While the Big Five descriptive personality model, first based on English language studies, had been replicated across several other languages, including Dutch, German, French, and Turkish (Saucier & Goldberg, 2006; Goldberg & Somer, 2000), other studies also began to emerge that questioned the universality of the Big Five model. Hungarian (Szirmak & De Raad, 1994) and Italian (Di Blas & Forzi, 1998) studies first pointed to a six-factor personality structure. Later, a paper by Michael Ashton et al. (2004) consolidated evidence for a six-factor model in seven different languages: In addition to the Hungarian and Italian work, Dutch, French, German, Korean, and Polish studies also found six factors.
The major difference between Ashton’s six-factor model and the Big Five (summarized in Table 13.1) was the emergence of a sixth factor they labeled Honesty/Humility, which later provided the “H” in their six-factor HEXACO Personality Inventory (Lee & Ashton, 2012). The Honesty/Humility dimension was basically Conscientiousness with morality features. In fact, Ashton suggested that Morality could be an alternate name for this extra factor (Ashton et al., 2004). The new Honesty/Humility factor included adjectives such as honest/dishonest and truthful/untruthful, among others, whereas the Conscientiousness factor retained adjectives such as orderly, careful, and disciplined versus careless and negligent (Ashton et al., 2004).
Goldberg had previously included the adjectives honest, moral, and truthful in an Honesty/Morality cluster of adjectives, along with dishonest in a Deceit cluster, all of which were derived from Warren Norman’s earlier work. Yet, these clusters did not load especially strongly in Goldberg’s reported Big Five factor structure (Goldberg, 1990). Furthermore, Goldberg did not include any of these key adjectives in his lists of one hundred Big Five unipolar or fifty bipolar adjective markers (Goldberg, 1992). The adjective moral was included in the 540 adjectives used in the classic paper on adjectives that statistically “blended” qualities of more than one Big Five dimension, but moral loaded only weakly as a blend of Agreeableness and Conscientiousness (Hofstee et al., 1992). By contrast, honest was the most frequent high-loading Honesty/Humility adjective in the studies cited by Ashton et al. (2004), appearing in six of the seven translations, truthful/untruthful appeared in four of the seven translations they studied.
Overall, Goldberg included less than 40 percent of the Honesty/Humility adjectives typically appearing in the translations cited by Ashton et al. (2004). It is possible that the increased sampling from the Honesty/Humility space in these other studies was crucial for the emergence of this new factor. Indeed, this paucity of Honesty/Humility adjectives in Goldberg’s Big Five work, compared to Ashton’s, is reminiscent of Jack Block’s (1995) critique of the Big Five, in which he noted that changing the variables in a factor analysis is likely to change the factor structure.
This discussion is not intended to be critical of Ashton et al.’s (2004) fine reanalysis and synthesis of the personality reports covering these seven languages. It seems very likely that honesty and humility are human traits that are valued in many different cultures. However, we are trying to illustrate the kinds of issues that begin appearing when starting with a tertiary top-down as opposed to a primary bottom-up approach to identifying basic personality dimensions.
OTHER “BIG” MODELS AROSE FROM THE SELECTION OF ADJECTIVES
Adding to this personality dimension conundrum, as mentioned previously, is a competing six-factor model offered by Gerard Saucier (2009) that emerged from an even different set of adjective descriptor studies (see Table 13.1). But before delving further, a short digression is needed to explain more about how different researchers have selected the adjectives for lexical personality research.
While some credit Sir Francis Galton as the first person to publish work on the lexical hypothesis (Goldberg, 1990), the systematic categorization of English-language adjectives as personality descriptors really began in earnest with Gordon Allport and Henry Odbert (1936). From the 1925 edition of Webster’s Unabridged New International Dictionary, these investigators compiled a list of 17,953 words that were descriptive of personality or personal behavior. They placed their words alphabetically in four columns, with each column corresponding to a specific category.
Table 13.1. Summary of Four Lexically Derived Personality Models
Goldberg:
Big Five Ashton:
Narrow Big Six Saucier:
Wide Big Six Tellegen & Waller, Goldberg & Somer: Big Seven
Extraversion Extraversion Extraversion Positive Emotionality
Agreeableness Agreeableness (more anger) Agreeableness Agreeableness
Conscientiousness Conscientiousness (not moral) Conscientiousness Dependability
Emotional Stability Emotionality (negative)(less anger) Resiliency versus Internalizing Negative Emotionality Negative Emotionality
Openness to Experience Intellect / Imagination / Unconventional Originality / Talent / Positive Valence Conventionality
__ Honesty / Humility Negative Valence Negative Valence
__ __ __ Positive Valence / Attractiveness
Column I included what they considered to be the “real traits” of personality that described stable and consistent behavior. such as aggressive, introverted, and sociable. Column II contained more “temporary states, current activities, and moods,” such as rejoicing, frantic, and annoyed. Column III was reserved for evaluative words implying “social judgment,” such as worthy, coarse, and crude. Column IV contained words describing “physical appearance as well as personal capacities and talents,” such as slender, pale, gifted, and athletic. While Allport and Odbert clearly felt that the 4,504 adjectives in Column I were the best descriptors for stable personality characteristics, they did not consider their placements absolute and did not discourage others from exploring the other columns for additional words. And herein lies the problem with the lexical hypothesis: Different investigators have used different criteria for selecting the trait terms they choose to study as they rely on factor analysis to extract personality dimensions. In other words, experimenter choices, rather than independent criteria, are guiding ultimate research outcomes. Importantly, the Big Five typically emerges from a narrower set of mostly Column I descriptors.
In his classic 1990 paper cited by over fourteen hundred other publications, Goldberg (1990) wanted to be more inclusive than Cattell and did include many more adjectives. Still, Goldberg had used a narrower set of adjectives, with most of his 1,431 trait terms coming from Allport and Odbert’s neutral Column I. However, in every factor even Goldberg included exceptions, especially in the small fifth factor (Intellect/Culture), in which seven of the twenty-eight words he used as examples were from Column III (social judgment) or Column IV (physical appearance or capacities). The seven non-English-language studies that Ashton et al. (2004) reanalyzed also largely followed narrower trait-term selection criteria but sampled more broadly from the honesty-humility domain. In other words, what you choose to include in your personality assessment has a powerful effect on what you will find. So how does one know what to include?
AND THE PLOT THICKENS
Gerard Saucier’s (2009) alternative six-factor model was drawn from studies in seven other languages (Chinese, English, Filipino, Greek, Hebrew, Spanish, and Turkish) that sampled even more widely from the domain of potential personality attributes. These studies tended to use more trait terms from the temporary states of Column II such as angry, and frustrated, and the social evaluations of Column III, such as charming, dangerous, and disgusting. Saucier (2009) labeled these the “wideband” lexical studies and referred to the group of studies used, for example, by Ashton et al. (2004), as “narrowband” (see Table 13.1)
Saucier (2009) reexamined the eight wideband studies (two were in English) and determined that, while six factors might be optimal for summarizing these studies, the wideband six factors did not completely replicate the narrowband six-factor model reported by Ashton et al. (2004): The main differences were (1) a Negative Valence factor—w
ith adjectives like cruel, corrupt, disgusting, and wicked—replaced Ashton et al.’s Honest/Humility factor, (2) the Emotionality factor from the narrower selection of adjectives had morphed into a Resiliency versus Internalizing Negative Emotionality factor when more evaluative trait terms like cowardly, depressed, fearful, frustrated, gloomy, and sad were added, and (3) the more traditional Intellect/Imagination factor from Ashton et al.’s narrowband six-factor model broadened out to include more positive-valence terms, such as impressive, outstanding, and admirable, with fewer terms like creative. While there were other more minor differences between the two six-factor models, Saucier concluded that the Big Six model would emerge differently depending on whether narrow or wider selection criteria were used to select the trait terms, but that in either form “the Big Six seems a more cross-culturally valid starting point than the Big Five” (2009, p. 1609).
However, Saucier also conceded that neither version of the Big Six represented a universal personality model: “The [six-factor] structure has not appeared identically in all studies in all languages. Its latent pattern can be detected across the 16 lexical studies reviewed here and by Ashton et al. (2004), but this pattern seems prone to have pieces missing in many single studies” (2009, p. 1609). In spite of Saucier’s herculean effort, the language and cultural differences expressed across these lexical studies epitomize the difficulty of identifying universal personality dimensions from the top-down, tertiary, human-language starting point.
BACK TO SCIENTIFIC “BASICS”: NARROWING THE FIELD ONCE MORE
The Turkish paper Saucier reexamined deserves some closer attention. Goldberg and Somer (2000) had started from scratch and broadly selected 2,200 person-descriptive adjectives from Turkish dictionaries. They narrowed their list to 1,300 adjectives by excluding many physical characteristics such as tall and thin, mere evaluations such as good or bad, and special abilities such as good dancer, and then further reduced the list to a manageable 498 by selecting those adjectives with the highest familiarity ratings. They concluded that seven factors best accounted for their 498-adjective data set. Each of the Big Five factors was present: Extraversion, Agreeableness, Conscientious, Emotional Stability, and Intellect (Openness to Experience). However, their set also included a Negative Valence factor and an Attractiveness factor to complete their seven factors. Notably, no Honesty/Humility factor emerged.
However, Goldberg and Somer next simulated a narrower, more restrictive item pool by eliminating terms that described physical appearance, such as eye-catching; one person’s effect on another, such as influential; and words that raters rarely used to describe themselves or others, such as idiotic, immoral, dirty, and uncivilized. The factor analysis of the remaining 440 personality terms yielded a clear Big Five structure, with no additional interpretable factors.
As it turned out, they had replicated work in English that had largely identified the same seven factors from an alternate wider selection of personality descriptors (Tellegen & Waller, 1987; see below), as well as the Big Five when the adjectives were more narrowly restricted. They had also provided evidence that, even in a language as remote from English and German as Turkish, indigenous trait terms can be independently selected and a Big Five personality structure can be derived “when the item pools were restricted to terms that are less pejorative and more clearly related to personality traits” (Goldberg & Somer, 2000, p. 523). Otherwise, additional dimensions such as Negative Valence and Attractiveness are likely to emerge from the more evaluative descriptors.
ONE MORE MODEL: THE BIG SEVEN
But the question remains: Using a top-down approach, how does one know what set of adjectives to use? Tellegen and Waller (1987), opted for a purposely nonrestrictive selection of trait terms originally sourced from The American Heritage Dictionary and reported what they named the Big Seven personality model due to the similarity of five of their dimensions to the Big Five. However, their additional two dimensions were Negative Valence and Positive Valence, the latter replacing the Attractiveness factor found in the Turkish study (Goldberg & Somer, 2000) and in an English language study by Saucier (1997). (This also raises the issue of how to distinguish “valence” and “emotionality” in these lexical studies.)
In another study using Spanish translations of a reduced set of the ten best marker terms from each of the Big Seven factors, Benet and Waller (1995) replicated the Big Seven model in Spanish, as well as across self and peer ratings. They argued that these negative and positive valence dimensions were not statistical or response-set artifacts but were stable measures of self-image that could be observed in nonclinical as well as clinical populations. In clinical samples, borderline or narcissistic personality disorders reflected extreme forms of negative and positive self-evaluations, respectively.
However, a more ambitious attempt to replicate the Big Seven in Hebrew using a nonrestrictive set of 252 trait terms drawn from a Hebrew dictionary (Almagor, Tellegen, & Waller, 1995) was not able to replicate the seven dimensions in their original factor analysis. They were only able to identify seven factors, including Conventionality, after identifying eight Hebrew Conventionality marker terms (erotic, sexual, unruly, liberal, individualist, permissive, naughty, and sensual) and factor analyzing those with the sixty strongest markers from the other six factors, although none of their eight Conventionality markers matched the markers reported by Benet and Waller (1995).
All this may start to sound like a Tower of “Babble” for readers not initiated in academic nitpicking. In any event, Table 13.1 summarizes the four lines of research discussed above and will, hopefully, alleviate any confusion.
In discussing the results of the Greek language study, which as in the Turkish and Hebrew studies started by gleaning descriptive terms from dictionaries, Saucier, Goldberg, and colleagues suggested that “the present results are not very supportive of the cross-cultural generalizability of [personality] structures at the five- to seven-factor level” (Saucier, Georgiades, Tsaousis, & Goldberg, 2005, p. 867) and even went so far as to suggest that while all the various lexical models “are variations within some underlying scheme. It may be that factor analysis does not directly reveal this underlying scheme” (Saucier et al., 2005, p. 870).
DENOUEMENT: PULLING THE THREADS OF THIS SHAKESPEARIAN “RAVELL’D SLEEVE OF CARE”
In a sense the overall results from the many factor analyses is a nightmare. Has the mathematical method that Cattell dreamed would resolve the personality dimensions dilemma run aground on its own glacial scree? Are human languages too far removed from our primary mammalian personality sources? Is it possible that starting from the top-down using a tertiary language approach even modern computers capable of probing immense databases incorporating hundreds of people rating themselves or colleagues on hundreds of trait terms cannot parse universal features of the human BrainMind? Interestingly, after discussing the value of adjective trait terms for describing personality, Allport and Odbert (1936) offered a prescient caveat, suggesting as much:
There is, unfortunately for scientific psychology, a second influence determining our lexicon of trait-names, namely the tendency of each epoch to characterize human qualities in the light of standards and interests peculiar to the times. Historically, the introduction of trait-names can be seen to follow this principle of cultural (not psychological) determination to a striking degree. Presumably human beings through countless ages had displayed such qualities as devotion, pity, and patience, but these terms were not established with their present meanings until the Church made of them recognized and articulated Christian virtues. (p. 2)
While they had focused exclusively on potential issues with the English lexicon, this quote strongly suggests there would be even greater problems attempting to establish lexical universals across a wide variety of cultures and languages, which is exactly what seems to be playing out.
By contrast, the bottom-up cross-species approach has provided an elegant parsing of evolutionarily adaptive emotional instinc
ts that are common to all mammals. The BrainMind allows relatively crude subcortical prods (deep brain stimulation) to inform us what well-organized adaptive systems it contains. The bottom-up DBS approach, which comports well with the natural emotional behaviors we see in the real world, supplies us a view of the most basic affective personality spaces, which is not only a real-life starting point for all humans but also remains visible and accessible throughout the human life-span, even as language and culture shape personality into more complex tertiary spaces, heavily influenced by individual learning and culture. We must accept that such complex tertiary personality spaces come to be strongly influenced by cultural learning, which also exerts abundant top-down cortical inhibition and arousal of emotions and behavior, especially in humans. This is what developmental maturation is all about.
How you behave and respond to others has enormous influences on your success as an individual—both reproductive and cultural. However, our comprehension of that tertiary complexity is likely to be limited profoundly until we appreciate the foundational importance of primary-emotional and other affective spaces in navigating interpersonal affairs. Might it not be that the emotional affects need to be understood first to gain a lasting and coherent vision of what it means to have a distinct personality? As these affective powers, along with influences arising from our bodily homeostatic needs and sensory feelings (perhaps comparatively minor in rich life-supportive environments), guide our progress, might we see a more coherent foundation for our personality structures? The major goal of this book is to share that affirmative vision, which can easily work across species, and even inform (and be informed by) studies of our genetic nature. In so doing, we may better see the deep structure of the Big Five.
The Emotional Foundations of Personality Page 27