Is an Emphasis on Dignity, Honor and Face more an Attribute of Individuals or of Cultural Groups?

This study compares the individual-level and sample-level predictive utility of a measure of the cultural logics of dignity, honor, and face. University students in 29 samples from 24 nations used a simple measure to rate their perceptions of the interpersonal cultural logic characterizing their local culture. The nomological net of these measures was then explored. Key dependent measures included three different facets of independent versus interdependent self-construal, relevant attitudes and values, reported handling of actual interpersonal conflicts, and responses to normative settings. Multilevel analyses revealed both individual- and sample-level effects but the dignity measure showed more individual-level effects, whereas sample-level effects were relatively more important with the face measure. The implications of this contrast are discussed.


Introduction
Attempts to characterize cultural differences over the past several decades have been dominated by varying conceptualizations of individualism and collectivism (Hofstede, 1980;House et al., 2004;Markus & Kitayama, 1991). While the focus on this particular contrast has proved in many ways to be a fruitful one (Smith et al., 2013), evidence has been mounting that the individualism-collectivism contrast may be too broadly defined. The particular focus of this study is on recent studies suggesting that it is possible to contrast two variants of collectivistic cultures, within one of which there is a distinctive emphasis on the preservation of honor, and in the other, on the preservation of face (Leung & Cohen, 2011). In this paper, we introduce a brief measure of the relative emphasis placed on dignity, honor and face in different cultural groups and test the validity of this measure by exploring the nomological net of its predicted correlates. We use multi-level analysis to distinguish the individual and sample-level effects attributable to the prevalence of each of these emphases.

The Three Cultural Logics
The first explicit formulation of cultural contrasts in terms of definitions of dignity, face, and honor was provided by Leung and Cohen (2011). Building on earlier classifications of cultural differences that had used personal values, beliefs and self-construals, Leung and Cohen proposed that groups differ in the basis of their "cultural logics." This perspective is based on the view that cultural differences are not simply embedded within the person but are elicited by the interaction between different kinds of persons and the types of situation that are more frequently encountered in a given cultural context. Thus, individualistic persons would act differently depending on whether they were embedded in an individualistic cultural group or a collectivistic group (Smith & Bond, 2019). Leung and Cohen define the cultural logic of dignity in terms of the inalienable worth of the individual and of the personal accountability of the individual for his or her actions. Where those actions are deemed wrong or inadequate, individuals are likely to experience guilt. They define the cultural logic of honor in terms of the creation and maintenance of respect for the individual as a member of a specific group. Honor can be gained or lost as a consequence of one's actions, and those who lose honor are more likely to experience shame. Leung and Cohen define the logic of face in terms of a distinctive emphasis on hierarchy, humility and harmony. Within this logic, an individual is given face by others on the basis of the extent to which their actions create and preserve in-group harmony, order, and coordination across status lines. Loss of face for oneself or for the other also incurs shame.
The research literature provides abundant studies comparing samples from North America and East Asia on various criteria, and these could be reinterpreted as distinguishing groups in which the dignity of the individual is predominant from those in which the creation and maintenance of face is a stronger priority. Cultural groups in which protection of the honor of one's group is a stronger priority have been much less fully explored until recently. Following the early initiatives of Nisbett and Cohen (1996) and Rodriguez Mosquera et al. (2000Mosquera et al. ( , 2002a, a literature has developed in which the emphasis on honor among samples from the southern US and the Mediterranean region is contrasted with data from the northern US and northern Europe (Cross et al., 2014;Uskul & Cross, 2019).

Studies Contrasting Dignity, Honor, and Face
The first studies built on these definitions of cultural logics were conducted among cultural groups within the US Leung & Cohen, 2011), and contrasted with a sample from Hong Kong . A distinctive attribute of all of these studies is that the authors included few measurements of the cultural logics assumed to be characteristic and operative within each sample. Their designs primarily involved the use of manipulations whose effects were predicted to vary on the basis of the predominant cultural logics within each sample. A similar rationale was used for the design of a comparison of response styles in USA, Turkey, and China as exemplars of dignity, honor, and face . Subsequent studies have each relied on more extensive forms of measurement for ascertaining the cultural logics carried by culture members, and it is the major goal of the present paper to develop and enhance such measures and explore their nomological nets.
While there is no reason to doubt that the variability of cultural logics within the US could enable the Cohen group to sample the utility of these contrasts, it is important to sample more widely in order to test the external validity of the results obtained in other, more varied cultural systems than the American. We note here relevant studies that have sampled from three or more nations.
There is no current consensus as to whether it is preferable to measure cultural differences on the basis of self-report measures (e.g., Schwartz, 2009) or on the basis of respondents' reported perceptions of their cultural norms (e.g., Fischer et al., 2009;Morris et al., 2015). Both types of measures have been employed separately in studies relevant to cultural logics. Mean scores on the measure of self-reported honor values developed by Rodriguez Mosquera et al. (2002b) have been compared across students from eight nations (Guerra et al., 2012), across students from USA, the Netherlands and Turkey (Novin et al., 2015), and across students from six nations (van Osch et al., 2019). Guerra et al. found the measure of person-integrity honor more strongly endorsed in dignity cultures and the measure of family honor more strongly endorsed in their other samples. Novin et al. found that honor values were related to self-esteem in honor cultures but not within their other samples, whereas van Osch et al. found no such relationship. Inconsistencies in such results could arise due to the narrow sampling of cultures in each study and across these studies.
Several researchers have employed a measure of perceived descriptive norms developed by Severance et al. (2013), drawing also on the measures of Rodriguez Mosquera et al. (2002b). Aslani et al. (2016) found that students in the US who were from the US, Qatar, and China scored highest on perceived dignity, honor, and face, respectively. Yao et al. (2017) classified MBA students in the US into cultural groups defined on the basis of respondent nationality as oriented toward either dignity, honor, or face, whereas Smith et al. (2017) surveyed perceived norms reported by students from eight diverse nations. These various results are used to guide our choice of measures.

Measuring Logics
Following Triandis (1995), Leung and Cohen (2011) define cultural logics as syndromes. By doing so, they acknowledge that cultural systems involve an interwoven set of norms, values, worldviews, self-construals, and behaviors, which are organized around a central theme. This highlights the question of how best to identify one or more syndrome elements that adequately represent the central theme. The studies of Smith et al. (2017) and Yao et al. (2017) both used items derived from the work of Severance et al. (2013). These two studies are the only ones known to us in which there has been an attempt to tap all three logics concurrently simultaneously. The items defining the logics of dignity and face focused on injunctive norms that are perceived to be frequently endorsed by "people in one's culture" (Smith et al.,p. 721). The items defining honor focused on a blend of behaviors and values believed to be widespread. Yao et al. found that this measure differentiated adequately between students from world regions thought to exemplify dignity, honor, and face. However, Smith et al. obtained mean scores from specific nations on these cultural logics and found that they did not accord with expectation. In particular, respondents from samples thought to be high on honor and those thought to be high on face did not differentiate their ratings on the honor items from the face items on the scales employed in this study. This lack of distinction may have arisen due to the greater tendency toward acquiescent responding to Likert scale items that is frequently found within these cultural groups (Johnson et al., 2011;Smith, 2004). The possible confound of response style is more thoroughly addressed within the present study.
Thus, there is evidence that measures based either on self-reports or on perceived norms can to some extent differentiate samples defined by the cultural logics of dignity, honor, and face. The measures of perceived norms more fully address all three logics, and we consequently base our measure on perceived cultural logics, taking as the point of departure relevant items from the scales employed by Severance et al. (2013). However, most of the studies that have used explicit measures of dignity, honor, or face have focused only on the comparison of sample means, sometimes also including their relation to other sample-level measures. Thus, the interaction between individual and context that is central to Leung and Cohen's (2011) conceptualization of cultural logics is not addressed. To address this crucial and usually neglected aspect of cross-cultural studies, measurement is required at both individual and sample levels, across a broad range of samples (Bond & van de Vijver, 2011).

Sample-level versus Individual-level analyses.
The relation between individual variability within cultural groups and variation between cultural groups has been increasingly debated in recent years. The monolithic characterization of cultural groups exemplified in the work of Hofstede (1980) has been challenged by the finding that variability between individuals greatly outweighs variability between nations, for instance in terms of values (Fischer & Schwartz, 2011) and of personality (Poortinga & van Hemert, 2001). Multilevel analyses make it possible to estimate sample-level effects where individual-level effects have already been taken into account. Where multilevel analyses have been conducted, it is found that some differences between nations can best be understood as an aggregation of individual-level effects, while other national differences are found to be independent of individuallevel predictors (Becker et al., 2012(Becker et al., , 2014Gheorghiu et al., 2009). In this study, we employ a similar approach in order to unconfound the conceptualization of cultural logics and their associated effects. Are these logics predominantly exemplified at the individual level, at the sample level, or by way of interactions between levels?

Defining a Nomological Net of Measures
In order to identify relevant measures, it is necessary to consider Leung and Cohen's (2011) formulation of the nature of cultural differences. Cultural logics are seen as elicited by particular combinations of circumstances. In contrast to prior dimensional characterizations of cultural difference (Hofstede, 1980;Schwartz, 2009), there is no strong impetus toward characterizing a given group exclusively in terms of its being defined as say, an honor or a face culture. Groups will differ in terms of the relative frequency with which each type of logic is elicited in different social contexts. This formulation falls within the constructivist conceptualization of cultural differences (Hong et al., 2000). From this perspective, it is unhelpful to refer to a particular group or nation as having an honor culture or a face culture. All cultural groups are to some extent dignity cultures and honor cultures and face cultures. Differences are seen as residing primarily in the frequency of the types of circumstance that elicit a specific logic.
Predictions relevant to a particular logic may therefore be tested within any cultural group, as illustrated by the studies of Leung and Cohen (2011). However, the implications of a given cultural logic may differ somewhat, depending on the circumstances in which it is elicited. For instance, elicitation of an honor logic can elicit interpersonal reciprocity in a context where honor is widely endorsed but elicit avoidance where it is not (Leung & Cohen, 2011). It is therefore necessary to examine the interrelation between individual perceptions of a cultural logic and the contexts within which they occur. Within dimensional formulations, a group's culture is seen as a moderator of individual-level effects. The predicted net of associated indices should therefore be more strongly supported in settings where a particular logic is more frequently elicited (Smith & Bond, 2019).
Habitual reliance upon one cultural logic rather than another should be distinctively associated with the ways in which one chooses to describe oneself. Self-construals, attitudes toward others, and descriptions of one's behavior in specific circumstances should therefore all be relevant. In selecting relevant measures, we gave preference where possible to those previously employed in multiple cultural samples. For self-construals, we draw on the findings of Vignoles et al. (2016), who revised and amplified the distinction between independent and interdependent self-construal (Markus & Kitayama, 1991), identifying seven cross-culturally valid facets of independent versus interdependent self-construal on the basis of 55 samples drawn from 33 nations. Since these facets were found to be differentially endorsed in various regions of the world, we can formulate hypotheses that are not simply based on overall independence versus interdependence. Based on analyses reported in this paper, we identified three dimensions of variation in self-construal, viz., Self-Containment versus Connection to Others, Self-Expression versus Harmony, and Consistency across Situations versus Variability, as relevant to our present study.
For attitudes that are relevant to the way that one describes oneself, we use Crocker et al.'s (2003) measure of awareness of others in general. For behavior descriptions, we draw on face-negotiation theory (Ting-Toomey, 1998), which distinguishes the types of concerns that persons may prioritize during the process of negotiation in contexts that are individualistic or collectivistic. This provides separate measures of concern for saving one's own face and for saving the face of the other party to a conflict (Oetzel & Ting-Toomey, 2003). We obtained ratings of the concerns experienced by respondents while handling a recently experienced interpersonal conflict. Face concerns are thus descriptions by respondents of two types of specific behavior in a given context, in contrast to the measure of face logic, which is a perception of the relative priority given to face by those around one.
Contexts in which particular logics are frequently elicited are likely to be characterized by the distinctive norms and behaviors that are apparent to respondents. Rather than sample cultural variation in specific norm adherence (Bond & Smith, 1996;Gelfand et al., 2011), we use newly developed measures of reported discomfort experienced in conforming to norms, and of avoiding situations that are strongly normative. While cultural logics can themselves be considered as subjective norms that prescribe ways of interpreting and reacting to events, our new measures are intended to reflect responses to settings where personal logics are at variance with the prevailing cultural logic in a given setting.
We also use a measure of relational mobility between membership groups (Thomson et al., 2018), which taps the extent to which persons in different cultural groups feel free to join and leave membership groups. As predicted, relational mobility has been shown to be high in individualistic cultures and low in collectivistic cultures, as permeability of group membership is a crucial aspect of a dignity cultural logic. Finally, our survey includes the measure of self-reported honor values (Rodriguez Mosquera et al., 2002b) that was included in preceding cross-national comparisons of honor (Guerra et al., 2012;Novin et al., 2015;van Osch et al., 2019).

Hypotheses
The hypotheses are formulated in terms of the predicted correlates of respondents' perception that a given cultural logic is prevalent in their local context.

Dignity
The cultural logic of dignity is here considered to be built upon the individual's inalienable autonomy and personal accountability for his or her actions. Predicted correlates derive from relevant aspects of the prior literature contrasting individualistic cultures with collectivistic cultures (Smith et al., 2013) and from the findings for self-construal in Western cultures obtained by Vignoles et al. (2016). In particular, members of individualistic cultural groups have been found to be more emotionally expressive (Matsumoto et al., 2008), to find generalized awareness of others aversive (Crocker et al., 2003), to have less concern for other-face than do those in collectivistic cultural groups (Oetzel & Ting-Toomey, 2003), to have greater flexibility in their group memberships, that is, to be high on relational mobility (Thomson et al., 2018), and to vary their behavior less between contexts (English & Chen, 2007). Their lesser concern for conformity (Bond & Smith, 1996) implies less discomfort in the face of normative pressures and greater willingness to express dissent rather than avoid such situations.
Hypothesis 1: Perception of the logic of dignity will correlate (a) positively with self-construals that emphasize self-containment, self-expression, and consistency across contexts; (b) negatively with awareness of others in general; (c) more positively with concern for self-face than other-face; (d) negatively with discomfort and avoidance; (e) positively with perceived relational mobility, but negatively with honor values.

Honor
The cultural logic of honor is here considered to be built upon the individual's personal responsibility to develop and protect the reputation of his or her group and to uphold its honor. Predicted correlates derive from the prior literature concerning honor cultures (Uskul et al., 2018), and from the findings for self-construal in Middle-Eastern cultures obtained by Vignoles et al. (2016). More specifically, members of cultural groups emphasizing honor are concerned both for their own reputation (Guerra et al., 2012) and for the reputation of their family (Uskul et al., 2012). In contrast to a dignity cultural logic, one's reputation as an individual also has implications for the reputation of one's membership groups (especially one's family) in relation to the reputations of outgroups (Uskul et al., 2018). In contrast, face logic is focused much more on the preservation of in-group harmony than with the external reputation of one's group (Yuki, 2003). Thus, within an honor cultural logic, adequate contribution to reputation will be principally a matter of self-face. Within the honor logic, a person will be aware of the reactions of others within their group, but that awareness will be focused on the need to fulfil one's obligations to the group, rather than on giving face to other group members. Adherence to an honor code involves being consistent and reliable in upholding and defending the honor of one's group (Uskul et al., 2018). In their study of self-construals, Vignoles et al. (2016) found that respondents from Middle Eastern samples scored higher on self-reliance and consistency than those from six other world regions. Members of honor groups are also found to be more polite than those in dignity cultures, in order to guard against threats to reputation that may elicit the need for retaliation (Cross et al., 2013). However, they react more forcefully when avoidance is not an option. This implies a preference for avoidance of potentially conflictual situations.
Hypothesis 2: Perception of the logic of honor will correlate (a) positively with self-construals that emphasize self-consistency but negatively with self-construals that emphasize self-containment; (b) positively with awareness of others in general; (c) positively with concern for self-face; (d) positively with avoidance; (e) positively with honor values, but negatively with perceived relational mobility.

Face
The cultural logic of face is here considered to be built upon the individual's responsibility to contribute to the preservation of harmony, humility, and hierarchy within their group (Leung & Cohen, 2011). Predicted correlates derive from the prior literature contrasting collectivistic cultures with individualistic cultures (Smith et al., 2013), and from the findings for self-construal in East Asian cultures obtained by Vignoles et al. (2016). This study found that East Asian samples scored notably low on expressiveness and consistency. The four samples from China, Malaysia, and Thailand also scored low on self-containment. However, those from Japan and Singapore scored high this aspect of self-construal. The prediction for this specific effect is therefore tentative. In other studies, members of East Asian cultural groups have been found to have greater concern for other-face than those in individualistic cultural groups (Oetzel & Ting-Toomey, 2003), to be low on relational mobility (Thomson et al., 2018), and to vary their behavior more between contexts (English & Chen, 2007). Higher conformity (Bond & Smith, 1996) could imply greater discomfort and avoidance, but only in contexts where the individual dissents from the norms, here then, no prediction is entered.
Hypothesis 3: Perception of the logic of face will correlate (a) negatively with self-construals that emphasize self-containment, expressiveness, and consistency across contexts; (b) positively with awareness of others in general; (c) positively with both self-face and other-face; (d) negatively with perceived relational mobility.

Participants
Participants were 5,064 students from 24 nations who completed the survey either online, in the classroom, or, in the Mexico City sample, by response to a request in public spaces. They either received course credit or were thanked for their participation. Ethical consent for the research project was obtained from each university that was sampled. In a small minority of cases, this was based upon the ethical scrutiny that had been conducted at the University of Sussex. Respondents provided details of their age, gender, country of birth, nationality, ethnicity, and religion, and also rated the location of their upbringing on a 7-point scale from rural (1) to urban (7). The survey was originally constructed in English and was then translated into the language for use at each location by first-language-speaking authors and their collaborators, with subsequent independent back-translation and correction based on discussion (van de Vijver & Leung, 1997). Respondents who were not nationals of the location sampled were excluded from the data analysis. Details of samples are provided in Table 1.

Measures
Measurement of cultural logics. For the present study, we obtained ratings of the cultural logics that were perceived to be prevalent in the respondent's local context. We chose to use three single items, with acquiescent responding controlled by standardizing each score against the mean for all three ratings. Single items clearly do not tap all aspects of a given logic, but they do make it clear which specific aspect is being tapped. Respondents were asked to rate how well each statement described ". . .the people around you (your school, workplace, town, neighbourhood, etc.)." The items used were newly developed: "These people think that they should be true to themselves regardless of what others think" (Dignity); "These people feel that they should uphold and defend their family's reputation" (Honor); "These people think they should be extremely careful not to embarrass others" (Face). The 6-point response scales were keyed from strongly disagree to strongly agree.
A test for cultural differences in response style showed that sample mean raw scores for all three cultural logics combined varied between 3.43 and 4.90 (SD = 0.93) on the 6-point scale. Standardization of scores for each logic was therefore essential. Following the procedures employed by Kashima et al. (1995), variations in response style between individuals was first discounted through within-subject standardization, yielding the individual-level   cultural logic scores. Differences in response style between samples were then discounted by standardization across samples. These standardized scores were then averaged at the sample-level to create the sample-level scores. The use of just three items as the basis for these measures does not permit tests of measurement equivalence.
Self-construal. Scales measuring three dimensions of self-construal were adapted from the larger number of items used by Vignoles et al. (2016). Each scale comprises six items, with some items phrased in terms of independence and other items phrased in terms of interdependence. Respondents were asked: "How well does each of these statements describe you?" 9-point response scales were used, ranging from 1 (not at all) to 9 (exactly), with three intermediate anchor-points (3 = a little, 5 = moderately, 7 = very well). Items were worded using "you," in order to make the task "easier" for all cultural groups. Scales measured: Self-Containment versus Connection to Others ("If a close friend or family member is sad, you feel the sadness as if it were your own"); Self-Expression versus Harmony ("You prefer to preserve harmony in your relationships, even if this means not expressing your true feelings"); and Consistency across Situations versus Variability ("You behave in the same way even when you are with different people"). Smith et al. (2020) report analyses of this data. Following Kashima et al. (1995), individual variations in response style were first discounted through within-subject standardization. Differences between samples in response style were then discounted by standardization across samples. A pan-cultural exploratory factor analysis with oblimin rotation explained 40.9 percent of variance and yielded three factors defining the three facets of self-construal. The factor scores for Harmony versus Expressiveness were reversed, to ensure that all scores were keyed toward independence rather than interdependence. To test the robustness of these scores, Tucker's phi (van de Vijver & Leung, 1997) was computed for each of ten clusters of samples that were judged culturally similar or geographically adjacent. The items defining the factor structure within each cluster were compared in turn with the pan-cultural structure. All factor coefficients exceeded .90 and 157 of 160 values for item congruence were at or above .90. Thus, there is evidence for a satisfactory structure for the three self-construal scales. Fuller detail of these analyses is provided by Smith et al. (2020).
Attitudes and values. The measure of Awareness of Others in General (Crocker et al., 2003) comprised five items, three of which are reversed (e.g., "I don't care what others think of me"). Responses were made on 4-point Likert scales, with anchors from strongly disagree to strongly agree. The measure of Honor values comprised five items provided by Rodriguez Mosquera et al. (2008) (e.g., "It is important to you that others see you as someone who deserves respect"). Responses to these items were based on 9-point scales, using the same anchors as described above for the self-construal items.
Face concerns. Respondents were asked to think of a recent interpersonal conflict that they had experienced, using a measure designed by Oetzel and Ting-Toomey (2003). They first identified the gender of the other party and indicated whether the conflict involved their romantic partner, a family member, a friend, or someone from work or college. They then rated four items referring to Self-Face (e.g. "I was concerned with protecting my selfimage"), and six items referring to Other-Face (e.g. "I tried to be sensitive to the other person's self-worth"). Responses were recorded on 7-point Likert scales, with anchors from strongly disagree to strongly agree. Lun et al. (2020) report analyses of these data. After within-sample standardization, factor analysis of the 10 face items for the total sample yielded a two-factor solution with oblimin rotation explaining 56.5 percent of variance. To test the adequacy of the two face scales within the samples, they computed Tucker-Lewis phi coefficients (van de Vijver & Leung, 1997) comparing each of ten culturally similar or geographically adjacent clusters of samples with the pan-cultural factor structure. For nine of the ten clusters, all 20 congruence coefficients for factors exceeded .96. For the Southeast Asian cluster, comprising the Malay and Thai data, the coefficients were .89 for Other-Face and .69 for Self-Face. Thus, there is overall evidence for consistent structure of the face concern scales. Relational mobility. Respondents were also asked to rate whether a series of statements accurately describe the people in the immediate society where they live. Perceived relational mobility was measured with 12 items devised by Masaki Yuki (Thomson et al., 2018) (e.g., "It is common for these people to have a conversation with someone they have never met before"). The 6-point response scales are keyed from strongly disagree to strongly agree. Using the same items as in the present study, Thomson et al. (2018) reported acceptable partial scalar invariance of the latent variable for relational mobility across their samples from 39 nations, after relaxing equality constraints for seven intercepts. Within the present data, sample-level Cronbach alpha was .86, and average individual-level alpha across samples was .78.
Reaction to norms. Nine items were newly created for the present study.
Respondents were asked to think of times when they had been in a situation in which certain behaviors were appropriate or expected, and how they reacted to such circumstances. Six items ("Discomfort") refer to discomfort in relation to perceived social expectations (example item: "In some situations, you are expected to behave in ways that would make you feel uncomfortable"). Three items ("Avoidance") refer to avoidance of perceived social expectations ("You usually avoid situations in which it is appropriate to behave in ways that would make you feel awkward"). Response scales for these items were the same as those used for the self-construal items. Smith et al. (2020) confirmed the intended factor structure of these items through multilevel confirmatory factor analysis, using MPlus Version 8 (Muthén & Muthén, 1998. They specified two covarying factors at both participant and sample levels, one representing a factor indicated by the six discomfort items, and one indicated by the three avoidance items. All items were highly significant indicators of their respective factors (ps < .001), and the model was an acceptable fit to the data (CFI = .93, RMSEA = .08, SRMR within = .046, SRMR between = .15). We tested our hypotheses through hierarchical linear modeling using MPlus Version 8 (Muthén & Muthén, 1998. We predicted the various dependent variables from the dignity, honor, and face mean scores, both at the individual and the sample level. As the means for the three measures of cultural logics had been standardized relative to the overall mean, they are not independent of one another. A separate set of analyses was therefore run for each logic. We also report estimates of random slopes, indicating the extent to which individual-level effects vary in consistent ways between samples. In evaluating random slope effects, we follow Nezlek (2011, p. 327), who notes that moderation may be stronger or weaker in different samples and that it is therefore possible to test for cross-level interactions even when the random slope term is not significant. Cross-level interaction terms indicate whether a given individual-level effect is strengthened or weakened in samples characterized by a given logic.

Results
The sample-level means shown in Table 3 indicate that the cultural logic measures have strong plausubility, with 22 of 29 samples showing highest mean scores for the logic that might be expected on the basis of the prior literature. At the sample-level, the standardized scores for face logic correlated with dignity logic at −.72 (p < .001) and with honor logic at −.15 (ns). Dignity logic correlated with honor logic at −58 (p < .001).
Preliminary multivariate analyses indicated that individual-level cultural logics were not significantly related to age, gender, or urban origin. Respondents on paper rather than online scored significantly higher on honor (p < .001) and on face (p < .05). However, after controlling for gender and age differences, there were no significant differences between paper and online responders in perceived culture means within each of the three samples that contained both types of respondent. The effect of response mode is therefore not controlled in the main analyses, as the choice by the authors of this paper as to the local appropriateness of paper or online response is more likely to be a side effect of the cultural and economic differences between samples. Online responders were from nations with average purchasing power parity of $37,767, whereas paper responders were from nations with average purchasing power parity of $26,782 (www.imf.org/external/pubs/ ft/weo/2018/01). Past research on cultural differences in values has also found no notable difference in data between online and pencil and paper data collection modes (Lilleoja et al., 2016). Table 4 shows the results of tests of Hypotheses 1-3. There are numerous individual-level and sample-level significant relations between the dependent measures and each of the cultural logics. We consider these results at each level of analysis in turn. Hypothesis 1. Hypothesis 1 concerns associations with the logic of dignity. We note the effects found in relation to the hypotheses. At the individual level, there are significant predicted associations with self-containment (H1a), low awareness of others (H1b), low avoidance (H1d), high relational mobility (H1e) and low honor values (H1e). Both types of face concern were low, but not in the manner predicted. Thus, there was evidence for five of nine   predicted effects. Samples with strong dignity logic show predicted effects for high consistency (H1a), lower other-face concern (H1c), low honor values (H1e) and high relational mobility (H1e), yielding four of nine predicted effects. There are also four significant cross-level effects. The association between individual-level perceptions of dignity and expressiveness, discomfort, and avoidance are all negative in samples where respondents perceived a dignity logic is widely endorsed, whereas the association between perceived dignity and relational mobility is more strongly positive in such samples.

Hypothesis Tests
Hypothesis 2. Hypothesis 2 concerns associations with the logic of honor. At the individual level there are significant predicted effects for low self-containment (H2a), high awareness of others (H2b), high self-face concern (H2c), high honor values (H2e), and a trend toward high avoidance (H2d), yielding five of seven predicted effects. Samples with strong honor logic show predicted effects for low self-containment (H2a), high honor values (H2e), and a trend toward avoidance (H2d), yielding three of seven predicted effects. However, there is also a significant negative relation with awareness of others, in contrast with the effect found at the individual level. There are no cross-level effects.
Hypothesis 3. Hypothesis 3 concerns association with the logic of face. At the individual level there are significant predicted effects for high other-face concern (H3c), and low relational mobility (H3e), yielding two of seven predicted effects. Samples with strong face logic show predicted effects for low expressiveness (H3a), high self-face concern (H3c), high other-face concern (H3c), and low relational mobility (H3d), yielding four of seven predicted effects. There are two significant cross-level effects. Within samples where face logic is widely perceived, the individual-level association between face logic and avoidance is negative, and the association between face logic and relational mobility is weakly positive.

Discussion
This study has explored the correlates of a set of three measures focused on the cultural logics of dignity, honor, or face. We first required evidence as to the validity of the simple three-item measure that was used to differentiate logics. If this separation can be established, it becomes possible to explore our interest in contrasts between individual and sample-level effects.

Measurement of Cultural Logics
The utility of our measures of cultural logics can be evaluated by the plausibility of the sample mean scores shown in Table 3, and by considering the outcome of hypothesis tests concerning the nomological net of each measure. The sample means did accord with prior expectation in 22 of 29 instances. Our hypothesis tests revealed 12 significant individual-level effects in the predicted direction, from the 23 for which a prediction was entered. There were also 11 significant predicted sample-level effects from the 23 that were computed. Considering both sets of effects together, in six instances predicted effects were found at both levels, in four instances they were at the individual level only and in three instances they were at the sample level only. In the seven of the remaining instances, there were no main effects. Six significant cross-level effects were also found, three of them in the absence of main effects. Thus, for the 23 predicted relationships, some support was found for 16 of them. Given these effects it appears that the three-item measure does capture elements of the contrast between these three cultural logics. Summarizing effects in this way risks overemphasizing the coherence of what has been found, unless a basis can be specified for when effects would be found at a given level. It is evident that there were slightly more individuallevel effects than sample-level effects. Furthermore, most sample-level effects provided evidence of the strengthening of individual-level effects in circumstances where the logic perceived by individuals was in fact widely endorsed by others in the same sample. The cross-level effects show a different pattern, some strengthening and others weakening individual-level effects where a given logic is strongly endorsed. These results enrich our understanding of cultural logics by showing the mutual relevance of individual-and sample-level data -these effects may or may not be compatible, as Leung and Bond (2007) maintain. Indeed, some of our results do show that logics are differently associated with various outcomes depending on the prevalent logic. For instance, individuals' honor logic was positively associated with generalized awareness of others, but not in samples where honor logic was widely endorsed.
The pattern of effects also appears to vary between logics. In relation to dignity logic, seven individual-level effects are found, compared to four sample-level effects. For honor logic there are four individual-level effects and three sample-level effects. For face logic, there are two individual-level effects and four sample-level effects. This set of results suggests that individuals' dignity logic may be somewhat more strongly predictive of effects, whereas face logic may be more rooted in the cultural context. Given that dignity logic is rooted within the individual, it is plausible that the effects associated with it will be most readily predictable from the individual's reading of the logic prevalent in his or her immediate social context. Conversely, given the contextualized nature of face logic, it is plausible that the effects associated with it would be more predictable from the properties of the broader social system in which face logic is frequently elicited. This contrast in results requires fuller investigation, when measures of cultural logics based on larger numbers of items are used. Since Leung and Cohen's (2011) conceptualization of logics emphasizes that each logic is elicited by distinctive types of context, the imbalance of individual and sample-level effects underlines the importance of examining more closely the eco-social and interpersonal contexts in which particular logics are most frequently elicited.

Cross-level Efects
Only six cross-level effects were found, and four of these refer to the same two predictors, namely avoidance and relational mobility. This provides only modest support for Leung and Cohen's (2011) proposition that differing cultural logics can modify the meaning of a given event. However, the measures of logics that we used were already based on ratings of one's immediately perceived context and may therefore tap context as well as person. Distinguishing persons and context as wholly separate is problematic (Bond, 2013). Although engaging 29 samples does greatly extend the number used in past studies, more extensive detection of cross-level effects might require still broader sampling.

The Distinction Between Honor and Face
Cultural groups characterized by emphasis on honor and face have until recently both been seen as exemplars of collectivism. The present results provide evidence for their distinctiveness, at both levels of analysis. The measure of relational mobility distinguished face logic from dignity logic, but was unrelated to honor logic. Thomson et al. (2018) found relational mobility low in both face cultures and Arab honor cultures, but they sampled adults, whereas we sampled students, who are likely to be more mobile. The measures of honor values distinguished honor logic from dignity, but were unrelated to face. The strength of these contrasting findings for honor logic and for face logic is underlined by the fact that they are built upon all four of the significant random slopes that were found in the entire analysis.
The associations between cultural logics and the ratings of past conflicts also provided clear contrasts between all three logics, perhaps because they were based on ratings of actual past behaviors. Our predictions for the handling of past conflicts were based upon prior studies comparing the relative salience of concerns for self-face and other-face (Oetzel & Ting-Toomey, 2003). The present results suggest an alternative perspective, with both types of face being salient where face logic is prevalent and neither type of face salient where dignity logic prevails.

Limitations
The principal limitation of this study is the use of three single-item measures. Multiple items would provide greater assurance that the three cultural logics have been adequately represented. However, we have presented evidence that the measures employed have sufficient value to take benefit from the broad range of samples that were surveyed. A second limitation is the variation between procedures used to collect data at different sites. At the three sites where data were collected both online and on paper, no mean differences in cultural logics were found. While this difference will have had some effect on the mean scores attributed to sites, it is not likely to have affected the probability of obtaining individual and sample-level effects.
None of the measures included in this study has been shown to have full metric equivalence across samples. This is not unusual when large numbers of samples from different nations are included. Tests for full scale equivalence are not achieved even with measures widely considered to be well established, such as the Big Five personality dimensions (Marsh et al., 2010). The very diversity of samples included may preclude conventional criteria for scale equivalence. In such circumstances, there is an argument for tradeoffs between broad sampling and measurement equivalence. In establishing the level of adequacy of the measures that have been employed in this study, an alternative basis for evaluation has been provided by examining the nomological network for the cultural logics of dignity, face, and honor.

Conclusion
All prior studies of cultural logics of which we are aware have either analyzed data at the individual level, or compared sample-level means. The use of hierarchical linear modeling in obtaining our results has made it apparent that the sample-level and cross-level effects that are found are explaining variance that is additional to that explained by individual-level effects. These results support the contention of Leung and Cohen (2011) that the implications of individual-and sample-level logics are mutually relevant to one another. To understand the actions of individuals, we need to know both the logics with which they interpret their context and the logics that are widely employed within that context. Those perceiving each logic to be most distinctively enacted around them do describe themselves and their actions in ways that are predominantly consistent with the logics of dignity, honor, and face. Cultural differences in logics are attributable both to individuals' distinctive perceptions of their context and to the contextual press of others with similar perceptions. Individual perceptions may be more relevant to the dignity logic, while contextual press may be more important in relation to face logic.

Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The work of Maria Efremova was supported by the framework of the HSE Basic research Program.
Matthew J. Easterbrook is a senior lecturer in psychology at the University of Sussex, UK. His research focuses on identity and group processes in different social and cultural contexts.
Yasin Koc is an assistant professor of social psychology at the University of Groningen, Netherlands. His research focuses on the integration of multiple conflicting identities, disadvantaged groups, and social change.
Vivian Miu-Chi Lun is an associate professor at Lingnan University in Hong Kong. Her research interests include critical thinking and well-being, and how these variables relate to culture.
Dona Papastylianou is a Professor in Social Psychology at the National and Kapodistrian University of Athens. Her research interests are focused on Applied Social Psychology and Cross-Cultural Psychology, and specifically in cross-cultural adaptation of migrants and remigrants, migrants and remigrants, multicultural and psychosocial issues in the community context, social values, social justice and social advocacy issues, emotions and group processes.
Lusine Grigoryan is a postdoctoral research fellow at Ruhr University Bochum, Germany. Her research interests include prejudice, social identity, values, and morality.
Cláudio Torres is a professor in the Institute of Psychology of the University of Brasilia, Brazil, Department of Basic Psychological Processes, and a senior researcher in the Graduate Program in Behavioral Sciences in the same university. He participated in research projects on indigenous approaches to achieve informal influence, validity tests of refined values theory, political consumerism, and core political values, among others. He teaches graduate courses on cross-cultural social psychology, consumer behavior, and cultural diversity and inclusion.

Maria Efremova is a Leading Research Fellow in the Center for Socio-Cultural
Research and associate professor at the Department of Psychology of National Research University Higher School of Economics (Russia). She mainly studies economic disadvantage, poverty attributions, social identity, and group-based guilt and shame.
Bushra Hassan is assistant professor at International Islamic University Islamabad (Pakistan). She received her PhD from the University of Sussex (United Kingdom). Her work focuses on culture, personal and social aspects of identity and well-being.
Ammar Abbas is a professor in the Department of Chemical Engineering, College of Engineering at the University of Baghdad. Beside his researches in chemical engineering, he studies a wide variety of psychological phenomena, including social cognition, motivation and cross-cultural psychologies.
Abd Halim Ahmad is an associate professor in cross-cultural social psychology at Universiti Utara Malaysia (Malaysia). He received his MSc in Occupational diagnosis of educational environments, relationships between social axioms, and personal autonomy in educational organizations.
Pelin Gul is a lecturer in psychology at the University of Twente, Netherlands. Her research focuses on identifying psychological motives and ecological predictors underlying gender relations, sexism, gender-based biases and honor norms.
Ceren Gunsoy is an assistant professor in psychology at Clemson University (USA). She completed her Ph.D. in social and cross-cultural psychology at Iowa State University. Her research examines cultural differences in interpersonal conflict, person perception, emotions, and social media behavior.
Anna Hakobjanyan is a lecturer at Yerevan State University, Armenia. She also works in the Personality and Social Context Laboratory on cross-cultural research methods and is involved in cross-cultural research collaboration.
Siugmin Lay is a researcher at the Measurement Center MIDE UC at Pontificia Universidad Católica de Chile. Her research focuses mainly on intergroup relations, in particular intergroup contact, prejudice, acculturation, social norms, and prosocial behaviours.
Olga Lopukhova is associate professor at the Institute of Psychology and Education, Kazan Federal University (Russia). She specializes in psychometrics, survey questionnaire design and adaptation to Russian culture. Her research interests are focused on the psychology of gender transformations, factors of psychological well-being, cultural and environmental psychology in personality development.
Ping Hu is a professor in social psychology at Renmin University of China (Beijing, P.R.China). She studies a wide variety of cultural psychological phenomena, including cross-cultural difference in thinking mind, emotional contagion and cross-cultural communication.
Diane Sunar is emeritus professor of psychology at Istanbul Bilgi University. Her interests include cultural changes in child-rearing practices; early childhood intervention programs; moral development; and the cross-cultural application of relational models theory to the study of moral psychology.

Maria Luisa Mendes
Teixeira is a full professor in organizational studies at Presbyterian Mackenzie University and president of the International Association for Transcultural and Organizational Studies. She studies organizational dignity and is interested in studies on cultural logics. Doriana Tripodi is a clinical psychologist and a cognitive behavioral psychotherapist in her private practice at Gioia Tauro, Calabria, Italy. In recent years she has collaborated with the Faculty of Psychology of the University of Padua in scientific research about learning disorders and with the University of Catanzaro in research into the personality of priests.