McCrae and Costa used the findings from the lexical tradition and combined them with results from factor analysis of personality questionnaires and personality assessment theory to develop the Five Factor Model (FFM). FFM is a hierarchical organisation of personality in five higher-order traits and thirty lower-order facets (McCrae & John, 1992). The instruments designed to assess the FFM, such as the NEO-PI-R (measuring both factors and facets) or the NEO-FFI (measuring factors only) (Costa & McCrae, 1992), conceptualise the five factors slightly differently than the original B5. The main difference is the inclusion of intellectual abilities on the B5 Intellect/Imagination dimension, whereas intellect-related personality traits are excluded from the corresponding FFM Openness to Experience factor. Another noteworthy difference is that the B5 Agreeableness dimension includes “warmth,” which is moved to Extraversion in FFM. (See Goldberg (1993) for more details on the differences between B5 and FFM.) Finally, McCrae and Costa aimed to develop a personality inventory that could measure distinctive lower-order constructs. Thus, traits in the FFM are generally somewhat broader conceptualised than in the B5. For instance, the FFM Agreeableness dimension includes traits related to sincerity, which in lexical studies are in the periphery of the B5 Agreeableness dimension (Ashton & Lee, 2005b).
After an era of investigations on the B5/FFM, the focus shifted towards exploring dimensions beyond the Big Five dimensions. This shift was partly prompted by lexical studies in Italian (Di Blas & Forzi, 1998) and Hungarian (Szirmák & Raad, 1994) which revealed five-factor solutions that did not map easily onto the Big Five. In these studies, the fifth factor was not characterised by openness-related personality adjectives as in the classical B5 but by personality terms that were summarised with the labels Trustworthiness (Italian) and Integrity (Hungarian), respectively. Ashton, Lee, & Perugini et al. (2004) explored six-factor structures in seven lexical studies (including Italian and Hungarian) and extended the finding of a dimension characterised by adjectives such as sincere, honest, and modest (positive pole) and deceitful, conceited, and greedy (negative pole) to French, Dutch, Korean, and Polish. Furthermore, their study showed that six factors were stable across the languages. Highly similar six-factor structures have also been found in German (Ashton, Lee, Marcus, & Vries, 2007), English (Ashton, Lee, & Goldberg, 2004), Turkish (Wasti, Lee, Ashton, & Somer, 2008) and Greek (Lee & Ashton, 2009).
This cross-language replicable six-factor structure is the basis for the HEXACO framework (Ashton & Lee, 2007) which depicts the personality traits Honesty-Humility (H), Emotionality (E), Extraversion (X), Agreeableness (vs. anger) (A), Conscientiousness (C), and Openness to Experience (O). Both the 100 item HEXACO-PI-R and the HEXACO-200 can be used to measure the six broad factors as well as 25 lower-order facets. The latter of these two, with 200 items, is recommended when researchers need to assess facets with high reliability (Lee & Ashton, 2016). The HEXACO-60 is available for the assessment of factors only (Ashton & Lee, 2009). The additional H-factor is the most prominent amendment from the B5/FFM. Persons with high scores on the H-factor avoid manipulating others for personal gain, obey rules, have little interest in luxuries, and feel no special entitlement to elevated social status. Conversely, low scorers on this scale will manipulate others, are inclined to break rules for personal profit, are motivated by material gain, and have a strong sense of entitlement (Ashton, Lee, & de Vries, 2014). Two other key conceptual differences between the models are: (1) the location of anger-related personality dispositions (hostility and temper) on Agreeableness in the HEXACO which in the B5/FFM are subsumed under low Emotional Stability/high Neuroticism and (2) Emotionality is characterised by dependency and sentimentality which are mainly associated with the high end of Agreeableness in the B5/FFM (Ashton et al., 2014). Thus, the content of HEXACO Agreeableness and Emotionality dimensions differ somewhat from B5/FFM Agreeableness and Emotional Stability/Neuroticism (Ashton et al., 2014; Ashton, Lee, & Perugini et al., 2004; Boies, Yoo, Ebacher, Lee, & Ashton, 2004). Finally, the HEXACO model also includes the interstitial facet Altruism (sympathetic and kind), where factor loadings vary between Honesty, Agreeableness, and Emotionality across samples (Ashton & Lee, 2007). The content of HEXACO’s Extraversion and Conscientiousness dimensions is approximately equivalent to the B5/FFM counterparts (Ashton et al., 2014). In contrast, adjectives related to intellectual abilities are excluded from the HEXACO Openness factor. That dimension is thus conceptualised similarly to the FFM Openness to Experience dimension (Ashton & Lee, 2007).
As mentioned above, the FFM Agreeableness factor is more broadly defined than the B5 Agreeableness dimension. FFM Agreeableness, as operationalised in the NEO-PI-R, includes the facets Straightforwardness and Modesty which conceptually and empirically show considerable overlap with Honesty-Humility (Ashton & Lee, 2005b). Furthermore, Dutifulness – a facet of the FFM Conscientiousness factor – shares some variance with Honesty-Humility (Ashton & Lee, 2005b). Thus, it has been argued that Honesty-Humility can be fitted within the five-factor framework (McCrae & Costa, 2008). However, the HEXACO model explains more of the total variance in the FFM than vice versa (Ashton & Lee, 2018; Gaughan, Miller, & Lynam, 2012), and not all facets within Honesty-Humility are adequately captured by the FFM Agreeableness dimension (Ashton & Lee, 2005b). Additionally, Honesty-Humility has higher predictive power than FFM Agreeableness on conceptually relevant criteria (Ashton & Lee, 2008). Honesty-Humility and Agreeableness are theoretically different, with (low) Honesty-Humility referring to the tendency to defect when there is an opportunity to exploit another person. In contrast, a person with low Agreeableness would defect when there is a perception of being exploited (Ashton et al., 2014). Finally, the HEXACO model’s predictive advantage over FFM might not be solely due to the additional Honesty-Humility factor but rather to its better organisational structure which is caused by the relocation of personality traits in the HEXACO model (Gaughan et al., 2012). In sum, this tendency implies that Honesty-Humility and the HEXACO model cannot fully be accounted for by the FFM or the B5 (Ashton & Lee, 2018).
The present study
Research on various Big Five inventories (or the FFM) has largely confirmed the five-factor structure of personality in Norwegian samples (Engvik & Føllesdal, 2005; Martinsen, Nordvik, & Østbø, 2005; Vassend & Skrondal, 1997) but exploration of personality structures beyond the B5/FFM within a Norwegian context remains scarce. A Norwegian version of the HEXACO-PI-R was developed by the author two years ago, and in a master thesis, the HEXACO-60, which consists of selected items from the Norwegian HEXACO-PI-R, showed adequate factor-scale internal reliability (Gundhus, 2018). However, the HEXACO-PI-Rs psychometric properties and factor structure have not yet been evaluated formally in a Norwegian sample.
In the present study, the adequacy of the Norwegian version of the HEXACO-PI-R for measuring the six-factor HEXACO model in a Norwegian sample was examined. A total of 484 college students completed the Norwegian HEXACO-PI-R and the inventory’s reliability and validity were evaluated. First, scale internal consistency was examined for factors and facet scores of the HEXACO-PI-R. Second, the theoretical six-factor internal structure of the inventory was tested with exploratory factor analysis. In light of the discussion in the introduction on the conceptual and empirical similarities between Honesty-Humility and Agreeableness, of particular interest in the factor analysis was examining the distinctiveness of these factors in terms of different factor loading patterns. Third, the inventory’s convergent validity was examined by exploring relationships between the HEXACO factors and the B5. Based on the described similarities and differences between the HEXACO and the B5 in the introduction, evidence for construct validity was indicated by (1) high correlations among HEXACO Extraversion, Conscientiousness, and Openness to Experience and their B5 counterparts; (2) medium correlation between HEXACO Agreeableness and B5 Agreeableness, with HEXACO Agreeableness also showing a low to modest negative correlation with Neuroticism; (3) medium correlation between Emotionality and Neuroticism, with Emotionality also having a low to modest correlation with B5 Agreeableness; (4) low to moderate correlation between Honesty-Humility and B5 Agreeableness; and (5) low relationships between Honesty-Humility and other Big Five factors. Finally, this study examined whether the HEXACO factors, and specifically Honesty-Humility, could be accommodated within the B5 framework. In line with previous findings (Ashton et al., 2014), it was expected that HEXACO factors collectively explained more of the total variance in the B5 factors than vice versa. And it was specifically anticipated that the B5 factors would predict relatively little of the variance in Honesty-Humility.
Method
Participants
A total of 484 college students (396 female and 85 male (three participants did not indicate gender)); mean age = 24.98 years, SD = 7.09 years, range = 18-56 years) participated in the study. They were recruited from three study programs, all taught in Norwegian. The author contacted the college administrations and gave information about the study. He was then put in contact with persons with course responsibility. These individuals informed the students of the study and a collective decision was made as to whether course time should be designated to study participation. The students gave informed consent to participate and answered the questionnaires in paper-and-pencil format. To incentivise the students to participate, they were promised, and received, feedback on group-level results from the author.
Ethics
The data gathered for the purpose of this study is not personal data and as such did not require prior approval from the Norwegian Centre for Research Data (NSD). Further, because the aim of this study is to validate existing theories, it did not require approval from the regional committees for medical and health research ethics (REK).
Measures
HEXACO-PI-R – The HEXACO-PI-R consists of 100 items in the form of statements and measures the personality factors Honesty-Humility (H), Emotionality (E), Extraversion (X), Agreeableness (vs. anger) (A), Conscientiousness (C), and Openness to Experience (O) as well as their lower-order facets (25 facets in total). The participants were instructed to read each statement carefully and indicate whether they: 1 = strongly disagree; 2 = disagree; 3 = neither agree nor disagree; 4 = agree; and 5 = strongly agree. Example items includes: “I wouldn’t pretend to like someone just to get that person to do favours for me” (Honesty-Humility/Sincerity); “I sometimes feel that I am a worthless person” (reversed); (Extraversion/Social Self-Esteem); and “I don’t allow my impulses to govern my behaviour” (Conscientiousness/Prudence).
The Norwegian version of the HEXACO-PI-R was used. The translation from English to Norwegian was done separately by the author and a research assistant, who discussed and resolved any inconsistencies before the final draft was back-translated independently by two bilingual Norwegian-English speakers. The two back-translations showed satisfactory correspondence in terms of item wording and item content with the original English HEXACO-PI-R. The Norwegian HEXACO-PI-R back-translation was deemed adequate for use by the original authors (Michael Ashton & Kibeom Lee).
HEXACO-60 – For reasons described below (see Data analysis section), the HEXACO-60 was used in the simultaneous regression analysis. A subset of items from the HEXACO-PI-R constitutes the HEXACO-60 (Ashton & Lee, 2009). The inventory measures the six HEXACO trait dimensions (Honesty-Humility (H), Emotionality (E), Extraversion (X), Agreeableness (vs. anger) (A), Conscientiousness (C) and Openness to Experience (O)). Each scale consists of 10 items. Cronbach’s alpha in this study was Honesty-Humility = .66, Emotionality = .73, Extraversion = .71, Agreeableness = .71, Conscientiousness = .71, and Openness to Experience = .71.
The Big Five Inventory – The Big Five Inventory (BFI) consists of 44 items (statements), and responses are given on a seven-point Likert scale ranging from disagree strongly (1) to agree strongly (7) (Engvik & Føllesdal, 2005). The instrument measures the Big Five factors (Extraversion, Agreeableness, Conscientiousness, Neuroticism, and Openness). Cronbach’s alpha in this study was .79 for Extraversion, .71 for Agreeableness, .78 for Conscientiousness, .83 for Neuroticism, and .80 for Openness. Examples of items are: “Is full of energy” (Extraversion), “Values artistic, aesthetic experiences” (Openness to Experience), and “Is depressed, blue” (Neuroticism). The BFI was chosen because it stems from lexical research, taps the core of the Big Five dimensions, and has a high correspondence with the factors in the Five Factor Model (John et al., 2008).
Data analysis
Descriptive statistics was computed. Cronbach’s alpha internal scale reliability was calculated for the facets and factor-scale scores. In light of the controversy associated with assessing the internal structure of omnibus personality inventories with confirmatory factor analysis (Hopwood & Donnellan, 2010), the theoretical internal structure of the inventory was tested with item-level and facet-level exploratory factor analysis. These two analyses served somewhat different, albeit complementary, purposes. The aim of the item-level factor analysis was predominantly descriptive, that is to investigate to what extent items designed to measure specific factors loaded onto their designated domain. Methodological artefacts inherent in personality measures (semantical-logical relationships, item wording) tends to yield problems (over-extraction of the number of latent factors, factors solely based on item distribution similarity) when personality inventories are factor analysed at the item level (Hopwood & Donnellan, 2010; Vassend & Skrondal, 1997). Consequently, conventional criteria for adequate factor structures (factor loadings > .40 on the intended factors, non-significant cross-loadings) are rarely met. Thus, the facet-level factor analysis aimed at a more robust test of the above-described HEXACO-PI-Rs theoretical factor structure and determining whether this structure offers the most parsimonious and complete model of the covariance in HEXACO-PI-R self-ratings. Principal axis factoring followed by both varimax (orthogonal) was applied to the data because the latent factor correlations and correlations of domain scores were small (see below).
On deciding the number of latent factors to be extracted, it is generally recommended to apply several criteria. For the item-level analysis, only the scree-test was applied because other criteria indicated a substantially higher number of latent factors than was deemed theoretically meaningful. For the facet-level analysis, the scree-test (DeVellis, 2012) and parallel analysis with raw data permutation (O’connor, 2000) were used in combination with the following fit statistics generated in R (Revelle, 2018): Tucker Lewis Index (TLI) and Root-Mean-Square Error of Approximation (RMSEA). In knowledge of the discrepancy concerning which value of fit statistics that can be regarded as indicating an adequate model fit (Lance, Butts, & Michels, 2006) and previous EFA studies on personality inventories (Hopwood & Donnellan, 2010), criteria on the more liberal side were used with a TLI of > .90 and RMSEA < .07, indicating acceptable fit (Hooper, Coughlan, & Mullen, 2008; Hopwood & Donnellan, 2010). For model comparisons (comparing solutions with a different number of latent factors), ΔTLI values > .01 imply model dissimilarity; a ΔRMSEA of > .010 indicates poorest fit for the model with the highest RMSEA (Chen, 2007).
To examine the outlined relationships between the HEXACO factors/facets and the B5 factors, Pearson correlation coefficients (r) were calculated. Finally, because several HEXACO factors and B5 factors are related to more than one factor in the other model, a simultaneous multiple regression analysis was performed to explicitly show the total variance in the B5 predicted by the HEXACO factors and vice versa. Factor scores from each model were used to predict the domains of the other model. To avoid bias due to different numbers of items in the HEXACO-PI-R (100 items) and the BFI (44 items), the HEXACO-60 (Ashton & Lee, 2009) was used in the multiple regression analysis. Except for the EFA fit statistics, all statistical analyses were carried out in IBM SPSS 25.
Results
Missing values
Four HEXACO-PI-R questionnaires and 17 BFI questionnaires were removed from further analysis due to more than 10% missing values. For the remaining questionnaires, missing values were replaced with series means. The sample size in the reliability analysis and in the factor analysis of the HEXACO-PI-R is therefore n = 480, whereas the sample size for the correlation analysis is n = 467.
Psychometric properties
Internal consistency (Cronbach’s alpha), means, and standard deviations for the six factor scales and 25 facet scales are shown in Table 1. Scores on all scales were normally distributed as indicated by skewness and kurtoses values. Means are close-to-scale midpoints and mostly situated between 3 and 4. Descriptive statistics is fairly similar to the original English version of the HEXACO-PI-R (Lee & Ashton, 2016) and other translations (e.g., Romero, Villar, & López-Romero, 2015).
The internal consistency of the factor scales ranged from α = .81 for Emotionality to α = .77 for Extraversion (mean α = .80). The average internal consistency for the 25 facet scales was α = .63 (median α = .64), ranging from α = .46 for Unconventionality and Altruism to α = .81 for Greed Avoidance. According to conventional criteria for evaluating scale reliability (DeVellis, 2012), all factor scales showed good to acceptable internal consistency, whereas five facet scales had questionable reliability. Unconventionality and Altruism showed unacceptable internal consistency. (See the Discussion section for possible reasons for the low internal reliability of the facet scales.)
Factor analysis
Factorability – Initially, the factorability of the HEXACO-PI-R items was assessed with several criteria. The vast majority of items correlated to at least > .30 with minimum one other variable, suggesting reasonable factorability. The KMO Measure of Sampling Adequacy was .765, above the recommended value of .6. Bartlett’s Test of Sphericity was significant (χ2 (4,950) = 16,025, p < .00) which together with a 5:1 respondent-to-variable ratio indicated that the data was suitable for factor analysis.
Item-level factor analysis – The 100 items of the HEXACO-PI-R were factor analysed with principal axis factoring followed by varimax rotation. The first 15 unrotated eigenvalues were 7.70, 5.46, 5.11, 4.31, 3.96, 2.88, 2.67, 2.19, 1.97, 1.89, 1.80, 1.69, 1.56, 1.55, and 1.50. The flattening of the eigenvalue curve occurred after the seventh factor, indicating that as many as seven factors could be retained. The seventh factor comprised items from Social Self-Esteem (X), Anxiety (negative factor loadings) (E), and Liveliness (X), whereas other items in the seven-factor solution loaded mainly in accordance with the HEXACO factorial structure. The seven-factor solution explained 32.1% of the total variance in the self-ratings.
When six-factors were derived to examine the theoretical structure of the inventory, a factor loading pattern in accordance with the HEXACO model emerged. Ninety-two of 96 items had the highest factor loading on the factor for which they were initially targeted. Eighty-four items loaded > .30 onto their designated dimension. Only item 76 (“I sometimes feel that I am a worthless person”) (X) had its highest and significant factor loading onto an unintended dimension (r = –.39 on Emotionality). Nevertheless, it loaded significantly onto Extraversion as expected (r = .36). Only two other significant cross-loadings were observed. Specifically, items designated to the Agreeableness and Honesty-Humility factors loaded onto separate factors, and cross-loadings of Honesty-Humility items on the Agreeableness factor and vice versa were negligible. Thus, the six-factor structure of the inventory, with Honesty-Humility and Agreeableness constituting distinctive factors, was generally reproducible. Altruism items loaded primarily onto Honesty-Humility and/or Emotionality. The first six factors collectively explained 29.4% of the total variance in the items.
Facet-level factor analysis – Turning to the facet-level factor analysis, the scores of the 25 facets were subjected to principal axis factoring followed by varimax rotation which yielded the following first 10 unrotated eigenvalues: 3.68, 2.66, 2.40, 2.08, 1.88, 1.34, 1.09, 0.91, 0.81, and 0.76. On deciding on the number of factors to be retained, the parallel analysis indicated that eight factors should be extracted. As seen in Table 1, the RMSEA was acceptable in a five-factor solution, whereas CFI suggested that seven or eight factors should be retained. However, ΔRMSEA and ΔTLI improved significantly from the five-factor model onward to the eight-factor model. The scree-test showed the clearest break after the fifth eigenvalue and a relatively smaller break after the sixth factor. This tendency indicated that it was appropriate to extract five or six latent factors. Because the criteria were inconclusive, rotated factor matrices for eight to five latent factors were inspected and are described in the following paragraphs.
In the eight-factor solution, no facet had the primary factor loading onto the eighth dimension. Only Prudence had significant (secondary) loading onto the eighth factor. Thus, this dimension showed insufficient numbers of factor loadings to be considered substantial. As seen in Table A1 in the Appendix, when seven factors were retained (explained variance 32.1%), the seventh factor, as in the item-level analysis, represented a mixture of the Social Self-Esteem, Anxiety, and Liveliness facets. Social Self-Esteem loaded strongly onto this factor (.76) whereas loadings for Anxiety and Liveliness were moderate and secondary to their loadings onto the dimensions defined by other Emotionality and Extraversion facets, respectively. As in the item-level factor analysis, the remaining facets loaded generally in accordance with the HEXACO structure. This outcome shows that seven meaningful factors could be extracted.
When six factors were requested (see Table 2), a factor pattern in accordance with the HEXACO-factorial structure was seen. All facets had their primary factor loadings (> .40) on their intended factor, and there were low secondary loadings (< .30) on other factors. Exceptions were Liveliness and Social Self-Esteem, with secondary loadings on Conscientiousness (.34 and .32, respectively) and Anxiety, with a small cross-loading on Agreeableness (–.31). Altruism loaded primarily onto Emotionality (.42) and secondarily onto Honesty-Humility, Agreeableness, and Conscientiousness. The first six factors collectively explained 56.2% percent of the total variance in the facets.
In a five-factor solution (see Table A2 Appendix), which explained 50.8% of the variance in the facets scores, facets within the Extraversion, Emotionality, Conscientiousness, and Openness dimensions loaded primarily onto one individual factor, whereas facets of the Agreeableness and Honesty-Humility dimensions showed the highest factor loadings on the same dimension. A comparison of the factor loadings showed that the variance in the Honesty-Humility facets was less explained by this dimension than the variance in Agreeableness facets was. The factor loadings on this dimension were for Agreeableness facets larger (ranging from .59 for Gentleness to .52 for Flexibility) than those for the Honesty-Humility facets (from .41 for Sincerity to .37 for Greed Avoidance). Moreover, in the five-factor solution, communalities for the Honesty-Humility facets ranged from .16 for Greed Avoidance to .31 for Fairness (see Table A2), indicating that variance in these facets was not very well accounted for by only five latent factors. As seen in Table 2, the communalities were clearly higher when six versus five factors were retained: .35 for Sincerity, .38 for Fairness, .36 for Greed Avoidance, and .34 for Modesty. The difference in communalities between the six-factor and five-factor solution for the other facets were smaller than for the Honesty-Humility facets. Thus, specifically more of the variance in the Honesty-Humility facets was accounted for by the six-factor solution. As seen in Table 1, the necessity of at least extracting six factors was also supported by significantly better goodness-of-fit statistics for ΔTLI for the six-factor solution than for the five-factor solution. The RMSEA approached a significant difference (ΔRMSEA(20) = .009, ns.).
Inter-correlations among the HEXACO factor scales
To examine the independence of the HEXACO factor scales, inter-correlations among the factor scale scores were computed. (See Table 3.) These correlations were low, indicating that relatively little overlap exists between the factors. The strongest relationships were seen between Honesty-Humility and Agreeableness (r = .29), Extraversion and Conscientiousness (r = .24), and Honesty-Humility and Conscientiousness (r = .23). Similarly, small latent factor correlations occurred when an oblique rotation (Promax (kappa = 4)) was performed on the 25 HEXACO facets. Honesty-Humility correlated .44 with Agreeableness, Extraversion correlated .32 with Conscientiousness, and Honesty-Humility correlated with .31 Conscientiousness. In sum, these analyses supported the independence of the HEXACO factor scales.
Relationships between the HEXACO and the BFI
To investigate convergent validity for the HEXACO-PI-R, bivariate correlations between the HEXACO factor and facet scales and the BFI factor scales were computed. As seen in Table 4, HEXACO Extraversion correlated r = .68 with BFI Extraversion, HEXACO Conscientiousness correlated r = .76 with B5 Conscientiousness, and HEXACO Openness correlated r =.73 with BFI Openness. Thus, HEXACO Extraversion, Conscientiousness, and Openness dimension corresponded to their BFI counterparts (hypothesis 1). As expected (hypothesis 2 and 3), somewhat weaker, albeit strong, relationships were seen between HEXACO Agreeableness and BFI Agreeableness (r = .59) and between HEXACO Emotionality and BFI Neuroticism (r = .61). HEXACO Agreeableness showed a small negative association with Neuroticism (r = –.25), which is in accordance with the two dimensions’ shared anger-related content (hypothesis 2). The expected facet-level associations between HEXACO Emotionality facets and BFI Agreeableness and Neuroticism (hypothesis 3) were seen. In particular, Sentimentality and Dependence correlated more weakly with Neuroticism than Anxiety and Fearfulness and showed a small relationship with BFI Agreeableness (r = .12). Supporting hypothesis 4, Honesty-Humility correlated modestly (r = .37) with BFI Agreeableness and had the lowest correlations of all the HEXACO domains with any of the B5 factors. The correlation between Honesty-Humility facets and the BFI Agreeableness dimension ranged from r = .35 for Fairness to r = .16 for Greed Avoidance. It is worth nothing that BFI Conscientiousness had nearly the same relationship strength with Honesty-Humility (r = .30) as BFI Agreeableness, with Fairness correlating .35 with BFI Conscientiousness. HEXACO Extraversion had a moderate negative correlation with BFI Neuroticism (r = –.44), mainly attributed to the facets Social Self-Esteem and Liveliness.
Finally, to examine if there is variance in the HEXACO model that cannot be explained by the factors in B5 and vice versa, a simultaneous regression analysis was conducted using the factors from each model to predict the domains of the other model. As noted above, in order to avoid bias due to different numbers of items in the HEXACO-PI-R (100 items) and BFI (44 items), the HEXACO-60 (Ashton & Lee, 2009) was used in the regression analysis. As seen in Table 5, the HEXACO domains explained from 47% to 60% of the BFI domains, with an average of 52%, whereas the B5 factors explained, on average, 45% of the variance in the HEXACO factors, with a range spanning 20% to 57%. The B5 factors predicted 20% of the variance in Honesty-Humility.
Discussion
The aim of this study was to validate the HEXACO model in a Norwegian sample by examining the psychometric properties, factor structure, and convergent validity of the Norwegian version of the 100-HEXACO-PI-R.
Internal consistency for the factor scales was good and in line with reliability coefficients obtained in other studies (Babarović & Šverko, 2013; Baiocco et al., 2017; Mededovic, Colovic, Dinic, & Smederevac, 2017; Romero et al., 2015). Scale reliability at the facet level was generally satisfactory but seven facet scales showed questionable internal consistency. However, the average facet scale reliability is similar to other translated versions of the inventory (Babarović & Šverko, 2013; Boies et al., 2004; Mededovic et al., 2017; Romero et al., 2015), and low reliability for some facets does not seem to be an attribute solely of the Norwegian translation. Rather, it appears to be an inherent feature of the HEXACO-PI-R caused by brief scales with broad content (Lee & Ashton, 2016). The relevance of alpha reliability in evaluating the psychometric properties of brief personality scales is debated, with research showing that scales can have acceptable test-retest reliability and validity in spite of low alpha levels (de Vries, 2013; McCrae, Kurtz, Yamagata, & Terracciano, 2011). Lee and Ashton (2016) recommend that researchers who examine associations between the HEXACO-PI-R facet scales and external criteria check the item-level correlations with those criteria to ensure that the facet-level associations are not due to the variance of a particular item.
Nonetheless, wording of some items might need modifications. For instance, more than half of the participants responded “neither agree nor disagree” to the Unconventionality items “I think of myself as a somewhat eccentric person” and “I think that paying attention to radical ideas is a waste of time.” When the author gave students feedback on the results, he also discussed potential reasons as to why a high percentage of them did not identify with these particular items. They pointed out that “eccentric” and “radical” are, to some extent, foreign words (with negative connotations) and that the meaning of these items was somewhat unclear. Compared to other samples (Lee & Ashton, 2016; Romero et al., 2015), the Norwegian sample seems particularly high on the Altruism dimension with more than 85% responding either disagree or strongly disagree to Item 100 (“People see me as a hard-hearted person”). This inclination leads to low variance and hence might contribute to the particularly low reliability for the Altruism facet. However, the Altruism facet generally shows low alpha reliability (Lee & Ashton, 2016; Romero et al., 2015).
As for the internal structure of the inventory, seven meaningful latent factors could be extracted in the factor analysis, with the seventh factor representing a mixture of items or facets from Social Self-Esteem, Liveliness, and Anxiety. In the English HEXACO-PI-R validation study, a small seventh factor in the item-level factor analysis was also found, but it corresponded with the item direction of keying and was described as an “acquiescence” factor (Lee & Ashton, 2016). Previous validation studies with facet-level factor analysis have consistently supported the theoretical six-factor structure of the inventory (e.g., Mededovic et al., 2017). However, from one sample to another, a factor solution similar to the one in the present study occurs; a factor structure where Extraversion divides into two factors, one of which has Social Self-Esteem and Liveliness domains and maybe some loadings for Anxiety (Michael C. Ashton, personal communication, 2nd of November 2018). Presently, it seems that the factor solution in this study must be considered sample specific. No consistent factors beyond the HEXACO/Big Six are repeatedly recovered in lexical studies (Saucier, 2009).
When six latent factors were requested, a factor loading pattern in line with the HEXACO model’s expectations was seen. In the facet-level factor analysis, all facets had their highest factor loading on the intended dimension and there were few cross-loadings. Total explained variance of the facets and the factor loadings was similar to other studies with principal axis factoring (Mededovic et al., 2017; Wakabayashi, 2014). Although the scree plot and the RMSEA suggested that variance in facets could perhaps be summarised by five latent variables, variance for Honesty-Humility facets specifically was better represented in a six- rather than a five-factor solution. Thus, extraction of six factors was necessary to adequately capture the variance in all facets, which also was supported by better EFA fit statistics for the six-factor solution than for the five-factor solution.
The item-level factor analysis, with a clean structure and the majority of items showing the highest and most significant loading on their target factor, gave further support to the inventory’s factor structure. Because single items in psychological research rarely meet statistical requirements for factor analysis, a few nonsignificant or anomalous loadings must be expected (Bernstein & Teng, 1989; Wirth & Edwards, 2007). In the present study, scores for items with nonsignificant factor loadings tended to be skewed or have excess kurtosis.
Inter-correlations between factor scales and latent factors were low, which indicates that the factors represent distinct constructs. No factor scales shared more than 6% variance. Similar small overlap in the HEXACO factors are found in other studies (Babarović & Šverko, 2013; Mededovic et al., 2017; Romero et al., 2015) and were also seen among the B5 factors. For example, Neuroticism correlated –.37 and –.24 with BFI Conscientiousness and BFI Extraversion, respectively. Thus, results supported the independence of the HEXACO factors.
Relationships between HEXACO and B5 were generally in line with theoretical assumptions (Ashton et al., 2014). HEXACO Extraversion, Conscientiousness, and Openness dimensions were strongly related to their B5 counterparts. Somewhat weaker yet still strong relationships were found between HEXACO Emotionality and Neuroticism and between HEXACO Agreeableness and B5 Agreeableness. Although these connections were higher than in other studies (Babarović & Šverko, 2013; Boies et al., 2004), the conceptual difference between Emotionality and Neuroticism was supported with only Anxiety (Emotionality facet) correlating strongly with Neuroticism. Contrarily, Dependence and Sentimentality had weak relationships with Neuroticism. Furthermore, HEXACO Agreeableness correlated significantly with Neuroticism, which is in line with the relocation of anger dispositions from Neuroticism in the Big Five onto Agreeableness in the HEXACO model. Finally, the Honesty-Humility domain showed the lowest correlation of the HEXACO factors with any of the B5 domains, and only 20% of the variance in Honesty-Humility could be explained by the B5 factors. In comparison, the explained variance for the other HEXACO factors with the B5 ranged from 42% to 57%. Thus, Honesty-Humility is not fully captured by the B5 as measured with the BFI. Also in line with expectations, the HEXACO model predicted somewhat more of the variance in the B5 than vice versa.
Although not explicitly outlined in the theoretical description of the HEXACO model, a significant negative relationship between HEXACO Extraversion and Neuroticism has been obtained in several studies (Gaughan et al., 2012; Mededovic et al., 2017; Romero et al., 2015) and was also found in the present study. This connection seems meaningful because Extraversion measures positive self-regard in social contexts (Social Self-Esteem) along with activity level and enthusiasm (Liveliness), which are clearly opposites of self-consciousness and nervousness in social situations and of depression and lethargy (Neuroticism). Another correlation that occurs repeatedly across studies (Mededovic et al., 2017; Romero et al., 2015) and was seen in the present study is a positive relationship between Honesty-Humility and B5 Conscientiousness. Arguably, at a lower level, impulse control is involved in resisting bribes and avoiding corruption (Honesty-Humility). These characteristics bear resemblance with the core of the B5 Conscientiousness dimension, which is characterised by personality adjectives such as reliable, thorough, and self-disciplined (Saucier & Goldberg, 1996).
The Norwegian predictive validity needs to be examined further.
In sum, this study shows that the Norwegian 100-HEXACO-PI-R is a valid measure for assessing the HEXACO dimensions. However, more studies with heterogeneous samples that specifically examine the Norwegian HEXACO-PI-R’s factor structure, predictive validity, and temporal stability are needed to fully establish the inventory’s psychometric properties.
References
Ashton, M. C., & Lee, K. (2005a). A defence of the lexical approach to the study of personality structure. European Journal of Personality, 19(1), 5–24. doi:10.1002/per.541
Ashton, M. C., & Lee, K. (2005b). Honesty‐Humility, the Big Five, and the Five‐Factor Model. Journal of Personality, 73(5), 1321–1354. doi:10.1111/j.1467-6494.2005.00351.x
Ashton, M. C., & Lee, K. (2007). Empirical, theoretical, and practical advantages of the HEXACO model of personality structure. Personality and Social Psychology Review, 11(2), 150–166. doi:10.1177/1088868306294907
Ashton, M. C., & Lee, K. (2008). The prediction of Honesty–Humility-related criteria by the HEXACO and Five-Factor Models of personality. Journal of Research in Personality, 42(5), 1216–1228. doi:10.1016/j.jrp.2008.03.006
Ashton, M. C., & Lee, K. (2009). The HEXACO-60: A short measure of the major dimensions of personality. Journal of Personality Assessment, 91(4), 340–345. doi:10.1080/00223890902935878
Ashton, M. C., & Lee, K. (2018). How well do Big Five measures capture HEXACO scale variance? Journal of Personality Assessment, 1–7. doi:10.1080/00223891.2018.1448986
Ashton, M. C., Lee, K., & de Vries, R. E. (2014). The HEXACO Honesty-Humility, Agreeableness, and Emotionality factors: A review of research and theory. Personality and Social Psychology Review, 18(2), 139–152. doi:10.1177/1088868314523838
Ashton, M. C., Lee, K., & Goldberg, L. R. (2004). A hierarchical analysis of 1,710 English personality-descriptive adjectives. Journal of Personality and Social Psycholology, 87(5), 707–721. doi:10.1037/0022-3514.87.5.707
Ashton, M. C., Lee, K., Marcus, B., & Vries, R. E. D. (2007). German lexical personality factors: Relations with the HEXACO model. European Journal of Personality, 21(1), 23–43. doi:10.1002/per.597
Ashton, M. C., Lee, K., Perugini, M., Szarota, P., de Vries, R. E., Di Blas, L., . . . De Raad, B. (2004). A six-factor structure of personality-descriptive adjectives: Solutions from psycholexical studies in seven languages. Journal of Personality and Social Psychology, 86(2), 356-366. doi:10.1037/0022-3514.86.2.356
Babarović, T., & Šverko, I. (2013). The HEXACO personality domains in the Croatian sample. Društvena istraživanja: časopis za opća društvena pitanja, 22(3). doi:10.5559/di.22.3.01
Baiocco, R., Chirumbolo, A., Bianchi, D., Ioverno, S., Morelli, M., & Nappa, M. R. (2017). How HEXACO personality traits predict different selfie-posting behaviors among adolescents and young adults. Frontiers in Psychology, 7, 2080. doi:10.3389/fpsyg.2016.02080
Bernstein, I. H., & Teng, G. (1989). Factoring items and factoring scales are different: Spurious evidence for multidimensionality due to item categorization. Psychological Bulletin, 105(3), 467–477. doi:10.1037/0033-2909.105.3.467
Boies, K., Yoo, T. Y., Ebacher, A., Lee, K., & Ashton, M. C. (2004). Psychometric properties of scores on the French and Korean versions of the HEXACO Personality Inventory. Educational and Psychological Measurement, 64(6), 992–1006. doi:10.1177/0013164404267277
Chen, F. F. (2007). Sensitivity of goodness of fit indexes to lack of measurement invariance. Structural Equation Modeling: A Multidisciplinary Journal, 14(3), 464–504. doi:10.1080/10705510701301834
Costa, P. T., McCrae, R. R., & Psychological Assessment Resources Inc. (1992). Revised NEO Personality Inventory (NEO PI-R) and NEO Five-Factor Inventory (NEO-FFI). Odessa, Fla. (P.O. Box 998, Odessa 33556): Psychological Assessment Resources.
de Vries, R. E. (2013). The 24-item Brief HEXACO Inventory (BHI). Journal of Research in Personality, 47(6), 871–880. doi:10.1016/j.jrp.2013.09.003
DeVellis, R. F. (2012). Scale Development: Theory and Applications (Vol. 3). California: SAGE.
Di Blas, L., & Forzi, M. (1998). An alternative taxonomic study of personality-descriptive adjectives in the Italian language. European Journal of Personality, 12(2), 75–101. doi:10.1002/%28SICI%291099-0984%28199803/04%2912:2%3C75::AID-PER288%3E3.0.CO;2-H
Engvik, H., & Føllesdal, H. (2005). The Big Five Inventory på norsk [The Big Five Inventory in Norwegian]. Tidskrift for Norsk Psykologforening, 42, 128–129.
Gaughan, E. T., Miller, J. D., & Lynam, D. R. (2012). Examining the utility of general models of personality in the study of psychopathy: A comparison of the HEXACO-PI-R and NEO PI-R. Journal of Personality Disorders, 26(4), 513–523. doi:10.1521/pedi.2012.26.4.513
Goldberg, L. R. (1990). An alternative «description of personality»: The big-five factor structure. Journal of Personality and Social Psychology, 59(6), 1216–1229. doi:10.1037/0022-3514.59.6.1216
Goldberg, L. R. (1993). The structure of phenotypic personality traits. American Psychologist, 48(1), 26. doi:10.1037/0003-066X.48.1.26
Gundhus, C. (2018). Exploring employee uncertainty in organisational change: Personality, trust and leader communication style. Retrieved from https://brage.bibsys.no/xmlui/bitstream/
Hooper, D., Coughlan, J., & Mullen, M. (2008). Structural equation modelling: Guidelines for determining model fit. Electronic Journal of Business Research Methods, 6(1), 53–60.
Hopwood, C. J., & Donnellan, M. B. (2010). How should the internal structure of personality inventories be evaluated? Personality and Social Psychology Review, 14(3), 332–346. doi:10.1177/1088868310361240
John, O. P., Angleitner, A., & Ostendorf, F. (1988). The lexical approach to personality: A historical review of trait taxonomic research. European Journal of Personality, 2(3), 171–203. doi:10.1002/per.2410020302
John, O. P., Naumann, L. P., & Soto, C. J. (2008). Paradigm shift to the integrative Big Five trait taxonomy: History, measurement, and conceptual issues. In O. P. John, R. W. Robins, & L. A. Pervin (Eds.), Handbook of Personality Theory and Research (Vol. 3). New York: Guilford Press.
Lance, C. E., Butts, M. M., & Michels, L. C. (2006). The sources of four commonly reported cutoff criteria: What did they really say? Organizational Research Methods, 9(2), 202–220. doi:10.1177/1094428105284919
Lee, K., & Ashton, M. C. (2008). The HEXACO personality factors in the indigenous personality lexicons of English and 11 other languages. Journal of Personality, 76(5), 1001–1054. doi:10.1111/j.1467-6494.2008.00512.x
Lee, K., & Ashton, M. C. (2009). Reanalysis of the structure of the Greek personality lexicon. Journal of Cross-Cultural Psychology, 40(4), 693–700. doi:10.1177/0022022109335183
Lee, K., & Ashton, M. (2016). Psychometric properties of the HEXACO-100. Assessment, 25(5), 543–556. doi:10.1177/1073191116659134
Martinsen, Ø. L., Nordvik, H., & Østbø, L. E. (2005). Norske versjoner av NEO PI-R og NEO FFI [Norwegian versions of NEO PI-R and NEO FFI]. Tidskrift for Norsk Psykologforening, 42(5), 421–423.
McCrae, R. R., & Costa, P. T. (2008). The five-factor theory of personality. In O. P. John, R. W. Robins, & L. A. Pervin (Eds.), Handbook of Personality Theory and Research (Vol. 3, pp. 159–181). New York: Guilford Press.
McCrae, R. R., & John, O. P. (1992). An introduction to the five-factor model and its applications. Journal of Personality, 60(2), 175–215. doi:https://doi.org/10.1111/j.1467-6494.1992.tb00970.x
McCrae, R. R., Kurtz, J. E., Yamagata, S., & Terracciano, A. (2011). Internal consistency, retest reliability, and their implications for personality scale validity. Personality and Social Psychology Review, 15(1), 28–50. doi:10.1177/1088868310366253
Mededovic, J., Colovic, P., Dinic, B. M., & Smederevac, S. (2017). The HEXACO personality inventory: Validation and psychometric properties in the Serbian language. Journal of Personality Assessment, 1–7. doi:10.1080/00223891.2017.1370426
O’Connor, B. P. (2000). SPSS and SAS programs for determining the number of components using parallel analysis and Velicer’s MAP test. Behavior Research Methods, Instruments, & Computers, 32(3), 396–402. doi:10.3758/bf03200807
Revelle, W. (2018). Psych: Procedures for Personality and Psychological Research (Version 1.8.4). Evanston, Illinois, USA: Northwestern University. Retrieved from https://CRAN.R-project.org/package=psych Version = 1.8.4.
Romero, E., Villar, P., & López-Romero, L. (2015). Assessing six factors in Spain: Validation of the HEXACO-100 in relation to the Five Factor Model and other conceptually relevant criteria. Personality and Individual Differences, 76, 75–81. doi:10.1016/j.paid.2014.11.056
Saucier, G. (2009). Recurrent personality dimensions in inclusive lexical studies: Indications for a Big Six structure. Journal of Personality, 77(5), 1577–1614. doi:10.1111/j.1467-6494.2009.00593.x
Saucier, G., & Goldberg, L. R. (1996). Evidence for the Big Five in analyses of familiar English personality adjectives. European Journal of Personality, 10(1), 61–77. doi:10.1002/%28SICI%291099-0984%28199603%2910:1%3C61::AID-PER246%3E3.0.CO;2-D
Szirmák, Z., & Raad, B. D. (1994). Taxonomy and structure of Hungarian personality traits. European Journal of Personality, 8(2), 95–117. doi:10.1002/per.2410080203
Vassend, O., & Skrondal, A. (1997). Validation of the NEO Personality Inventory and the five-factor model. Can findings from exploratory and confirmatory factor analysis be reconciled? European Journal of Personality, 11(2), 147–166. doi:10.1002/%28SICI%291099-0984%28199706%2911:2%3C147::AID-PER278%3E3.0.CO;2-E
Wakabayashi, A. (2014). A sixth personality domain that is independent of the Big Five domains: The psychometric properties of the HEXACO Personality Inventory in a Japanese sample. Japanese Psychological Research, 56(3), 211–223. doi:10.1111/jpr.12045
Wasti, S. A., Lee, K., Ashton, M. C., & Somer, O. (2008). Six Turkish personality factors and the HEXACO model of personality structure. Journal of Cross-Cultural Psychology, 39(6), 665–684. doi:10.1177/0022022108323783
Wirth, R. J., & Edwards, M. C. (2007). Item factor analysis: Current approaches and future directions. Psychological Methods, 12(1), 58–79. doi:10.1037/1082-989X.12.1.58
Appendix
