007 (Kuckartz, 2007) in terms of how frequently they occurred. Idiosyncratic features (i.e. those that were mentioned by only one person) and also attributes not directly related to the question (e.g. title of a children's book) were dropped. This was done separately for each sample. The responses were then pooled across samples and further analyzed in a pancultural fashion by two independent raters. The main aim herein was to identify synonyms of moral attributes. These synonyms were then summarized under a moral trait category. Pooling the attributes across samples ensured that this could be done without 'cultural bias', i.e. without treating an attribute as culture-specific when in fact it was synonymous to a similar attribute mentioned in other cultural samples. For instance, stable was only mentioned in the Brazilian sample; however, well-balanced was identified as a synonym occurring in the German and Filipino sample. Hence, the moral trait category was labelled well-balanced (because wellbalanced occurred more frequently than stable) and Brazilian responses which yielded the trait stable were counted as falling under the category of well-balanced. Discrepancies in the categorization by the two raters were discussed with a colleague, who is a native speaker of English and who was naïve to the hypothesis, until agreement and consistency in the categorization and labelling of the categories was achieved. The final solution consisted of 49 distinct moral trait categories. The reliability of the 49 categories was tested with a third judge who was instructed to rate on a 3-point scale (1 = fitting, 0 = debatable, -1 = unfitting) how well the attributes from the pan-cultural data fitted into each of the 49 categories (cf. Gregg et al., 2008). In order to keep this task manageable for the judge, we only requested it for half of all attributes which were selected randomly. A total of 88% of the categorizations were judged as fitting, 10% as debatable and 2% as unfitting. Similar percentages have been found previously, so that we concluded that our categorization scheme was reliable.
The frequencies of the final attributes in the respective trait categories were summed. Only categories that were mentioned by at least 5% of the respondents in any one of the samples were retained and further analyzed cross-culturally.
Central vs peripheral trait categories. We analyzed moral attributes that emerged as central in each cultural sample. Central attributes are those that are highly accessible and therefore occur with a relatively high frequency compared to peripheral attributes (cf. Gregg et al., 2008;Smith et al., 2007). A scree test was used as an aid to differentiate between central (frequent) and peripheral (non-frequent) trait categories in which the relative frequencies were arranged in descending order along the ordinate and the features along the abscissa. The cut-off point between central and peripheral trait categories is where there is a substantial change in the gradient of the slope. We then applied a correspondence analysis on the central trait categories.
The scree test yielded 11 central trait categories in the Brazilian and German sample, 12 in the New Zealand and 9 in the Filipino sample. Central and peripheral trait categories for all four cultural groups and the associated frequencies can be seen in Table 2.
Correspondence analysis. We applied a correspondence analysis to the central trait categories from all four cultural samples. Correspondence analysis is a multidimensional scaling technique that creates a perceptual map. A biplot is commonly used as a visual representation of the data. It displays both row and column points (here trait categories and cultural samples) in a single multidimensional map based on the statistical association between them. In order to compare frequencies of column and row profiles they need to be standardized. We chose symmetrical normalization because it facilitates the interpretation of the biplot: row/column points that are close together are more alike than points that are far apart. Yet, it is important to note that even under this standardization procedure, one cannot precisely interpret the distance between a row point and a column point. It is preferable to make rather general interpretations, for instance, whether row and column points appear in the same quadrant of the multidimensional space (Hair, Black, Babin, Anderson, & Tatham, 2006). In our case, this method allowed us: (1) to statistically test whether there is a significant relationship between the samples and trait categories; (2) to visually inspect how similar or different cultural samples are in their freelisted associations; and (3) what kinds of trait categories are associated with which samples.

Comparing laypeople's associations of moral character across cultures
The results show that honest was a shared central moral trait category (occurring in all four cultural samples), followed by friendly, good and just (central in three samples). These attributes are consistent with the moral universalism perspective that morality is a matter of welfare (to be friendly and good to others) as well as fairness (e.g. to be honest; see also Vauclair & Fischer, 2011) and justice. Culturally idiosyncratic trait categories were open, critical, reliable and obedient in the German sample and correct, serious, educated, responsible, loyal and sociable in the Brazilian sample. Well-mannered was an idiosyncratic central moral category in the Filipino sample and caring, strong, and trustworthy were central and specific to the New Zealand sample.
Correlating the frequencies of the central moral trait categories of the different samples with each other showed that none of the samples were significantly similar to each other, except for the New Zealand and Filipino sample which reached Cultural conceptions of morality 61 marginal significance (r s = .41, p = .05). The correspondence analysis on all central moral trait categories across the four cultural samples produced a significant chi-square ( χ 2 (75) = 745.23, p < .001), therefore there was a significant relationship between the samples and trait categories. The first and second dimensions accounted for a cumulative total of 73.64% of the inertia and were therefore retained. The resulting two-dimensional biplot is depicted in Figure 1.
The triangular shape of the biplot is due to the distinctive responses of the German, Filipino and Brazilian sample, while traits shared across the four cultural samples are located at the centroid of the map. Culturally idiosyncratic traits from these three samples occurred with relatively high frequencies in the respective samples and relatively low frequencies in all the other samples; this tendency was somewhat less accentuated in the case of the New Zealand sample. The biplot shows indeed that New Zealand has the most overlap with all other samples which is somewhat unexpected, the Filipino and Brazilian sample overlap more with each other than they do with the German sample and the German sample has the least overlap with all other samples.
Taking the trait categories in the triangle of the biplot as an anchor for interpretation, we found that Brazilian trait categories are more about the 'moral self' (e.g. being serious, educated), whereas Filipino, German and New Zealand trait categories are more about relations to others (e.g. being kind, helpful and trustworthy). Walker and Pitts (1998) also reported a self-other distinction in their study on attributes of a moral person, however, they found self-focused attributes that are more directly related to morality (e.g. being righteous, having high standards, p. 414). The self-focused associations of the Brazilian sample are only distantly related to a person with moral character and could in fact be applied to a number of other persons, too. This may point to a broader definition of morality in the Brazilian culture. Yet, given that this finding is somewhat unexpected, we are cautious in interpreting it any further.
As expected, German trait categories that are highly central are those that reflect a rights-based morality, characterized by moral trait categories that indicate an individualized and flexible orientation towards moral standards (e.g. being Note: Features in boldface refer to those that were determined to be central moral attributes. Percentages are rounded to the nearest full percent. The percentages do not sum to 100% as respondents were encouraged to list more than one attribute. * 'personality' in the sense of backbone or fortitude.

Cultural conceptions of morality 63
open-minded, a trait category that was mostly defined by the freelisted attributes 'tolerant'; and being critical which was mainly freelisted as 'critical thinking'). Both the Brazilian and Filipino trait categories have in common that they deal more with a 'duty-based' morality (e.g. being religious, well-mannered and responsible). Their perception of a moral character reflects a more obligatory and communal kind of morality targeted at maintaining already existing moral standards. The pattern of association remained virtually the same when we repeated the analyses with all 49 moral trait categories (more details can be obtained from the first author).   * values that are usually excluded from cross-cultural comparison due to their unstable meaning (Schwartz, 1992).

Comparison of freelisted moral trait categories with explicit value ratings
We compared moral values that are implicitly assessed with the freelisting procedure to explicit value ratings. In order to do so, we matched the moral trait categories to the 10 value types from the SVS (Schwartz, 1992). We did not use the more refined 19 value types (Schwartz et al., 2012) because each value type would have consisted of very few matched moral trait categories and our purpose was not to examine the motivational dynamics of values in greater detail. Only values that Schwartz (1992) found to be equivalent across cultures were included in the matching procedure. To enhance the number of matches, moral trait categories that occurred with a frequency of less than 5% in all samples were also included. When matching these attributes, we were guided not only by the trait category and value label, but also by all freelisted attributes that make up a moral trait category and the defining phrases of the values from the SVS (e.g. the moral trait category well-mannered consists of the freelisted attributes 'courteous, tactful, polite, well-raised' and therefore corresponds to the Schwartz value politeness, defined as 'courtesy, good manners'). There were moral trait categories that corresponded directly to a value from the SVS (e.g. helpful) and moral trait categories that were semantically equivalent (e.g. the trait category conservative/conventional corresponding to the SVS value respect for tradition defined as 'preservation of time-honoured customs'). A total of 36 moral trait categories corresponded to values in the SVS and were further categorized into value types according to Schwartz's (1992) value theory . Note that a number of trait categories could not be matched with a Schwartz value (e.g. being friendly, generous, kind ) which points to a set of moral values that is not assessed in the SVS. Table 3 provides an overview of pan-cultural and culture-specific frequencies of the moral trait categories falling under Schwartz's value types. Note that hedonism values did not occur at all. Some of the value types were accessed significantly more frequently than others (pan-cultural sample: χ 2 (8) = 266.09, p < .001; Germany: χ 2 (7) = 53.54, p < .001; New Zealand: χ 2 (7) = 70.00, p < .001; Brazil: χ 2 (5) = 66.75, p < .001; The Philippines: χ 2 (7) = 56.13, p < .001); computed on frequencies corrected for number of values in a value type).
Moral trait categories that were accessed most frequently, after adjusting by both number of participants and number of values in the respective value type, belonged to benevolence followed by conformity and universalism values. One exception is the Filipino sample for which moral trait categories belonging to tradition and conformity value types were accessed most frequently.
As can be seen from Table 3, the rank ordering of the moral trait categories is fairly similar across cultural samples. The average Spearman correlation between cultural samples was relatively high with a value of .88 (for frequencies adjusted by the number of participants in the respective sample) and .80 (for frequencies adjusted by both number of participants and number of values in the respective value type). Therefore matching moral trait categories with Schwartz (1992) value types rendered the cultural samples more similar to each other.
In the next step, we explored whether the moral value hierarchy based on frequencies is associated with the value hierarchy based on explicit value ratings. Values are usually rated according to their personal importance; yet, they can also be rated according to their moral and normative relevance (Vauclair, 2010). We used secondary data in which values have been assessed in different ways. First, we used data from Schwartz and Bardi (2001) pan-cultural value hierarchy (across 54 nations) based on students' self-importance ratings of the SVS to obtain a highly reliable value hierarchy. In addition, we used culture-specific data from an unpublished values study (Vauclair, 2010): students from Brazil (N = 141, M = 27.20, SD = 9.11, 67% female), Germany (N = 68, M = 26.92, SD = 6.54, 75% female), New Zealand (N = 88, M = 23.13, SD = 5.30, 73% female) and the Note: † p < .10, * p < .05, two-tailed, N = 9 value types. IMP = value hierarchy based on the personal importance rating scale; MOR = value hierarchy based on the personal moral relevance rating scale; EXP = value hierarchy based on the societal expectation rating scale.

Cultural conceptions of morality 67
Philippines (N = 97, M = 20.14, SD = 1.22, 73% female) responded to a shortened version of the SVS and rated on three 7-point bipolar scales whether each of the values were: (1) personal value rating: 'important to me' vs 'unimportant to me' (comparable to the original SVS instruction); (2) moral value rating: 'a moral to me' vs an 'immoral value to me'; and (3) societal expectations ratings: 'something that I am expected to strive for in my society' vs 'something that is my free choice to strive for in my society'. All responses were recoded so that higher numbers reflected more personal importance, moral relevance or perceived societal expectation.
The previously gathered moral trait categories were not reliably related to personal value ratings or prescriptive value ratings. However, they correlated significantly with the moral value ratings (see Table 4). This correlation pattern was prevalent in both the pan-cultural sample as well as in the four culture-specific samples. Hence, people's associations of what it means to be a person with moral character seem to reflect their moral values (and not their perception of societal expectations or prescriptive norms). Moreover, the moral values hierarchy is remarkably similar across cultures when matched with Schwartz (1992) value types. Benevolence, universalism and conformity ranked highest which is consistent with Helkama's (2004) functional theory of morality.
To conclude, similarly to Smith et al.'s (2007) findings on freelisted associations of the good person, we found that: (1) attributes related to benevolence, conformity, tradition and universalism are highly accessible descriptors of moral character across different cultures; (2) there is a greater emphasis on tradition and conformity and a lesser emphasis on self-direction compared to Schwartz and Bardi's (2001) pan-cultural personal value hierarchy. Contrary to Smith et al.'s (2007) study, we found that: (1) hedonism values were not accessible at all as attributes of moral character; and (2) there was no evidence for the fact that some attributes (especially those related to self-direction and achievement) varied widely in terms of their accessibility across cultures. In three of the four cultures, the five least important values followed the same order: achievement, self-direction, stimulation, security and power (except in Brazil where power came after achievement).

General discussion
Our study showed that there are cultural differences and similarities in lay conceptions of moral character. Consistent with our expectations, we found that there were widely shared attributes which are related to issues of justice and welfare. This is in line with the moral universalism perspective (Kohlberg, 1984;Turiel, 1983) and corroborates the widespread 'human obsession' with fairness, reciprocity and justice (Graham et al., 2009) as well as the importance of social cooperation as emphasized by evolutionary theorists (Richerson & Boyd, 2005). As expected, we found that duty-based concerns, i.e. concerns about obligations, obedience and respect for (social or religious) authority, were mentioned more often in collectivistic-oriented samples. Yet, rights-based concerns, i.e. concerns about self-determined moral thinking that is not affected by any other (social or religious) authority, were more salient in the individualistic-oriented German sample. It is noteworthy, that the biplot of the correspondence analysis also shows trait categories that diverge from our prediction (e.g., obedient). As so often in psychological research, the findings show a trend and no clear cut results in absolute terms. It is the pattern of relative accessibility of trait categories that illustrates which moral orientation is more salient in which cultural sample. We also expected that the New Zealand sample would be more closely associated with the rightsbased concerns; however, the findings show that it shares many attributes with all other three samples. One tentative explanation might be that New Zealand's culture is a mix of individualistic (Pakeha culture, i.e., New Zealand European) and collectivistic tradition (Maori culture). This biculturalism might contribute to the finding that it bears elements of both individualistic and collectivistic cultures in the conceptualization of moral character.
The findings extend our knowledge on moral psychology in important ways. They point to both cultural similarities (justice and welfare) and an underlying bipolar dimension of morality (rights-vs duty-based concerns) that varies with culture. This distinction is not incorporated in contemporary moral theories and studies. For instance, Iyer, Koleva, Graham, Ditto, and Haidt (2012) recently stated that 'standard morality scales, including the Moral Foundations Questionnaire, do a poor job of measuring libertarian values' (Iyer et al., 2012, p. 10). Political libertarians strongly endorse individual liberty and the freedom to pursue their own happiness as their most important guiding principle which bears strong resemblance with the rights-based moral orientation we found in our study. Iyer et al. did not study cultural variation in libertarian values, yet there is another study that lends support to the notion that some cultures might be more focused on rights or duties which predicts their attitudes towards moralized issues (e.g., homosexuality, abortion, Vauclair & Fischer, 2011). Hence, incorporating a rights-based moral orientation in future studies might be important in order to get a better understanding of cultural variation in moral outlooks.
Finally, when moral trait categories were matched with Schwartz's (1992) 10 value types, moral value hierarchies based on freelistings correlated significantly with explicit ratings of moral values, but not with personal values or prescriptive value ratings. Moreover, the three most important moral values in the moral value hierarchy were benevolence, universalism and conformity supporting Helkama's (2004) reasoning that these values serve universal social functions of morality in regard to promotion of prosocial actions, moral problem-solving and control of antisocial action.
Forcing the freelisted moral attributes into Schwartz (1992) value types diminished the cultural differences and rendered the samples more similar to one another. We think that there are two reasons for this finding, one is theoretical and the other methodological in nature. First, universalism and benevolence refer to the previously claimed universal justice and welfare concerns (see e.g. Graham et al., 2009) and therefore it is not surprising that they ranked highly across sam-Cultural conceptions of morality 69 ples. Somewhat more contradictory is the rank order for conformity, given that we found an important cultural variation in the freelistings regarding the salience of duties. Conformity is in fact, to some extent, fundamental for every social group in ascertaining smooth social relations and has developed very early in human evolution (Moghaddam, Slocum, Finkel, Mor, & Harré, 2000). People from different cultures might commonly agree that conformity is important in upholding the moral order. However, there may be subtle differences in the meaning of conformity when it is assessed with different methods. Conformity, as assessed with the values of the SVS (e.g., politeness in the sense of good manners), may be interpreted as conforming to certain abstract standards or moral ground rules necessary for the smooth functioning of every social group. On the other hand, conformity as it emerges with the freelisting procedure, or in ethnographic interviews (Shweder et al., 1997), or more contextualized self-report measures on moral judgments (Graham et al., 2009), might be interpreted as conforming to social or religious authorities that define the moral standards. These two conceptualizations of conformity are fundamentally different in regard to the motivational underpinnings of human morality, juxtaposing an intrinsic and extrinsic motivation to conform. Second, closed-ended self-report questionnaires contain an important methodological limitation (see also Smith, Tü rk Smith, & Christopher, 2007). They are expert-designed category systems, that, as in the case of the SVS, ask respondents to rate very general and abstract goals. Regardless of culture, the general goals of morality are to promote and protect positive relations of self to others. This can be fulfilled with the values of universalism, benevolence and conformity and this is why respondents recognize these values as moral values across different cultures. However, if there are no constraints at all imposed upon respondents, people from different cultures can access their cultural conception of morality. For instance, being critical emerged as a trait category from an individualistic culture, yet it is not a value that appears in the SVS. Hence, the group-generated freelistings provide insight into which attributes are most salient in a specific culture without constraining individuals' responses to pre-defined categories and yields therefore a more culturally refined conceptualization of morality than expert-designed category systems can provide. Future cross-cultural research on naturalistic conceptions of morality could focus on other moral concepts (cf. Walker & Hennig, 2004). Additional quantitative measures, such as prototypicality ratings of freelisted attributes could also be used in future research.

Limitations
One limitation is the different proportion of females in the samples. Yet, reviews of previous research on morality (Walker, 1991) as well as studies examining laypeople's moral conceptions (Hardy et al., 2011;Walker & Pitts, 1998) have not found any strong evidence for gender differences. Hence, it seems very unlikely that it has induced a substantial bias in our study.
Another limitation concerns the age differences across our cultural samples. Previous research does not indicate age differences in the content of freelisted moral attributes between young, midlife and older adults (Walker & Pitts, 1998) nor between early and late adolescents (Hardy et al., 2011). Nevertheless, we conducted additional empirical tests with our data. We pooled all data and created age groups that might reflect different developmental stages and that are more or less of equal sample size (15-17 years, 19% of the sample; 18-19 years 30% of the sample, 20-23, 31% of the sample, and over 24 years, 20% of the sample). The Chi square test showed that there was a significant association between moral trait categories and the four age groups ( χ²(75) = 322.28, p < .001), yet the association between trait categories and the four cultural groups was much stronger ( χ²(75) = 766.95, p < .001). We also conducted age group analyses (of more or less equal sample sizes) within each cultural sample and found that they were not significantly associated with moral trait categories in any of the four samples (Filipino [below 17 and over 18 years of age]: ( χ² (22) (20) = 19.86, ns)). These analyses demonstrate that although culture and age are to some extent confounded in our study, age differences between the samples are unlikely to be the ultimate explanation for our findings.
Religion may also be an alternative explanation for the differences we found between the cultural samples. It is noteworthy that it is difficult to delineate religious from cultural duty-based moral codes. Both are based on obligations, duties, obedience and respect for authorities, which may have a social or religious source (Haidt et al., 1993). This is also evident in Haidt's (2008) moral foundation theory in which he merged religious moral concerns (purity/sancity) with community moral concerns (ingroup/loyalty and authority/respect) into the binding foundation of morality.
Although our samples were matched in regard to their level of education (they were all undergraduate students) and are likely not to have any significant income on their own, it might be that the level of income of their parents differs across countries. Previous studies have found that moral judgments vary as a function of socio-economic status (SES; e.g. Haidt et al., 1993). In a similar vein, it might be that parents' SES may have an effect on their children's conceptualization of morality even if they are young adults. Future dyadic studies with students' and parents' responses about moral character may be an intriguing way to explore these relationships.
We took great care in translating the freelisted attributes from the two non-English speaking samples. Yet, alike quantitative research in cross-cultural psychology, there is the issue of equivalence of abstract concepts. For instance, does 'tolerant' in German mean the same as the word in English? The bilingual translators were very well acquainted with both cultural backgrounds and should have been able to detect different nuances in the meaning of abstract concepts between the two cultures. Nevertheless, some culture-specific information will always get lost when Cultural conceptions of morality 71 translating concepts into one common language. It might be desirable to conduct follow up interviews with a second sample of respondents from the respective cultures to clarify the meaning of central concepts and even to ask about normativity, social punishment and context-dependency (Turiel, 1983) in order to further distinguish between common and culture-specific aspects of morality.

Conclusion
There are two main conclusions that we draw from our study. First, if cultural differences are the research focus in morality, imposing constraints through expertdesigned category systems may mask some differences. Moreover, it narrows the scope of inquiry onto the researcher's definition of morality which may reflect the universal functions of morality and obscure cultural conceptions of morality. Adopting a descriptive approach and examining a layperson's perspective provides important insights into everyday morality as conceived in different cultures with the potential to inform theory development, such as the inclusion of a separate moral code defined by the importance of personal rights (e.g., as expressed in gay rights).
Last but not least, an important theoretical contribution of this study is that we provide evidence for the notion that there is an opposite pole to the concept of duties (i.e., the concept of rights) that is not equivalent to the universal justice and welfare concerns as suggested in contemporary theories and research on human morality. Cultures may agree on basic aspects of social cooperation (e.g., not intentionally harming someone with no reason, not to deceive one another), yet there may be large disagreements as to whether the moral order is seen as fixed (pre-defined by a social or religious authority) or malleable (subjected to challenge and social change). This has important implications for the moral discourse between different (sub)cultures and therefore, intergroup and intercultural relations.