Seven-Factor Model of Personality
KeywordsPersonality Disorder Borderline Personality Disorder Negative Valence Incremental Validity Positive Valence
The seven factor model of personality was developed by Tellegen and Waller (1987) using the lexical approach and represents personality traits in terms of seven broad dimensions including positive emotionality, negative emotionality, dependability, agreeability, conventionality, positive valence, and negative valence.
Within the past few decades, personality psychologists have generally reached consensus regarding the structure of personality traits. Despite a multitude of trait models ranging from as few as one (Musek 2007) or two (Block and Block 1980; Digman 1997) traits to as many as 20 (Gough 1987), evidence for a five factor model is robust. Indeed, there is evidence that the “Big Five” traits subsumed in this model – extraversion/surgency, agreeableness, conscientiousness, emotional stability (vs. neuroticism), and culture or openness to experience – are generalizable across numerous populations and settings (Marsh et al. 2013; McCrae and Costa 2003; McCrae et al. 2005). Despite this consensus, as noted, there are alternative models that are arguably superior in some ways. One alternative is the Big Seven model, which may be especially relevant for psychopathology assessment and research. In this entry, we will describe the Big Seven and the origins of the model, review empirical evidence supporting the model, and review criticisms of the model. Lastly, we will focus on its potential role in psychopathology research, exploring how it can expand our understanding of mental disorders.
Description of the Seven Factor Model
In order to appreciate how the Big Five and Seven differ, it is necessary to revisit the origins and “discovery” of the Big Five. The Big Five were born out of the lexical tradition, which rests on the assumption that salient and socially relevant individual differences are encoded into the natural language (John et al. 1988; Goldberg 1990). Allport and Odbert (1936) identified nearly 18,000 descriptive terms amassed from the English language dictionary and categorized the terms into one of four categories including stable traits, temporary moods, social evaluations, and a miscellaneous category encompassing physical characteristics and talents, among other things. Only those in the first category, the stable traits, were further considered and subsequently factor analyzed, and these factor analyses resulted in the five replicable factors we now refer to as the Big Five (Cattell 1943; Goldberg 1990; Norman 1967; Tupes and Christal 1961; see John et al. 1988, for a complete historical account). Notably, the other types of descriptors, such as social evaluations, were excluded. This exclusion has prompted some critics of the five factor model to argue that the model is not comprehensive enough to appreciate the true complexity of personality (Block 1995).
Tellegen and Waller (1987) are among those who believe this assertion. Tellegen (1993) argued that it is critical to include evaluative terms, characteristics to which one ascribes a desirability or value, such as “evil” (a moral judgment of negative quality) or “unworthy” (a devaluation of self). He emphasized that neglecting evaluative terms results in elimination of key individual differences in personality, such as self-esteem. Tellegen further argued that early researchers’ choice to exclude affective terms results in “a systematic underrepresentation of potential emotional-temperament-related trait descriptors” (Tellegen 1993, p. 124). In an effort to remedy this, Tellegen and Waller (1987) used a similar natural language approach to identifying the major personality traits represented in the lexicon, but they used less stringent criteria when selecting words, including stable trait terms in addition to previously neglected terms falling outside this realm, such as evaluative terms. Using a stratified sampling method, 400 personality descriptors were chosen from the dictionary. When self-report ratings of these terms were organized by factor analysis, seven broad personality dimensions emerged, which were named the “Big Seven.” The first five factors (positive emotionality, negatively emotionality, dependability, agreeability, and conventionality) were largely similar to the Big Five dimensions, though there were some notable differences. As an example, positive and negative emotionality were similar to extraversion and neuroticism, respectively, but differed in that they appeared to be more emotional-temperament dimensions. The final two factors were named positive and negative valence. Waller (1999) described the positive valence factor as measuring “a sense of self-worth and personal value at moderate levels and a grandiose sense of self-importance and specialness at the upper extreme” (p. 170), exemplified by adjectives such as “special” or “excellent.” Negative valence has been defined as “self-perceptions of evilness or awfulness” (Waller 1999, p. 170) and is associated with words such as “wicked” and “terrible.” On the heels of this research, Tellegen and colleagues created an inventory designed to measure the Big Seven, the Inventory of Personal Characteristics #7 (IPC-7; Tellegen et al. 1991).
Support for the Big Seven Model
Considerable support for the Big Seven model has since emerged, much of which hinges on robustness across cultures. Benet and Waller (1995) sought to evaluate the cross-cultural generalizability of the model, using a Spanish translation of the IPC-7. They obtained self-reports of the IPC-7 in English- and Spanish-speaking samples and peer-reports of the IPC-7 in a Spanish-speaking sample. A seven factor solution fits the IPC-7 items well in all samples, and the solutions were remarkably similar across samples. There were some differences, though, highlighting the impact of culture on self-evaluations. For example, in self-reports, “odd” had a negative loading on the conventionality factor in American samples but loaded on positive valence in the Spanish sample. Benet and Waller also collected self-reports of the Big Five in English- and Spanish-speaking samples. In a joint factor analysis of the Big Five and Seven items, a seven factor model fit the data better. Seven factor models have also been seen in Hebrew (Almagor et al. 1995) and Tagalog (Church et al. 1998), though these factor structures were not exact replications of Tellegen and colleagues’ model. For example, in the study of Hebrew, the typical fifth factor, openness to experience or conventionality, did not emerge. Instead, Almagor and colleagues described it as a continuation of the extraversion/positive emotionality factor with adjectives such as talkative, eccentric, and silent loading on it. In the study of Tagalog, an entirely different factor structure emerged with a solid negative valence factor but an absent positive valence factor. A negative valence factor has also appeared in studies of the Dutch lexicon (De Raad and Hoskens 1990; De Raad et al. 1988). First, De Raad et al. (1988) collected self- and peer-reports of verbs and verbal phrases, rather than adjectives, and found a ten factor solution with factors resembling the Big Seven negative and positive valence factors. Next, De Raad and Hoskens (1990) collected self- and peer-reports of nouns and found support for a seven factor model. In this second study, a negative valence dimension was clearly identified, though a positive valence dimension was not. Finally, a study of Serbian provided partial support for the Big Seven (Smederevac et al. 2007). Smederavac and colleagues administered an adjective rating questionnaire and uncovered a factor solution identical to Tellegen and Waller’s (1987). However, in their principal component analysis of a questionnaire of statements (rather than adjectives), a five factor solution emerged, one different from the Big Five.
In addition to exploring cross-cultural replicability, researchers have sought to understand the organization and psychological relevance of evaluative terms. Benet-Martinez and Waller (2002) selected adjectives from a list of evaluative terms that Norman (1967) excluded from his factor analysis of personal descriptors. In two separate studies, they examined the structure of these evaluative terms. In the first study, participants sorted the adjectives into groups based on similar characteristics. The sortings were converted to co-occurrence matrices, which were factor analyzed in an effort to discern the internal/conceptual structure of evaluative terms. In the second study, participants rated evaluative adjectives in terms of how strongly they agreed or disagreed in terms of how well that term described them. These ratings were also factor analyzed in order to appreciate the external structure of the terms. There was generally a great deal of convergence across studies. In both, five factors emerged, which included distinction and worthlessness (two facets of positive valence), and depravity, stupidity, and unconventionality. Benet-Martinez and Waller (2002) concluded that these dimensions represent important personality variance untapped in Big Five and other personality trait models.
Finally, there is evidence for the incremental validity of the Big Seven in psychopathology research. Indeed, there is research documenting how the Big Seven relate to mental disorders and maladaptive personality and explain variance unaccounted for by the Big Five (Durrett and Trull 2005; Simms 2007). This is discussed in detail below.
Critiques of the Seven Factor Model
Despite some evidence in support of its validity, the seven factor model has been met with some resistance. A common criticism is that positive and negative valence reflect response styles and are artifactual. For example, they may reflect high and low social desirability, respectively, and thus are artifacts that should be either subsumed by other factors or not included at all. This possibility is noted by not only the authors (Tellegen and Waller 1987) but other scholars as well (McCrae and Costa 1995; McCrae and John 1992). Related, the valence factors arguably could be artifactual factors resulting from highly skewed items. That is, the items loading on the valence factors tend to be those that are endorsed relatively infrequently (McCrae and John 1992). It has also been argued that positive and negative valence are simply maladaptive extremes of normal personality traits. Widiger (1993) argued this supposition, indicating that negative valence represents an extreme of low agreeableness, and positive valence represents an extreme of high emotional stability. Tellegen (1993) responded to these latter two critiques directly, explaining that effects of difficulty and skew are removed when tetrachoric and polychoric correlations are computed and that factor analyses of Big Seven markers based on both polychoric correlations and product-moment correlations are nearly identical. That is, the factor analytic method rules out the possibility of “difficulty factors.” Waller (1999) also noted that the nature of the resulting seven factors directly challenges these criticisms. The positive and negative valence dimensions are comprised of items with strong loadings on them. Further, the items that loaded highly on these dimensions had near-zero correlations with all other factors suggesting they are orthogonal to the Big Five (Tellegen and Waller 1987; Waller 1999). That is, the factors are relatively pure and clearly defined.
Critics have also suggested that the positive and negative valence factors could be accounted for by other confounding constructs (e.g., self-esteem and/or narcissistic self-regard). For example, Ashton and Lee (2001) deemed the Big Seven model’s inclusion of self-evaluative terms to be inappropriate, as this type of self-evaluation can be confounded with self-esteem. However, Tellegen (1993) had already acknowledged this, stating that the two additional factors could indeed reflect differences in self-esteem. He argued that this does not imply that the factors should be removed, rather, that these state- and trait-level aspects of self-evaluation are important to the research of individual differences. Waller (1999) defined both factors as representative of those who endorse extreme adjectives as descriptors of themselves. It would follow logically that those individuals with high self-esteem (perhaps extremely high) are most likely going to endorse items such as “special” or “outstanding.”
Another concern McCrae and John (1992) raised is the need to identify outcomes and/or behaviors which are linked to, or accounted for, by the two additional factors of the Big Seven model. Unquestionably, establishing the criterion validity of the valence factors is of paramount importance. This discussion echoes Box’s (1976), and more recently, Hand’s (2014), sentiments: a model is always wrong, but if it serves an important, specific purpose, it has use. That being said, it seems the seven factor model may have utility in the area of psychopathology, which we discuss in detail below.
The Seven Factor Model in Psychopathology Research
One arena in which the validity of the positive and negative valence dimensions has potential to gain traction is psychopathology research. The literature examining links between personality and psychopathology is vast, and the connections between traits and disorders are fairly well-established. For example, there are meta-analyses summarizing Big Five connections with Axis I disorders (Kotov et al. 2010) and personality disorders (Saulsman and Page 2004). There is decidedly less research on the utility of the Seven Factor model in the realm of psychopathology although conceptually at least, the sixth and seventh factors may play a critical role.
Benet-Martinez and Waller (2002) and others (Waller and Zavala 1993) made the case that positive and negative valence could shed light on the maladaptive self-evaluative processes characteristic of several personality disorders. As they noted, this would include narcissistic personality disorder, a hallmark of which is a grandiose sense of self-importance, borderline personality disorder, which involves an unstable self-image, and avoidant personality disorder, which entails intense negative self-evaluation (American Psychiatric Association [APA] 2013). Others have argued that they might also be related to Axis I disorders (we acknowledge that the Diagnostic and Statistical Manual of Mental Disorders – Fifth Edition abandoned the multiaxial system, though we continue to use the term “Axis I” to distinguish clinical disorders from personality disorders in order to simplify our discussion). Durrett and Trull (2005), for instance, noted that at least one key symptom of depression, intense feelings of guilt/worthlessness (APA 2013), certainly involves negative self-evaluation.
Despite the potential links between psychopathology and positive and negative valence being transparent, there is a scarcity of research in this area. One exception is Durrett and Trull’s (2005) research in which they examined the links between positive and negative valence and personality disorders as well as Axis I disorders. In a first series of hierarchical regression models, they sought to predict personality disorder symptom counts from the Big Five and the Big Seven, with the latter entered in the last step of the regression. The Big Seven accounted for a small (i.e., no more than 8%) yet statistically significant amount of variance above and beyond the Big Five in all but three disorders, paranoid, schizoid, and narcissistic. In that final step of the model, though, negative valence was often a significant, positive predictor. This was true for models predicting paranoid, schizotypal, antisocial, borderline, avoidant, dependent, and obsessive compulsive personality disorders. Positive valence was a significant, negative predictor for schizotypal only and surprisingly did not significantly predict narcissistic personality disorder. In a separate series of hierarchical regression models, they sought to predict lifetime Axis I diagnoses from the Big Five and the Big Seven, with the latter entered in the last step of the regression. The Big Seven did not have any incremental validity, although low positive valence significantly predicted any Axis I disorder, any substance use disorder, and any eating disorder. This finding was unexpected as they anticipated that negative, not positive, valence would predict Axis I disorders.
Simms (2007) followed up on this study by examining the relationships among positive and negative valence and maladaptive personality. He included scales on the IPC-7 (Tellegen et al. 1991), a measure of the Big Five, the Narcissistic Personality Inventory (NPI; Raskin and Terry 1988), the Rosenberg Self-Esteem Scale (RSES; Rosenberg 1965), and the Schedule for Nonadaptive and Adaptive Personality – Second Edition (SNAP-2; Clark et al. 2009), which is a measure of maladaptive personality traits. In terms of the SNAP-2 temperament and trait scales, positive valence was positively correlated with entitlement, positive temperament, and exhibitionism, and negatively with detachment and self-harm, particularly the subscale of low self-esteem. In terms of the SNAP-2 diagnostic scales, positive valence was positively related to narcissism and negatively to schizoid, avoidant, and dependent personality disorders. Finally, positive valence had strong positive correlations with the RSES and NPI. Negative valence had moderate-strong positive correlations with several SNAP-2 temperament and trait scales, as well as the diagnostic scales, including mistrust, manipulativeness, aggression, self-harm (both low self-esteem and suicide proneness), eccentric perceptions, and detachment, as well as paranoid, schizoid, schizotypal, antisocial, borderline, and avoidant personality disorders. Negative valance correlated negatively with the RSES.
Next, Simms (2007) fit a series of hierarchical regression models to determine whether positive and negative valence had incremental validity in predicting personality pathology over the Big Five. Positive valence led to increased predictive validity for narcissistic and histrionic personality disorders, accounting for an additional 13% and 2% of variance, respectively. It also accounted for a significant amount of variance for self-harm (including low self-esteem and suicide proneness), positive temperament, exhibitionism, and entitlement, accounting for an additional 15% of the variance in entitlement. Negative valence significantly predicted schizoid, schizotypal, antisocial, borderline, narcissistic, and avoidant personality disorders (the amount of variance accounted for ranged from 1% to 6%), as well as several trait and temperament scales including manipulativeness, aggression, self-harm (including low self-esteem and suicide proneness), eccentric perceptions, and workaholism. In sum, positive and negative valence had links with many scales related to self-evaluative processes, as expected, but several unexpected connections were also observed, indicating they have a broader relevance for personality pathology than previously though.
Although their findings differed somewhat (perhaps due to methodological differences), collectively, Durrett and Trull’s (2005) and Simms’s (2007) results confirm the primary hypothesis; that is, positive and negative valence are significantly related to particular aspects of psychopathology and have incremental validity over the Big Five. In terms of specific hypotheses, results from these two studies offer mixed support.
As described above, positive valence entails a high sense of self-worth and grandiose feelings of self-importance (Waller 1999). As expected, it is related to high self-esteem and narcissism and related traits such as entitlement (Simms 2007; cf. Durrett and Trull 2005, who conjectured the lack of significant association between narcissism and positive valence was due to incongruent measures of the constructs). In fact, the strength of association is nontrivial. For example, the Big Five collectively accounted for 15% of the variance in narcissistic personality disorder while positive valence accounted for an additional 13%. Positive valence was unexpectedly related to any Axis I, substance use, or eating disorder (Durrett and Trull 2005), thus the lack of positive self-evaluation is highly relevant for Axis I disorders. Earlier work carried out by Benet-Martinez and Waller (2002) may shed some light on this. As discussed above, they factor analyzed evaluative adjectives and settled on a five factor solution. Two factors tapped different aspects of positive valence; the factor labeled “distinction” was marked by adjectives related to high status and being exceptional, and the factor labeled “worthlessness” was marked by adjectives related to being meaningless and unable. The former, distinction, is clearly related to the grandiose self-importance felt by those with narcissistic tendencies. The latter, worthlessness, taps into feelings germane to many Axis I disorders, such as mood disorders (APA 2013).
In contrast, negative valence, which is marked by adjectives such as evil (Waller 1999), was surprisingly unrelated to Axis I disorders. Durrett and Trull (2005) maintained that such disorders entail an absence of positive self-evaluation, rather than a presence of negative self-evaluations. Such negative self-evaluations do, however, seem to be salient for personality disorders and maladaptive personality traits (Durrett and Trull 2005; Simms 2007). Indeed, negative valence was significantly related to many personality disorders and related traits. For example, as expected, Simms found that negative valence was positively related to antisocial and borderline personality disorders and manipulativeness, aggression, and self-harm. Durrett and Trull’s hypotheses that personality disorders characterized by low self-esteem, namely borderline, avoidant, and dependent personality disorders, would be positively related to negative valence were supported. Unexpectedly, several disorders and traits characterized by odd or eccentric features – paranoid, schizoid, and schizotypal disorders, and eccentric perceptions – were related to negative valence. Simms offered two possible explanations for this finding. First, cognitive and perceptual distortions typical of these disorders could apply to self-evaluations. That is, those with schizotypal personality disorder, for example, have a distorted negative self-image, which is inconsistent with others’ view of them. Alternatively, those individuals might actually be in tune with their unusual thoughts and behaviors so the increased negative valence is accurate.
It is clear that positive and negative valence play a role in mental disorders and maladaptive personality. However, it is also apparent that the exact nature of their relationship with specific disorders needs further clarification. With regard to personality disorders, some relationships were as expected, some unexpected relationships arose, and some findings were not replicated across studies (Durrett and Trull 2005; Simms 2007). With regard to Axis I disorders, the hypotheses of the single study assessing their relationship with positive and negative valence were not supported. Durrett and Trull concluded that, “The lack of predictable, differential associations with personality or Axis I psychopathology that are theoretically coherent call into question the utility of these evaluative terms in clinical assessment” (2005, p. 366). We would argue that additional research needs to be conducted before drawing this conclusion. As Simms noted, Durrett and Trull’s methodology was different from his own, which may explain the lack of replication. As far as the emergence of unanticipated relationships, it is important to recognize that the outcomes of negative and positive valence are, to date, understudied so it is not surprising that we do not have a full appreciation of how broad their function is in psychopathology or any other domain.
As discussed above, outcomes associated with positive and negative valence, including psychopathology, remain understudied. In addition to clearly establishing the connections between these two evaluative factors and specific disorders, potential explanations for the connections ought to be considered as well. For example, one should consider why low positive valence and substance use disorders are connected. There are at least six possible explanations for this. The first two postulate about the effect of personality on mental disorders. The vulnerability model suggests that premorbid personality traits act as risk factors. For example, individuals who do not feel special might be more inclined to engage in excessive drinking. The pathoplastic relationship hypothesis would suggest that positive valance would influence substance use disorders once they have already surfaced (i.e., their course, expression over time, and susceptibility to treatment). On the flip side, the scar and complication models examine the disorders’ effect on personality traits. According to the scar model, mental disorders have an impact on traits. For example, those engaging in problematic substance use might begin to have less positive self-evaluations as a result of the behavior. The complication model is similar, but the disorder’s effect on the trait is assumed to be temporary, at its height when the expression of the disorder is most severe. While these four hypotheses consider personality and psychopathology to be distinct constructs, the common cause and spectrum hypotheses assume differently. The common cause model suggests the two share a common root, such as a genetic risk factor. The spectrum model assumes, for example, that low positive valence and substance use disorders are different manifestations of a single process. Some combination of these six is also possible (Andersen and Bienvenu 2011; South et al. 2010). Longitudinal data would be ideal to disentangle the processes through which self-evaluations and psychopathology are related.
In addition to more fully understanding outcomes associated with positive and negative valence and the explanations for the associations, we need to gain an understanding of the lower-order structure of the traits. McCrae and Costa (1995) reasoned that basic dimensions of personality must meet three criteria, one of which is that the dimension must subsume more specific facets. Indeed, there is agreement in the field that personality is hierarchically structured with broad traits, such as the Big Five, subsuming more narrow traits (Markon et al. 2005). Various five factor models outline a number of lower-order facets. For example, Costa and McCrae’s (1995) model includes six facets for each of the Big Five. To our knowledge, there is only one study examining the structure of evaluative traits (Benet-Martínez and Waller 2002). Factor analyses point to five factors including unconventionality, stupidity, depravity, worthlessness, and distinction. The authors noted that the depravity factor resembles negative valence and that distinction and worthlessness are two facets of positive valence. Additional work is needed to elucidate the lower-order structure of these traits. This could potentially lead to increased understanding of their predictive validity. Examining only broad-level traits can obscure important personality-psychopathology (Walton et al. 2016) and other associations (Paunonen and Ashton 2001).
Not only is there a need for understanding the lower-order structure of personality traits, but it is also imperative to understand how these evaluative traits fit into the higher-order structure of personality and psychopathology. For example, Markon et al. (2005) identified an integrated hierarchical structure of normal and abnormal personality. Efforts have also been made to identify a metastructure that accounts for the covariation among mental disorders, personality disorders, and maladaptive personality traits (Wright and Simms 2015). Prior research points to two broad psychopathology dimensions, an internalizing factor that accounts for comorbidity in mood and anxiety disorders characterized by negative emotion, and an externalizing factor that accounts for comorbidity in substance use disorders and antisocial personality disorder characterized by disinhibition (e.g., Krueger et al. 1998). Adding normal and abnormal personality traits into the mix results in a five factor solution including dimensions of internalizing, disinhibition, psychoticism, antagonism, and detachment (Wright and Simms 2015). Knowledge of the location of positive and negative valence in such metastructures is lacking.
The seven factor model (Tellegen and Waller 1987) is one of many personality trait models put forth. While it has not garnered as much empirical support as the five factor model (McCrae and Costa 2003), for example, and it has its critics (Ashton and Lee 2001; McCrae and Costa 1995; McCrae and John 1992), there is some empirical evidence to support its validity (Almagor et al. 1995; Benet and Waller 1995; Church et al. 1998; De Raad and Hoskens 1990; De Raad et al. 1988; Smederevac et al. 2007) and utility (Durrett and Trull 2005; Simms 2007). As evidenced by the dates of the citations provided here, interest in validating the model seems to have waned, and the one field in which it is likely to have the greatest utility, psychopathology, is remiss in its attention to the model. Conceptually, it is clear how positive and negative self-evaluations play a role in psychopathology. What is missing is the body of empirical evidence elucidating this role. We have outlined several steps we feel as though ought to be taken in order for the model to make a real contribution to our understanding of psychopathology and possibly other areas.
- Benet, V., & Waller, N. G. (1995). The ‘Big Seven’ model of personality description: Evidence for its cross-cultural generality in a Spanish sample. Journal of Personality and Social Psychology, 69, 701–718.Google Scholar
- Block, J. H., & Block, J. (1980). The role of ego-control and ego-resiliency in the organization of behavior. In W. A. Collins (Ed.), Minnesota symposia on child psychology (pp. 39–101). Hillsdale: Erlbaum.Google Scholar
- Clark, L. A., Simms, L. J., Wu, K. D., & Casillas, A. (2009). Schedule for nonadaptive and adaptive personality – Second edition (SNAP-2). Minneapolis: University of Minnesota Press.Google Scholar
- Gough, H. (1987). California psychological inventory administrator’s guide. Palo Alto: Consulting Psychologists Press.Google Scholar
- McCrae, R. R., Terracciano, A., & 78 Members of the Personality Profiles of Cultures Project. (2005). Universal features of personality traits from the observer’s perspective: Data from 50 cultures. Journal of Personality and Social Psychology, 88, 547–561.Google Scholar
- Norman, W. T. (1967). 2800 personality trait descriptors: Normative operating characteristics for a university population. Unpublished manuscript, Department of Psychology, University of Michigan.Google Scholar
- South, S. C., Eaton, N. R., & Krueger, R. F. (2010). The connections between personality and psychopathology. In T. Millon, R. F., Krueger, R.F., & E. Simonsen (Eds.), Contemporary directions in psychopathology: Scientific foundations of the DSM-V and ICD-11 (pp. 242–262). New York: Guilford Press.Google Scholar
- Tellegen, A., & Waller, N. G. (1987). Reexamining basic dimensions of natural language trait descriptors. Paper presented at the 95th annual meeting of the American Psychological Association, New York.Google Scholar
- Tellegen, A., Grove, W., & Waller, N. G. (1991). Inventory of personal characteristics #7. Unpublished measure, University of Minnesota.Google Scholar
- Tupes, E. C., & Christal, R. E. (1961). Recurrent personality factors based on trait ratings (USAF ASD Tech. Rep. No. 61–97). Lackland Air Force Base, TX: U.S. Air Force. (Republished in Journal of Personality, 1992, 60, 225–251).Google Scholar
- Waller, N. G. (1999). Evaluating the structure of personality. In C. R. Cloninger (Ed.), Personality and psychopathology (pp. 155–197). Washington, DC: American Psychiatric Press.Google Scholar
- Walton, K. E., Pantoja, G. A., & McDermut, W. (2016). Association between lower-order facets of personality and dimensions of mental disorder. Manuscript submitted for publication.Google Scholar
- Widiger, T. A. (1993). The DSM-III-R categorical personality disorder diagnoses: A critique and an alternative. Psychological Inquiry, 4, 71–93.Google Scholar