The HEXACO Model of Personality Structure

(1)

The HEXACO Model of Personality Structure

Michael C. Ashton and Kibeom Lee

By the 1990s, many personality researchers had reached a consensus that the domain of personality variation was best summarized in terms of five broad and roughly independent dimensions known as the Big Five factors. These dimensions were discovered in investigations of English-language personality-descriptive adjectives, and were later popularized via the Five-Factor Model of personality structure and its associated questionnaire instruments.

In recent years, however, there has emerged much new evidence suggesting an alternative representation of personality structure. This newer framework, which we have called the HEXACO model, consists of six rather than five dimensions. Three of the six HEXACO factors have close counterparts in the Big Five or Five-Factor Model (B5/FFM), whereas the other three HEXACO factors have more complex relations with the remaining two dimensions of the B5/FFM.

In this chapter, we will begin by describing the origins and content of the HEXACO dimensions, which were discovered in analyses of the personality lexicons of various languages. After summarizing the results of those

studies and the content of the six HEXACO dimensions, we then discuss the theoretical bases of these factors, exploring the likely adaptive trade-offs between high and low levels of each dimension. We then turn to the predictive validity of the HEXACO constructs in accommodating a variety of personality characteristics, particularly those that may be only modestly associated with the B5/FFM dimensions.¹ Next, we discuss the differences between the HEXACO model and some other recently proposed frame- works, as judged both by interpretability of factors as personality constructs and also by the replicability of the structures. Finally, we discuss the results of some recent lexical studies of personality structure conducted in the Croatian, Greek, and Filipino languages.

EMPIRICAL BASIS: RESULTS OF LEXICAL STUDIES OF PERSONALITY STRUCTURE IN MANY LANGUAGES

The statistical technique of factor analysis is useful for identifying the major dimensions

12

(2)

that underlie the domain of personality characteristics. But if factor analysis is to gener- ate an accurate model of personality structure, it is necessary to analyze variable sets that are representative of that domain. Such variable sets can be obtained through the lexical strategy, whereby the familiar personality- descriptive words (usually adjectives) of a language are selected as the variables on which participants’ self- or peer ratings can be obtained and then factor-analyzed. The important strength of the lexical approach is that the selected variables constitute the full array of subjectively important personality characteristics that have been observed and described by generations of speakers of a language. Thus, this approach greatly reduces the problem of researcher biases in the selection of personality variables (see Ashton and Lee, 2008a, for responses to various criti- cisms of the lexical approach).

EARLY ENGLISH-LANGUAGE LEXICAL STUDIES: ORIGINS OF THE BIG FIVE AND FIVE-FACTOR MODEL

The first lexically based investigations of personality structure were conducted in the English language. In those early studies (e.g. Cattell, 1947), limited computing power forced researchers to use rather small variable sets (typically less than 50 variables) that represented only a small fraction of the English personality lexicon. Those investigations consistently produced a common set of five factors (see Tupes and Christal, 1961,1992) now called the ‘Big Five’ (e.g. Goldberg, 1990).

The names of these Big Five factors are extra- version (e.g. talkative, outgoing versus quiet, shy), agreeableness (e.g. gentle, sympathetic versus harsh, cold-hearted), conscientiousness (e.g. organized, disciplined versus sloppy, lazy), emotional stability (e.g. relaxed versus moody, anxious), and intellect/imagination (e.g. intellectual, imaginative versus shallow).

The Big Five factors have been popularized through the closely related Five-Factor Model,

whose dimensions are operationalized by the the NEO Personality Inventory – Revised and the NEO Five-Factor Inventory (NEO- PI-R and NEO-FFI; Costa and McCrae, 1992). In these instruments, the five factors are conceptualized in ways that depart slightly from the original Big Five identities, but remain very similar overall to the Big Five structure. (One noteworthy difference is that the five-factor model openness to experi- ence factor excludes the intellectual ability aspects of the corresponding Big Five factor of intellect/imagination.) Note that the Five- Factor Model does owe its ultimate origins to lexical studies of personality structure (see McCrae, 1989). Three of its dimensions – neuroticism (i.e. low emotional stability), extraversion, and openness to experience – were identified in analyses of Cattell’s personality scales, which had been derived from his earlier lexical research (Cattell, 1947). Its other two dimensions – agreeableness and conscientiousness – were later added in response to findings from lexical studies of personality structure in the English language.

During the past decade, the factor structure of the NEO-PI-R has been examined in countries around the world. The results show that the five-factor space consistently corresponds closely to that obtained in samples from the US. But although this result supports the validity of the NEO-PI-R as an operational- ization of the B5/FFM, it does not provide independent evidence that the B5/FFM is the optimal cross-culturally replicated representation of personality structure (cf. McCrae and Costa, 1997, 2003). Such evidence can only be forthcoming from analyses of variable sets that are indigenous to the cultures in question (not imported from the culture of the model’s developers) and representative of the personality domain (not selected as mark- ers of a specified set of factor axes).

If the B5/FFM did, in fact, represent the optimal structural model of personality variation, then one would not expect to observe any widespread recovery of a larger set of factors from analyses of personality variable sets that are culturally indigenous and representative

(3)

of the personality domain. Some proponents of the Five-Factor Model have suggested that there exist five and only five factors of per- sonality, a point that is viewed as ‘an empirical fact, like the fact that there are seven continents or eight American presidents from Virginia’ (McCrae and John, 1992: 194).

From the perspective of the B5/FFM, the pre- dicted result of investigations of personality structure is that the familiar five factors – but no more than five factors – should be replicated when the personality lexicons of various languages are analyzed.

LEXICAL STUDIES IN MANY LANGUAGES: EMERGENCE OF THE SIX-DIMENSIONAL (HEXACO) FRAMEWORK

Since the 1980s, lexical studies of personality structure have been conducted in various languages other than English, using adjective selection criteria that have generally followed the logic of the lexical approach by excluding terms that are not plausible descriptors of personality (e.g. terms that are chiefly evaluative or that describe physical characteristics and abilities; see Ashton and Lee, 2001, 2002). In most languages, the five-factor solutions have recovered the space of the Big Five factors as found in English, although often with some differences in the rotational positions of the factor axes. In a few languages, however, five- factor solutions have not recovered the Big Five factor space. Most notably, an intellect/

imagination dimension has failed to emerge among the first five factors in lexical studies conducted in Italian (Di Blas and Forzi, 1998), in Hungarian (Szirmak and De Raad, 1994), and in Greek (Saucier et al., 2005; see also reanalysis by Lee and Ashton, 2006a).

The most striking findings from these investigations, however, have been the results obtained in six-factor solutions. Contrary to the suggestion that there are five and only five replicable dimensions of personality,

a common set of six factors has emerged across at least 12 languages (see list in Table 12.1).

As reported by Ashton et al. (2004), the personality lexicons of the Dutch, French, German, Hungarian, Italian, Korean, and Polish languages produced very similar six- factor solutions. A similar six-dimensional structure has also been recovered more recently in reanalyses of archival data based on the English personality lexicon (Ashton et al., 2004a), and also in reanalyses of investigations of the personality lexicons of the Croatian (Lee et al., 2005a), Greek (Lee and Ashton, 2008), Filipino (Ashton et al., 2006a), and Turkish (Wasti et al., in press) languages. (Later in this chapter, we provide a detailed description of the Croatian, Greek, and Filipino results.)

Listed in Table 12.2 (in English translation) are the adjectives that typically define these six factors as observed in the lexical studies conducted in various languages. Two of the factors in that table are very similar in content to the extraversion and conscientiousness dimensions obtained in five-factor solutions of previous English-language investigations. A third dimension of that table is similar to the English intellect/imagination dimension, with an additional promi- nent element of unconventionality.

Table 12.1 Languages having known six- factor solutions from lexical studies of personality structure (personality- descriptive adjectives only), classified by language family and (for Indo-European family) by branch of language family Language Family (Branch) Languages Indo-European (Italic/Romance) French, Italian Indo-European (Germanic) Dutch, English, German Indo-European (Hellenic) Greek

Indo-European (Slavic) Croatian, Polish

Korean Korean*

Malayo-Polynesian Filipino

Turkic Turkish*

Uralic Hungarian

Note: See reviews of lexical study results in Ashton, Lee, Perugini, et al. (2004), Ashton and Lee (in press).

* Some linguists (e.g. Ruhlen, 1991) classify Korean and Turkish within a broad Altaic language family

(4)

Table 12.2Summary of defining content and theoretical interpretations of HEXACO factors Common defining adjectives in HEXACO-PI-R facet Theoretical Factorlexical studiesscalesinterpretationBenefits of high level?Costs of high level? Honesty-humilitySincere,honest,faithful/loyal,modest/Sincerity,fairness,greed-Reciprocal altruismGains from cooperation (mutualLoss of potential gains that unassuming,fair-minded versus sly,avoidance,modesty(fairness)help and non-aggression)would result from exploitation greedy,pretentious,hypocritical,of others boastful,pompous Agreeableness Patient,tolerant,peaceful,mild,Forgiveness,gentleness,Reciprocal altruismGains from cooperation (mutualLosses due to being exploited by (versers anger)agreeable,lenient,gentle versus flexibility,patience(tolerance)help and non-aggression)others ill-tempered,quarrelsome,stubborn, choleric EmotionalityEmotional,oversensitive,sentimental,Fearfulness,anxiety,Kin altruismSurvival of kin (especiallyLoss of potential gains associated fearful,anxious,vulnerable versusdependence,offspring);personal survivalwith risks to self and kin brave,tough,independent,sentimentality(especially as favors kin self-assured,stablesurvival) ExtraversionOutgoing,lively,extraverted,sociable,Social self-esteem,social Engagement in social Social gains (i.e.friends,mates,Energy and time;risks from social talkative,cheerful,active versus shy,boldness,sociability,endeavorsallies)environment passive,withdrawn,introverted,liveliness quiet,reserved ConscientiousnessOrganized,disciplined,diligent,careful,Organization,diligence,Engagement in task-Material gains (i.e.,improved useEnergy and time thorough,precise versus sloppy,perfectionism,prudencerelated endeavorsof resources),reduced risks negligent,reckless,lazy,irresponsible, absent-minded Openness to Intellectual,creative,unconventional,Aesthetic appreciation,Engagement in idea-Materials and social gains (i.e.Energy and time;risks from social Experienceinnovative,ironic versus shallow,inquisitiveness,creativity,related endeavorsresulting from discovery)and natural environment unimaginative,conventionalunconventionality

(5)

Two other factors of Table 12.2 have more complex relations with the classic English lexical Big Five factors. One factor has elements in common with (low) Big Five emotional stability; however, this factor excludes the anger that mainly defines (low) emotional stability, and includes the sentimentality that mainly defines Big Five agreeableness (cf. Saucier and Goldberg, 1996, table 2). We have called this Table 12.2 factor ‘emotionality’, a less pejorative term than ‘emotional instability’ or ‘neuroticism’ (Ashton et al., 2004b). Conversely, another factor of Table 12.2 resembles Big Five agreeableness to some extent, but excludes sentimentality and includes (lack of) anger. This factor of Table 12.2, being defined by patience, gentleness, and flexibility, is perhaps better summarized by the name ‘agreeableness’ than is the B5/FFM version, which is defined by sympathy, gentleness, and sentimentality; to differentiate the two versions of the factor, we sometimes refer to the six-dimensional variant as ‘agreeableness (versus anger)’. Given these shifts of factor content, emotionality, and agreeableness (versus anger) can be seen roughly as rotational variants of B5/FFM (low) emotional stability and B5/FFM agreeableness.

We have labeled the remaining factor in Table 12.2 as ‘honesty-humility’ (e.g. Ashton et al., 2004b). As seen in the table, the terms that frequently define this factor suggest hon- esty (e.g. sincere, fair) and humility (e.g. at the low pole, pretentious, greedy). Thus, the emer- gence of these three dimensions – emotionality, agreeableness (versus anger), and honesty- humility – is the feature that distinguishes the cross-language six-factor structure from the B5/FFM.

The content of the six lexical personality factors, as shown in Table 12.2, tends to be very similar across studies and across languages. Nevertheless, there is some variation across investigations in the rotational positions of the factor axes that emerge from the use of simple-structure-seeking algorithms (e.g. varimax). As a result, the core content of the six factors does tend to show some shifts between investigations. An example of this

involves the positions of some emotionality- related terms, such as those describing fearfulness and sentimentality, within the plane of emotionality and extraversion. From one study to the next, there may be small variations in the positions of those two vectors, with the result that fearfulness terms may show substantial negative loadings on extraversion, or that sentimentality terms may show substantial positive loadings on extraversion. Despite variations such as these, however, the factor spaces observed in the various studies remain very similar. Moreover, the factors that typically emerge in simple- structure solutions from lexical studies conducted in various languages are those that are summarized in Table 12.2.

The six-factor structure is also characterized by an interesting variability of the locations of a group of related terms describing sympathy, soft-heartedness, generosity, and associated characteristics. These terms repre- sent blends of factors rather than manifesta- tions of any one factor alone, as shown by their tendency, both to have substantial second- ary loadings and also to ‘migrate’ between factors across studies (see discussions in Ashton et al. 2006, 2007). In several investigations, most of these sympathy- related terms have shown their strongest loadings on the agreeableness (versus anger) factor.

However, in several other investigations, most of those terms have instead shown their strongest loadings on the honesty-humility factor. Also, in a few other studies, several sympathy-related terms have shown substantial loadings on an emotionality factor; for example, the low pole of the English lexical emotionality factor was defined by terms such as pitiless, unsympathetic, and cold- hearted (Ashton et al., 2004a).

To conclude this section, we should briefly mention the nature of solutions involving more than six factors, as obtained from standard lexical studies of personality structure (i.e. those based on personality-descriptive terms only). There is evidence from a few studies that seven-factor solutions produces a separate factor for intellectual ability

(6)

(Ashton et al., 2004b), which defines a factor different from that defined by intellectual openness (i.e. creativity, unconventionality, etc.). Also, there is evidence from a few studies for a separation of two aspects of emotionality – specifically, fearfulness and sentimenta-lity – onto two separate dimensions within seven-factor solutions (Ashton et al., 2004b). In addition, some investigations that have included descriptors of religiosity and spirituality have produced a factor defined by those terms. This suggests that religiosity or spirituality represents an additional dimension beyond the B5/FFM (e.g. Piedmont, 1999), but as we have discussed elsewhere (e.g. Ashton et al., 2004a;

Lee et al., 2005c), we believe that religiosity or spirituality falls outside the domain of personality proper.

Summary

Lexical studies of personality structure have produced a common set of six dimensions across at least a dozen languages. Three of these factors are interpreted as extraversion, conscientiousness, and intellect/imagination/unconventionality, and closely resemble three of the English lexical Big Five factors.

Of the remaining factors, one is interpreted as honesty-humility, and the other two are interpreted as emotionality and agreeableness (versus anger); these last two factors differ in important respects from B5/FFM (low) emotional stability and B5/FFM agreeableness.

One noteworthy result of these studies has been the position of terms describing sympathy and soft-heartedness, which represent blends of honesty-humility, agreeableness (versus anger), and (to a lesser extent) emotionality.

The finding of this six-dimensional structure across diverse cultures has some important implications, as it contradicts the expectation that only five factors of personal- ity description would be replicated. In fact, the six-factor solution is actually somewhat more widely replicated than is the B5/FFM, which has failed to emerge in some languages that recovered the six-factor structure.

Conversely, there are apparently no languages in which standard lexical studies (i.e. those based solely on personality-descriptive terms) have failed to recover the above six dimensions in the six-factor solution. The emergence of the six-dimensional structure across diverse cultures and languages, from variable sets that are both indigenous and representative of the personality domain, gives this model a wider basis of independent empirical support than is possessed by the B5/FFM.

Recently, we have referred to the six- factor framework as the HEXACO model of personality structure, and we have operationalized that framework in the HEXACO Personality Inventory (HEXACO-PI) (Lee and Ashton, 2004, 2006b). The name of this model reflects the names of the factors – honesty-humility (H), emotionality (E), extraversion (X), agreeableness (A), conscientiousness (C), and openness to experience (O) – and also their number (i.e. the Greek hexa, six). (Note that, despite its lexical ori- gins, the HEXACO model uses the name openness to experience rather than intellect/

imagination/ unconventionality.) Names of the revised HEXACO Personality Inventory (HEXACO-PI-R) facet scales are listed in Table 12.2.

THEORETICAL INTERPRETATION:

LINKS WITH CONSTRUCTS FROM EVOLUTIONARY BIOLOGY

As shown in the above summary, lexical studies of personality structure in diverse languages have generated a six-dimensional space that we call the HEXACO structure. But this raises the question of why human person- ality variation would be characterized by this particular set of dimensions. We believe that the emergence of the HEXACO factors fol- lows from the importance of the adaptive trade-offs that are associated with behaviors corresponding to several important constructs from theoretical biology. Also, we suggest that these interpretations can explain and predict

(7)

several important personality phenomena that would not otherwise be understood.

As we describe below, the theoretical framework associated with the HEXACO model is based on two broad concepts. First, the honesty-humility, agreeableness (versus anger), and emotionality factors correspond to biologists’ constructs of reciprocal and kin altruism. Second, the extraversion, conscientiousness, and openness to experience factors represent three parallel dimensions each describing engagement or investment within its own respective area or variety of endeavor.

Our interpretations of the six factors are summarized below and in Table 12.2, with a focus on the probable adaptive trade-offs associated with high and low levels of each dimension (see also Ashton and Lee, 2001;

Lee and Ashton, 2004).

First, we have suggested that the honesty- humility and agreeableness (versus anger) correspond to two complementary aspects of the construct of reciprocal altruism (Trivers, 1971). Honesty-humility represents the tendency to be fair and genuine in dealing with others, in the sense of cooperating with others even when one might exploit them without suffering retaliation. Agreeableness (versus anger) represents the tendency to be forgiving and tolerant of others, in the sense of cooperating with others even when one might have suffered some exploitation by them. According to this interpretation, high levels of honesty-humility are associated with decreased opportunities for personal gains from the exploitation of others, but also with decreased risks of losses from with- drawal of cooperation (or further retaliation) by others. By the same logic, high levels of agreeableness (versus anger) are associated with increased opportunities for personal gains from long-run reciprocal cooperation with others but also with increased risks of losses from exploitation by others. (Here we use the term altruism in terms of a dimension of altruism versus antagonism, to include both a willingness to help or provide benefits to others but also an unwillingness to harm or impose costs on others.)

We have also suggested that emotionality corresponds to tendencies that are associated with the construct of kin altruism (Hamilton, 1964). These tendencies include empathic concern and emotional attachment toward close others (who tend to be one’s kin), but also the harm-avoidant and help-seeking behaviors that are associated with investment in kin (see also Lee and Ashton, 2004). Interpreted in this way, high levels of emotionality are associated with increased likelihood of personal and kin survival, but also with decreased opportunities for gains that are often associated with risks to personal and kin survival.

The remaining three personality factors – extraversion, conscientiousness, and openness to experience – are interpreted in our theoretical framework as tendencies to become engaged in three different areas or varieties of endeavor. Specifically, extraversion represents engagement in social endeavors (such as socializing, leading, or entertaining), conscientiousness represents engagement in task-related endeavors (such as working, planning, and organizing), and openness to experience represents engagement in idea-related endeavors (such as learning, imagining, and thinking). According to these interpretations, high levels of any of these three dimensions are associated with increased opportunities for gains resulting from the investment of one’s energy and time in those areas. For example, depending on the social and ecological circumstances in a given environment, high extraversion may promote gains of a social nature (i.e. access to friends, allies, and mates), high conscientiousness may promote gains of a material or economic nature (as well as improved health and safety), and high openness to experience may promote social and material gains that result from new discoveries. On the other hand, however, high levels of any of these dimensions would also be associated with increased costs in terms of expended energy and time and, in some cases, of risks from the natural and social environment.

The interpretations of extraversion, conscientiousness, and openness to experience

(8)

as dimensions of engagement or endeavor suggest links with some constructs from theoretical biology (see details in Ashton and Lee, 2007). As noted elsewhere (Ashton et al., 2002), the interpretation of extraversion as social endeavor suggests similarities to the concept of social-attention holding power, a variant of resource-holding potential. Similarly, the interpretation of conscientiousness as task- related endeavor suggests similarities to some of the (non-altruistic) aspects of the r versus K continuum, particularly, the consistent exploitation of resources. Finally, the interpretation of openness to experience as idea-related endeavor suggests similarities to the contrast between individual learning and imitation.

A noteworthy advantage of the theoretical framework described above is that it parsi- moniously explains the existence of three separate factors relevant to altruism. That is, the recurrent emergence of the honesty- humility, agreeableness (versus anger), and emotionality factors is immediately understood when these are interpreted with refer- ence to reciprocal altruism and with kin altruism. In contrast, the emergence of these three factors is not explained by any interpretation of the B5/FFM framework, which includes only two dimensions within this three-dimensional segment of the HEXACO space, and provides no hint as to the nature of any more differentiated structure. Moreover, the interpretation of these dimensions in terms of altruism-related constructs allows an understanding of several phenomena that would otherwise be unexplained. For example, the large sex differences on and the diverse content of the emotionality factor are readily grasped when this factor is explained in terms of kin-altruistic tendencies (see Ashton and Lee, 2007, for a detailed discussion).

PRACTICAL SIGNIFICANCE:

ACCOMMODATION OF IMPORTANT PERSONALITY VARIABLES

Earlier in this chapter we described the empirical basis of the HEXACO framework

in the results of lexical studies of personality structure, and we also discussed the theoretical interpretations of the six dimensions.

But there remains the question of practical implications: In particular, does the larger space of the HEXACO structure provide any advantages beyond the B5/FFM space in accommodating personality traits and personality-relevant criteria? Such advantages could be expected, given that some variables associated with the three-dimensional space of the HEXACO honesty-humility, agreeableness (versus anger), and emotionality factors might have weaker projections within the two-dimensional space of B5/FFM agreeableness and emotional stability (versus neuroticism). Below, we briefly summarize some previous studies that have investigated the ability of the HEXACO model and the B5/FFM to capture a variety of personality traits and related criteria.

The HEXACO honesty-humility factor is defined by characteristics that tend either to be weakly represented within measures of the B5/FFM or to be subsumed within a very broad version of B5/FFM agreeableness (see Ashton and Lee, 2005b). As a consequence of this heavier representation of honesty- humility-related traits within the HEXACO model, that framework has outperformed the B5/FFM in predicting several variables of practical importance. These variables – which have shown strong negative associa- tions with honesty-humility – include work- place delinquency (see Lee et al., 2005b), sexual harassment proclivities (see Lee et al., 2003), and the traits of primary psychopathy, Machiavellianism, and narcissism (see Lee and Ashton, 2005).

The practical usefulness of the honesty- humility factor can also be seen by examin- ing the correlates of the NEO-PI-R (Costa and McCrae, 1992), which assesses several narrow traits or facets within each B5/FFM domain. In one study (Ashton and Lee, 2005b), we found that the straightforward- ness and modesty facets of NEO-PI-R agreeableness were associated with HEXACO honesty-humility rather than with the English

(9)

lexical Big Five version of agreeableness.

Moreover, these same two facets were also strong predictors of two personality variables involving insincerity – namely (low) self- monitoring and (low) social adroitness.

(Note, however, that these two NEO-PI-R facets do not represent the full range of honesty-humility content, and would be unlikely to provide optimal prediction of some other honesty-humility-related variables, including (at the negative pole) materi- alism or corruption-proneness.)

The HEXACO framework also had some success in accommodating a range of variables that have been suggested to fall

‘beyond the Big Five’. A diverse set of such variables has been operationalized in the scales of the Supernumerary Personality Inventory (SPI) (Paunonen et al., 2003), which we examined in a recent study (Lee et al., 2005c). The results of that investigation showed that the HEXACO-PI scales outperformed measures of the B5/FFM dimensions in predicting several variables that are conceptually similar to honesty-humility, including integrity, (low) manipulativeness, (low) egotism, and (low) seductiveness. There was also a substantial predictive advantage of the HEXACO variables over those of the B5/FFM in predicting two SPI scales – femininity and (low) risk taking – that are conceptually related to emotionality. This latter result suggests that the predictive advantages associated with the HEXACO model are not limited to honesty-humility, but also extend to the emotionality domain.

Some traits within that domain, including femininity, harm-avoidance, and depend- ence, are typically not included within measures of any B5/FFM dimension.

Note that although most of the studies summarized above were based exclusively on self-reported personality, similar results have been obtained from peer reports (e.g. in prediction of likelihood to sexually harass;

see Lee et al., 2003). In a recent study by Lee et al. (in press), self- and peer reports of honesty-humility, each outperformed self-reports of all B5/FFM variables in

predicting scores on an overt integrity test and a business ethical dilemmas task. That is, the cross-source correlations between these outcome variables and honesty-humility exceeded any of the within-source correlations between the outcome variables and the B5/FFM. This result suggests that the strong correlations, previously observed, between honesty-humility and outcome variables are due to actual behavioral co-occurrence, and not merely to artifactual covariation resulting from the use of self-report assessment.

Future validity studies of the kind described above are still warranted, but the results of comparisons conducted thus far do suggest some advantage of the HEXACO model over the B5/FFM in accommodating several important personality variables. As summarized above, the HEXACO honesty- humility and emotionality factors appear to accommodate several personality traits and personality-related constructs that are less well assimilated within the space of the B5/FFM.

CONTRASTS WITH OTHER LEXICALLY DERIVED MODELS: FACTOR

INTERPRETATION AND REPLICABILITY

As summarized above, the results of lexical studies of personality structure have shown a rather high degree of consistency, producing a similar six-dimensional structure even across very different languages. In fact, this six-factor solution has actually been replicated somewhat more widely across languages than the Big Five structure, which was not recovered in the lexical studies conducted in Hungarian and in Italian (Trieste). Thus, the six-dimensional framework not only provides a more differentiated representation of the personality domain than the five- dimensional framework does, but also does so without incurring any cost in terms of cross-cultural replicability.

Recently, Saucier (2003) has proposed an alternative lexically derived structural model

(10)

of personality that also involves a space larger than that of the Big Five. This alternative structure, which has been called the ‘Multi- Language Seven’ (ML7; Saucier, 2003), was derived from results of investigations of the personality lexicons of two languages – Hebrew and Filipino – as supplemented by some analyses of English-language personality-descriptive adjectives. We will discuss the content of the proposed ML7 dimensions below, but we first note that six of the ML7 factors share elements of content with the six factors that have been replicated across languages. The remaining ML7 factor (‘negative valence’) is defined by terms that have been excluded from most lexical studies of personality structure on the grounds that they provide extreme negative evaluations, rather than descriptions corresponding to normal personality dispositions.

Given the proposal of this alternative lexically based model of personality structure, it would be useful to review some features of that model. Below, we discuss two features of the ML7 structure, specifically (a) the interpretation of the negative valence factor, and (b) the extent to which the remaining six substantive factors of the ML7 are consistent with the six dimensions that have been recovered across many languages.

Interpretation of negative valence Recall that the purpose of lexical studies of personality structure is to find the major dimensions that underlie the domain of personality characteristics. Therefore, the variable sets of these investigations have generally been restricted to words that serve chiefly as personality descriptors (e.g. outgo- ing, stubborn, organized, unemotional, pre- tentious, inquisitive); in contrast, adjectives that serve chiefly to evaluate an individual (e.g. terrible, good-for-nothing, wonderful, excellent) have been excluded.

In some lexically based investigations, however, researchers have included primarily evaluative terms, on the grounds that such

terms might define factors that would indicate an individual’s level of self-esteem (Tellegen, 1993). For example, an individual who describes himself or herself as terrible or good-for-nothing is likely to have very low self-esteem; therefore, a factor defined by these terms would represent a dimension of low self-esteem.

This inference about the self-esteem of respondents who endorse extremely undesirable terms is probably accurate. However, this does not justify the inclusion of primarily evaluative terms in lexical studies of personality structure. According to the logic of the lexical approach, any hypothesized personality dimension of low self-esteem would have to be defined by terms that describe persons having low self-esteem – terms such as self-critical, self-disparaging, and so on (see Ashton and Lee, 2001, 2002). To include other terms as potential signs (rather than descriptors) of low self-esteem is inconsistent with the purpose and rationale of the lexical approach. (Note, of course, that the above argument does not deny the importance of low self-esteem as a personality characteristic.)

Note that when primarily evaluative terms are included in the variable sets of lexical investigations, a ‘negative valence’ factor often does emerge, being defined by extremely undesirable adjectives that are infrequently endorsed in self-ratings (e.g.

Saucier, 1997).² Recently, it has been suggested that this factor is indeed a meaningful descriptive dimension of personality variation, on the grounds that some of the defining adjectives of this factor do, in fact, have descriptive aspects (Saucier, 2002, 2003).

For example, stupid suggests low intellect, evil suggests low kindness, and crazy sug- gests low conventionality.

These terms likely do have descriptive aspects, and would be expected to correlate somewhat with various personality scales that share similar descriptive content. However, this does not mean that the negative valence factor is a meaningful descriptive dimension.

To the extent that terms such as stupid and evil and crazy together define a negative

(11)

valence factor – instead of separately defining the factors that describe high versus low levels of intellect, of kindness, and of conventionality – this suggests that those terms covary because of a strong common element of extreme undesirability and/or infrequent endorsement. This interpretation is supported by the prevalence on negative valence of the more purely evaluative terms (see Saucier, 1997, for example), which further suggests that terms such as stupid, evil, and crazy (despite their descriptive aspects) serve mainly as terms of negative evaluation rather than as descriptors of personality dispositions.

Related to the above point, it is likely that persons who endorse negative valence adjectives will be rather poorly socialized, and will tend to have markedly low scores on self-report measures of socialization-related traits. This might suggest an interpretation of negative valence as a descriptive dimension of ‘social deviance’ (e.g. Saucier, 2002) or ‘social impropriety’ (Saucier, 2003).

However, we believe that interpretation is inconsistent with the fact that the negative valence factor is not defined by most of the terms that actually describe persons’ levels of socialization or social conformity (e.g.

conventional, self-controlled, polite, tame, proper versus rebellious, impulsive, rude, wild, nonconforming). Those terms instead load on other dimensions, leaving a negative valence factor that is defined by various extremely undesirable terms that have very low mean self-ratings.³

The findings discussed above indicate that the negative valence factor cannot be interpreted substantively. Although some terms that define negative valence do have some descriptive aspects, the crucial fact remains that those terms do not load on the same factors as do the terms that share those descrip- tive aspects without being characterized by extreme negative evaluation. Instead, the distinguishing common element of the terms that define negative valence is their extreme social undesirability and/or their low endorsement rates.

Relations of other ML7 factors with cross-language

(HEXACO) dimensions

As we have noted elsewhere (see Ashton and Lee, 2002), the remaining six substantive factors of the ML7 correspond fairly closely to the factors that we have proposed as the major dimensions of personality (e.g.

Ashton and Lee, 2001). Four ML7 factors – gregariousness, conscientiousness, intellect, and even temper – correspond fairly closely overall to the extraversion, conscientiousness, intellect/imagination/unconventionality (openness to experience), and agreeableness (versus anger) dimensions as observed in lexical studies of personality structure, apart from the somewhat narrower content of the ML7 factors. In addition, ML7 self- assurance resembles to some extent the low pole of cross-language emotionality, but lacks the sentimentality, hypersensitivity, and emotionality that typically define the latter dimension. Finally, ML7 concern for others has some elements in common with cross- language honesty-humility, particularly (at the low pole) conceit. However, ML7 concern for others excludes much prototypical honesty-humility content (e.g. sincere and unassuming versus deceitful, hypocritical, materialistic, and pretentious).

RECENT FINDINGS: THE STRUCTURE OF THE CROATIAN, GREEK, AND FILIPINO PERSONALITY LEXICONS

In our 2004 review of lexical studies of personality structure, we drew attention to the similar six-factor solutions obtained from investigations conducted in various languages (Ashton et al., 2004b). Since that time, studies of the personality lexicons of two other languages – Croatian (Mlacic and Ostendorf, 2005) and Greek (Saucier et al., 2005)–have also been published. Although the results of those investigations show six- factor solutions very similar to those that

(12)

have been replicated across languages, the original articles did not draw attention to these similarities. In the Croatian study, no systematic comparison with previous six- factor solutions was undertaken, and in the Greek study, the analyses do not allow clear comparisons with the results of previous investigations. Therefore, we describe below the six-factor solutions obtained from the Croatian and Greek lexicons (see Lee et al., 2005a; Lee and Ashton, 2008). In addition to the recent Croatian and Greek studies, lexically based investigations of personality structure were conducted in the Filipino language during the 1990s (Church et al., 1997, 1998). However, those investigations were based on variable sets that included ‘negative valence’ terms, and therefore, we describe here the solutions derived from Filipino variable sets that exclude such terms, and hence are more directly comparable to solutions from other languages (see Ashton et al., 2006a).

Croatian lexical personality factors In their report describing the structure of the Croatian personality lexicon, Mlacic and Ostendorf (2005) focused on the five-factor solutions obtained from 453 adjectives (self- ratings) and 455 adjectives (peer ratings), but also indicated some features of the six-factor solutions. Here, we will focus mainly on the Croatian self-rating results, but we also mention briefly the similar results obtained from Croatian peerrating data. Using the informa- tion reported by Mlacic and Ostendorf in combination with that reported for the same data set by Ostendorf et al. (2004), we have listed in Table 12.3 the Croatian adjectives that best represent each of the varimax-rotated factors in the self-rating six-factor solution.

Below, we reinterpret the Croatian six-factor solution on the basis of that factor content.

The adjectives representing the Croatian factors that we interpret as extraversion, conscientiousness, and intellect were taken from the corresponding factors in the five-factor solution reported by Mlacic and Ostendorf

(2005), who arrived at the same interpretations. (Correlations between the corresponding five-factor and six-factor versions of these three factors were 0.98 or above.) Note that the Croatian intellect factor is, indeed, heavily defined by intellect-related terms, with many of those terms describing cogni- tive ability (e.g. intelligent, smart, bright), rather than a disposition toward intellectual endeavor; in addition, there are several terms suggesting imagination, but no terms suggesting unconventionality (such terms appear to have been largely absent from the Croatian variable set).

For two of the remaining three dimensions of the Croatian six-factor solution, the highest- loading terms were listed explicitly by Ostendorf et al. (2004). One of these factors was very similar in content to the cross- language agreeableness (i.e. agreeableness versus anger) factor (cf. Ashton et al., 2004b, table 2), being defined by adjectives translated as unirritable and gentle versus explosive, quick-tempered, stubborn, and quarrelsome.

(Recall that this content corresponds to the cross-language agreeableness factor, but not to Big Five agreeableness.)

The other factor whose defining terms were listed by Ostendorf et al. (2004) corresponded closely to the cross-language honesty-humility factor (cf. Ashton et al., 2004b, table 5), being defined by adjectives trans- lated as honest, fair, and just versus greedy, perfidious, rapacious, covetous, overbear- ing, conceited, fame-thirsty, self-important, self-interested, and hypocritical. Interestingly, the Croatian honesty-humility factor also shows strong loadings for terms such as generous and good-hearted; as noted above, these traits largely represent blends of the cross-language, honesty-humility, and agreeableness (versus anger) factors, sometimes showing their strongest loadings on one factor, and sometimes on the other (see discussion above).

For the last of the Croatian self-rating factors, we have listed the strongest defining terms from this dimension in the five-factor solution, which correlated 0.92 with its

(13)

counterpart in the six-factor solution (cf. Mlacic and Ostendorf, 2005, figure 2, table 3); however, we have removed those terms that shifted to the cross-language agreeableness (versus anger) factor of the six-factor solution. As seen in Table 12.3, the self-rating version of the factor is characterized by adjectives suggesting unemotionality versus oversensitivity. Thus, the adjectives defining the low pole of this dimension are among those that have consistently defined the emotionality factor in previous lexical studies (cf. Ashton et al., 2004b, table 4).

Inspection of the content of this factor, however, suggests that this Croatian dimension is somewhat more narrowly defined than are its counterparts in most other languages, being defined most strongly only by terms related to emotionality and sensitivity, and not by terms suggesting fearfulness and vulnerability.

Turning briefly to the peer rating six-factor solution, the six factors were generally very similar to those of the self-rating solution:

three factors were interpreted by Mlacic and Ostendorf (2005) as extraversion, conscientiousness, and intellect, and two others were nearly identical in content to the self-rating dimensions that we interpreted as cross-language agreeableness and honesty-humility.

The remaining peer rating factor was small, being defined by only 11 terms, but these were similar to those of the self-rating solution;

the six terms listed by Mlacic and Ostendorf were romantic, emotional, sentimental, and naïve versus unemotional and insensitive.

To summarize, the Croatian personality lexicon has produced six factors whose defining content corresponds closely to that of the six factors observed in many other languages. The Croatian six-factor solution con- tains dimensions clearly interpretable as extraversion, conscientiousness, agreeableness (versus anger), and honesty-humility, and two other dimensions that – although rather narrow in content, probably because of the specific rules used in variable selection – are defined exclusively by terms associated with the intellect and emotionality factors of other languages.

Greek lexical personality factors The main analyses of Saucier et al. (2005) involved self-ratings of 991 respondents on the 400 most frequently used Greek personality terms. The authors reported summaries of analyses on both ipsatized and original (i.e., non-ipsatized) data, but focused mainly on the non-ipsatized responses. However, because almost all other lexical studies of personality structure have been based on ipsatized data, we will focus exclusively on the results obtained from the ipsatized responses. (See Lee and Ashton, 2007, for a discussion of the use of ipsatized and non- ipsatized data.)

As noted in the previous section, another feature of the investigation by Saucier et al.

(2005) was its use of a variable set that included primarily evaluative terms (abject, unpleasant, loser, barbarian, inhuman, dis- gusting, and useless). As would be expected, these terms defined a negative valence factor, which emerged in solutions containing three or more factors. This dimension was defined exclusively by terms with low mean responses:

Of the 20 highest-loading terms on negative valence, all had response means below 1.35 on the 1-to-5 response scale; conversely, of the 20 terms with the lowest response means, all loaded at least 0.30 on negative valence and less than 0.30 on all other factors.

We have reanalyzed the Greek adjective data set of Saucier et al. (2005) after excluding the negative valence terms. (As noted above, we ipsatized participants’ responses by standardizing across variables within respondents prior to performing factor analyses; in this way, our procedures would be consistent with those of all previous lexical studies, and would allow meaningful comparisons with results of those studies.) Also for the sake of consistency with previous lexical studies, we excluded the Greek adjectives that represented ‘negative valence’, by removing the 40 adjectives that loaded 0.30 or above on the negative valence factor in Saucier et al.’s ipsatized seven-factor solution. We factor analyzed the self-ratings of

(14)

Table 12.3Adjectives having highest loadings on the dimensions of the Croatian self- and peer rating six-factor solutions FactorSelf-rating adjectivesPeer rating adjectives ConscientiousnessOrganiziran (organized),marljiv (industrious),temeljit (thorough),radiˇsan Organiziran (organized),odgovoran (responsible),temeljit (thorough),radin (hard-working),odgovoran (responsible),uredan (orderly),vrijedan (sedulous),(diligent),radiˇsan (hard-working),vrijedan (sedulous),marljiv (industrious), sistematiˇcan (systematic),radin (diligent),precizan (precise) sistematiˇcan an (systematic),sustavan (systematic),metodiˇcan (methodical) versusversus neorganiziran (disorganized),nemaran (negligent),neodgovoran (irresponsible),neorganiziran (disorganized),neodgovoran (irresponsible),nemaran lijen (lazy),neoprezan (incautious),nesistematiˇcan (unsystematic),(negligent),nesistematiˇcan (unsystematic),lijen (lazy),povrˇsan (superficial), nesustavan (unsystematic),neuredan (disorderly),netoˇcan (inaccurate),nesustavan (unsystematic),zaboravljiv (forgetful),neozbiljan (frivolous), nesavjestan (unconscientious)nestalan (unsteady) Honesty-humilitySućutan (sympathetic),dobroduˇsan (good-hearted),obazriv (regardful),humanPoˇsten (honest),sirokogrudan (broad-minded),velikoduˇsan (generous), (humane),dobronamjeran (well-intentioned),poˇsten (honest),velikoduˇsan dobroduˇsan (good-hearted),human (humane),milosrdan (charitable), (generous),praviˇcan (fair),milosrdan (charitable),sirokogrudan (broad-minded)pravedan (just),osjećajan (compassionate),topao (warm),sućutan versus(sympathetic),dobronamjeran (well-intentioned) licemjeran (hypocritical),pohlepan (greedy),grabezljiv (rapacious),sebiˇcan versus (selfish),koristoljubiv (self-interested),gramziv (covetous),uobraˇzen pohlepan (greedy),podmukao (perfidious),grabezljiv (rapacious),sebiˇcan (conceited),egocentriˇcan (self-centered),podmukao (perfidious),umiˇsljen (selfish),podao (mean),gramziv (covetous),prepotentan (overbearing), (self-important),podao (mean),slavohlepan (fame-thirsty)uobraˇzen (conceited),slavohlepan (fame-thirsty) ExtraversionKomunikativan (communicative),ekstravertiran (extraverted),razgovorljivOtvoren (open),razgovorljiv (loquacious),komunikativan (communicative), (loquacious),otvoren (open),druˇstven (sociable),energican (energetic),govorljivpokretljiv (brisk),dinamiˇcan (dynamic),poduzetan (enterprising),okretan (talkative),dinamiˇcan (dynamic),druˇzeljubiv (companionable),aktivan (active)(nimble),energiˇcan (energetic),pokretan (agile),ekstravertiran (extraverted) versus versus zatvoren (reserved) povuˇcen (withdrawn),nedruˇstven (unsociable),stidljivzatvoren (reserved),ˇsutljiv (taciturn),stidljiv (bashful),negovorljiv (bashful),negovorljiv (untalkative),introvertan (introverted),introvertiran(untalkative),povuˇcen (withdrawn),srameˇzljiv (shy),introvertiran (introverted),ˇsutljiv (taciturn),srameˇzljiv (shy),neenergiˇcan (unenergetic)(introverted),introvertan (introverted),pasivan (passive),stidan (demure) AgreeablenessNerazdraˇzljiv (unirritable),smiren (tranquil),blag (gentle),miroljubiv (peaceful)Nerazdraˇzljiv (unirritable),smiren (tranquil),popustljiv (complying),posluˇsan (versus Anger)versus (obedient),tolerantan (tolerant),strpljiv (patient),blag (gentle) ratoboran (belligerent),eksplozivan (explosive),raspaljiv (quick-tempered),versus ˇzestok (fierce),tvrdoglav (stubborn),razdraˇzljiv (irritable),prkosan (defiant),eksplozivan (explosive),svadljiv (quarrelsome),raspaljiv (quick-tempered), inatljiv (spiteful),oˇstar (harsh),nagao (rash),agresivan (aggressive),impulzivansvojeglav (self-willed),tvrdoglav (stubborn),inatljiv (spiteful),razdraˇzljiv (impulsive),svojeglav (self-willed),buntovan (rebellious),svadljiv (quarrelsome),(irritable),oˇstar (harsh),ˇzestok (fierce),nagao (rash),ratoboran (belligerent), nepopustljiv (unyielding)prkosan (defiant),otresit (surly)

(15)

IntellectDarovit (gifted),talentiran (talented),uman (sagacious),stvaralaˇcki (originative),Bistar (bright),bistrouman (quick-witted),uman (sagacious),oˇstrouman bistar (bright),pametan (smart),intelektualan (intellectual),nadaren (endowed),(sharp-witted),domiˇsljat (ingenious),intelektualan (intellectual),talentiran oˇstrouman (sharp-witted),bistrouman (quick-witted)(talented),misaon (reflective),maˇstovit (imaginative),pametan (smart) versus versus nekreativan (uncreative),nedarovit (ungifted),neinteligentan (unintelligent),neinteligentan (unintelligent),nedarovit (ungifted),povodljiv (suggestible), neintelektualan (unintellectual),nemaˇstovit (unimaginative),nevjeˇst nekreativan (uncreative),neintelektualan (unintellectual) (unskillful),neprofinjen (unsophisticated) (low) Emotionality*Neemocionalan (unemotional),neosjetljiv (insensitive),neosje´cajan Neemocionalan? (unemotional),neosjetljiv? (insensitive) uncompassionate),( bezosje´cajan (without compassion),hladnokrvanversus (cool-blooded),flegmatiˇcan (phlegmatic)romantican? (romantic),emocionalan? (emotional),sentimentalan? versus(sentimental),naivan? (naïve) preosjetljiv (oversensitive),osjetljiv (sensitive),romantiˇcan (romantic),strastan (passionate),strastven (impassioned),povodljiv (suggestible), emocionalan? (emotional),sentimentalan? (sentimental) Note:Adjectives for the self- and peer rating factors interpreted as conscientiousness,extraversion,and intellect and for the peer rating factor interpreted as agreeableness (versus anger) were taken from the five-factor solutions reported by Mlacic and Ostendorf (2005,tables 4 and 6).Adjectives for the self- and peer rating factors interpreted as honesty-humility and for the self-rating factor interpreted as agreeableness (versus anger) were taken from Ostendorf et al.(2004).Adjectives for the self-rating factor interpreted as low Emotionality were selected from Mlacic and Ostendorf’s (2005) table 4,according to the method described in the text of this article.English translations of adjectives for peer rating factor interpreted as Intellect were those listed by Mlacic and Ostendorf (2005).See text for cautions regarding content of the peer rating factor interpreted as intellect *The peer-rating version of this factor was defined by 11 terms,6 of which were listed,in English translation only,by Mlacic and Ostendorf (2005)

(16)

all 991 respondents on the remaining set of 360 adjectives by the method of principal components (see Lee and Ashton, 2008).

We applied a varimax rotation to the six Greek lexical factors, and obtained dimensions that corresponded closely to those observed in other languages. However, we noticed that the extraversion and emotionality factors were rotated slightly away from their usual axis locations, so we rotated these two varimax axes by 15 degrees within that plane. Table 12.4 shows the loadings of the highest-loading adjectives on each of the six factors, as obtained from varimax followed by this additional re-rotation.

With regard to interpretation of these dimensions, the first factor corresponded fairly closely to extraversion, being defined by such characteristics as sociability and cheerfulness versus their opposites.

The second factor represented conscientiousness, being defined by terms describing organization and industriousness versus their opposites. The third factor resem- bled the cross-language agreeableness dimen- sion, being defined by content suggesting patience and politeness versus irritability and stubbornness.

The fourth factor resembled the cross- language emotionality factor, being defined by content suggesting vulnerability, (hyper) sensitivity, fearfulness, and emotionality versus their opposites. The fifth factor resembled honesty-humility, being defined by adjectives relevant to the various aspects of that factor, including fairness (e.g. ethical versus unscrupulous) and modesty (e.g. modest versus grandiose and exhibitionist) versus insincerity (e.g. sly) and greed (e.g. profiteer).

Finally, the sixth factor can be interpreted as an intellect/imagination dimension, being defined by such terms as talented, inventive, and ingenious. This factor is relatively small compared to the other five factors, but its defining content is broadly similar to that of the intellect/imagination/unconventionality factor observed across languages, except for the absence of content related to unconventionality or philosophicalness in the Greek

version of the factor. Overall, the content of the six factors suggests that this solution is very similar to the common six-factor structure observed across other languages.⁴

Filipino lexical personality factors The Filipino personality lexicon has been examined in a series of studies (Church et al., 1997, 1998) in which ‘negative valence’

terms were included along with adjectives that serve primarily as personality descriptors. To facilitate comparisons with results of previous studies, the authors of those studies did comment on the five-factor solutions obtained when negative valence terms were excluded from analyses, noting that the Big Five structure was not recovered. Given the widespread cross-language replication of a six-dimensional solution, the six-factor space derived from the personality-descriptive adjectives of the Filipino language is of some interest, and we summarize those results here.

Table 12.5 shows the highest-loading terms on the varimax-rotated factors of the six-factor solution derived from the set of 232 adjectives obtained by Church et al. (1997) after removal of negative valence terms, and based on self- ratings from 1,529 persons. As seen in that table, the first factor was defined by a variety of terms related to overall altruism and to the core aspects of honesty-humility, and thus represents a broad variant of that factor.

The second factor contained terms typical of conscientiousness (e.g. disciplined, orderly versus lazy), but also has strong elements of religiosity and thriftiness, which obscure somewhat the interpretation of this factor.

The content of the third factor involved irritability and ill-temper versus patience, and thus corresponds to cross-language agreeableness versus anger (but not to classic Big Five agreeableness); interestingly, terms describing overall altruism (e.g. kind, understanding) tended to divide their loadings between this factor and the first factor (i.e. honesty- humility). The fourth factor corresponded