Personality or Factor-analytically Developed, Lay Person, Self-reported, Single-word, Adjectival Descriptors of Global Characteristics of Personality Structure: The NEO Five Factor Model and Skimming the Surface of the Wetlands of Personality.

Darek Dawda
Department of Psychology
Simon Fraser University

Abstract

Despite criticisms (Block, 1995), the NEO Five Factor Model (NEO FFM) continues to be one of the most researched models of personality. In fact, numerous authors appear to have accepted the NEO FFM as a criterion to validate other personality and clinical constructs. Moreover, proponents of the NEO FFM advocate the model's usefulness in clinical assessment (e.g., Costa, 1991). In light of these developments, I would like to present a brief and non-technical overview of the problems inherent in the NEO FFM, emphasizing its potential lack of utility in clinical settings. The article first briefly summarizes criticisms raised by Block (1995): the limits of the method of factor-analysis and the abuse of its arbitrariness in the development of the NEO FFM; the exclusion of psychological language, theory, and insight in the development of the NEO FFM; the lack of control for response bias in the NEO self-report questionnaires; and the fact that the NEO questionnaires are only capable of reliably assessing their five global dimensions. The core of the present paper expands upon and goes beyond Block's (1995) criticism. I argue that the use of single-word adjectives makes the NEO FFM, or any trait theory of personality, incapable of describing the dynamics of personality organization or the variation of personality characteristics over situations. I concluded with the claim that the term personality has a broader meaning than that referred to by the proponents of the NEO FFM.

Personality ...

Although different personality theories work at different levels of explanatory complexity, psychologists generally agree that a complete theory of personality should address the questions of what are the characteristics of the person and how are they organized (i.e., personality structure), as well as how these characteristics develop and change over different contexts (i.e., personality process) (Pervin, 1980). Only such a theory can adequately address the issue of both individual differences and similarities in personality characteristics (nomothetic view), and the intra-individual complexity of personality organization and dynamics (idiographic view). Further, the claims personologists can make are limited by the methodology they use to develop and study a construct (Pervin, 1980). The kinds of phenomena attended to and the kinds of statistical procedures used determine the level of complexity and the limits of validity of a given framework.

I argue that trait theory in general, and the NEO Five Factor Model (NEO FFM; Costa & McCrae, 1985) in particular, have a defined place in personality research and theory. While the NEO FFM supplies an easy to administer research tool, its utility is limited to the description of individual differences in global characteristics of personality structure or to the prediction of behavior. Such a model inadequately addresses the dynamic organization of personality characteristics. Nor does it address cross-situational variations in behavior. As a consequence, the NEO FFM has little utility in clinical settings, where comprehensive idiographic and dynamic assessment of an individual is a necessity.

Or Factor-analytically Developed ...

The factor-analytic technique, which was used to develop the Five Factor Model (FFM) of personality, is often misjudged as a purely empirical and atheoretical method. Such a fallacy allows the proponents of the FFM to speak about the "discovery" of basic personality dimensions or the "empirical fact" (McCrae & John, 1992, p.194). However, the method of factor-analysis is not an automatic truth generator. At its best, it can be a useful aid in making sense of a large amount of multivariate data by reducing it to a comprehensible number of clusters. At its worst, it can be a tool in the hands of data analysts who strive to prove their points, while disguising their attempts by appealing to the authority of factor-analysis as an objective technique. In the following section, I present a brief summary of some potential problems inherent in the factor-analytic technique. I point to several arbitrary decisions during the development of the NEO FFM which have influenced its present form.

The method of factor-analysis has a number of inherent features which make it less than a purely objective method. First, loss of information may result from the fact that factor-analysis works on bivariate linear correlations, and is thus insensitive to non-linear relationships. Second, the relationship among a group of bivariate correlations is not always as simple as the factor-analytic method assumes. For example, in a moderator model, one variable moderates the relationship between two other variables without necessarily being related to either of them. Unless specified, factor analysis will miss such a relationship and disregard the moderator. Another example is a suppressor variable, which "cleans" another variable of "bad" variance thus enhancing its relationship with the third variable, without necessarily being correlated to that third variable. The suppressor variable may hinder the interpretation of a factor, which is normally proceeds through examining the content of the variables that load on that factor. The seriousness of the limitations inherent to the factor-analytic method, then, depends on the presence of non-linear, moderator, or suppressor relationships. As the impact of these issues has not been addressed by the proponents of the NEO FFM (Costa & McCrae, 1992a; 1992b), it remains unknown how many descriptors of personality or the relationships among the descriptors have been omitted or misrepresented by the model.

Additionally, the method of factor-analysis requires arbitrary decisions on the part of the analyst. First, the method of factor rotation chosen affects the subsequent grouping and loadings of variables. For instance, orthogonal rotation, the method preferred by Costa and McCrae (1992a) because of its interpretational elegance, often imposes an artificial structure and may lead to the elimination of important variables. Second, the decision about the number of factors which adequately represent the data set is to a great degree arbitrary. Hence, the controversy about the number of factors which optimally account for the set of adjectival descriptors of personality has been ongoing (Eysenck, 1991). Third, typical development of a measure such as the NEO questionnaires consists of a series of analyses, in which the variable set is modified for each subsequent analysis. In order to achieve a reliable and replicable factor structure, the analyst retains the variables which load highly on their own, and only on their own, factors. Such a procedure leads to the elimination of double loading variables, and/or variables which themselves have few correlates and do not have their own axis. These variables, however, may be important for the validity of the final solution.

In summary, the arbitrariness of the factor-analytic method allows for a significant degree of data prestructuring. The main goal of the process is to achieve a clean factor structure, but reliability is achieved at the potential loss of validity. The final solution is only one of the many possible solutions. The interested reader may find a more elaborate discussion of these issues in any advanced book on multivariate statistics (e.g., Cliff, 1987).

John, Angleitner, and Ostendorf (1988) and Block (1995) identified a number of arbitrary steps which crucially contributed to the present form of the FFM model in general, and the NEO FFM in particular. A very brief summary of the history of, and methodology behind, these trait models follows. First, Allport and Odbert (1936) extracted nearly 18,000 potentially useful trait descriptors from Webster's dictionary. This initial list was subsequently abridged by Cattell (1943a; 1945; 1947), first to 171 terms, then to a set of 35 variables using a mixture of correlational techniques and arbitrary decisions. These variables were then used by Tupes and Christal (Tupes, 1957; Tupes & Christal, 1992), the acclaimed discoverers of the Big Five (McCrae, 1992). A series of six analyses in which a consistent five-factor solution was found were reported. However, for all but the first analysis the data were prestructured to fit the initially found five-factor solution. The final solution consisted of five groups of adjectives which were so closely semantically related that they could be regarded redundant. The Big Five was born. Subsequent research on the FFM has used either Cattell's 35 variable set, or the later developed Norman's (1963; 1967) set of 75 categories. Norman's categories, however, were prestructured to fit the FFM by being rationally assigned to the positive or negative pole of one of the five Tupes and Christal's dimensions. Goldberg's (1990; 1992; 1993) role in the development of the Big Five model was also prominent, though his effort was again marked by a significant degree of data prestructuring. Faced with difficulties in obtaining a clear five-factor solution for Norman's list, Goldberg (1990) modified the set by eliminating the categories that misbehaved.

The sum of this research provided the foundation for the NEO FFM. Costa and McCrae (1976) initially started with a three-factor model obtained from Cattell's variables. The first two factors, Neuroticism and Extraversion, were well established in trait theory research, but the third factor was less congruent and more controversial. Costa and McCrae (1976) somewhat arbitrarily interpreted this third factor as Openness to Experience, and recreated the dimension by factor analyzing a number of items from existing Openness to Experience scales (Costa & McCrae, 1978). It was not until later that Costa and McCrae (1985) decided to add two new dimensions so as to fit their paradigm to the Big Five model. The selection of items for the remaining two factors, Agreeableness and Conscientiousness, was governed by the desire to obtain as orthogonal a solution as possible. The final result of Costa and McCrae's collaboration were two self-report, paper-and-pencil questionnaires (the NEO Personality Inventory - NEO-PI-R, and its shorter version the NEO Five Factor Inventory - NEO-FFI, Costa & McCrae, 1992b). Due to the emergence of these easily administered self-report measures, the Big Five model became one of the most popular personality research tools. The interested reader is referred to John, Angleitner, and Ostendorf (1988), and Block (1995) for a detailed discussion of the development of the FFM and the NEO FFM.

In conclusion, the method of factor-analysis is not a completely objective technique, as it has a number of inherent limitations and requires arbitrary decisions. The use of factor-analysis in the development of the NEO FFM was particularly remote from objective discovery. A series of inadequately elaborated rational decisions, motivated by the need to achieve reliable, orthogonal factor structure, guided the selection of items. A significant degree of data prestructuring at many stages of the development of the FFM calls for skepticism as to the validity of the NEO FFM. First, is the magic number five optimal? Second, are there alternative Big Five frameworks which have better explanatory relevance? Before further research addresses these questions, an awareness of the discussed limitations of factor-analysis and the history of development of the NEO FFM is crucial.

Lay person ...

The source of information for the development of the FFM model was a dictionary (Allport & Odbert, 1936; Norman, 1963; 1967). Cattell's (1943b) lexical hypothesis assumes that the important aspects of human nature have corresponding terms in everyday language. Moreover, the employment of factor analysis assumes that the more important features of personality will have correspondingly more adjectival descriptors. It is possible, however, that there are large clusters of adjectives which have little importance for personality description, and that some important personality features will have relatively few corresponding trait names (Block, 1995). Block (1995) proposed that an adequate theory of personality should pay more attention to the terminology invented by personality theorists. This marginalization of psychological language in the creation of the Five Factor model warrants concern.

Self-reported ...

Most of the research that contributed to the development of the FFM used other people's ratings of subjects (e.g., Tupes & Crystal, 1992; Norman, 1963; Goldberg, 1992). Costa and McCrae (1985; 1992), however, adopted self-report, paper-and-pencil instruments (NEO-PI, Costa & McCrae, 1985; NEO-PI-R, NEO-FFI, Costa & McCrae, 1992b). The creation of self-report questionnaires greatly advanced the research popularity of the NEO FFM. However, the reader should be reminded that an inherent problem of self-reports is response bias (Cohen, 1992). While some personality measures have scales to control for such response biases as social desirability or faking bad (e.g., MMPI, Hathaway & McKinley, 1989; Eysenck Personality Inventory, Eysenck & Eysenck, 1965), the NEO questionnaires provide no means for such control. This may further impede the validity of the NEO FFM.

Single-word, Adjectival Descriptors ...

In this section, I argue that a set of single adjectives is inadequate to describe the organization and dynamics of personality. While the trait approach may be used for describing static personality structure or for the purposes of prediction, the validity and utility of such description for individual idiographic assessment is questionable.

The trait theory's greatest limitation is its reliance on simple relations among single adjectival descriptors of personality. For such an approach to work, it must assume the immutability of the meaning of, and relationships among, the adjectives for different individuals over different contexts. However, with the exception of the trait theory itself, no personality theory allows for making such an assumption. Freud (1920) was the first to point out that the meaning of symptoms is only an expression of underlying dynamic personality organization, and that the meaning is lost when a single symptom is considered outside the total motivational context. Other psychodynamically oriented theories (Erikson, 1980; attachment theory, Bowlby, 1969; object relation theory, see Hamilton, 1989) also recognize the importance of the individual's developmental history in interpreting the meaning of personality characteristics. Many theorists (Jung, 1928; Fromm, 1941; May, 1969; Rogers, 1965; Markus & Ruvolo, 1988) have pointed to the importance of personal goals and values in determining the significance of existing personality characteristics. Self-discrepancy theorists (Higgins, 1987) regard the discrepancy between the perceived actual self and one's self-guides (i.e., ought self, ideal self), rather than the attributes of the actual self per se, as crucial markers of personality. Similarly, a less dynamic, organizational perspective (Sroufe, 1990) views the meaning of a trait descriptor as stemming from a complex intra-individual cognitive organization of personality. An attribute cannot be adequately considered outside the idiographic context.

An additional line of criticism arises from considering the impact of the situation on personality variation. Behaviorism (Skinner, 1987) provides ample evidence for the claim that environmental contingencies regulate behavior. Vaillant's (1987) principle of adaptability points to the person's efforts to adjust and adapt to different situations and implies the modification of personal strategies in different situational contexts. Interpersonal theorists (Baldwin, 1992; Horney, 1937) stress the primacy of interpersonal interactions in determining identity, where different social roles are adopted in different interpersonal contexts. The impact of the situation on behavior and the lack of stability of personality characteristics over situations has been demonstrated by Mischel (1968). In summary, the meaning of trait descriptors depends on the intra-individual, interpersonal, and situational context. In order to adequately understand a personality trait as it specifically relates to an individual, one's developmental history, motivation, personal goals and values, the interaction of these characteristics, and their interaction with different situational contexts have to be considered. This complexity cannot be captured using single-word, adjectival descriptors.

A brief description of an alternative approach to structured personality assessment should make the limitations of the single-adjective approach more evident. Mischel's (1995) recently created a structural, generic, content free framework based on person/situation interaction rather than stable personality features or contingencies of the environment. He proposed that one's personality characteristics be measured over a diverse range of situations, with the results presented as a personality profile. Such a profile would tap both the stability and variability of a characteristic over situations. Individual profiles may differ in the degree of cross-situational consistency of a characteristic. Thus, it is the person/situation interaction that determines one's personality. Additionally, Mischel's framework provides the potential for a more comprehensive measurement of the intra-individual complexity of personality organization. Motivation, developmental history, or personal goals may each determine the broader meaning of a trait and its relevance to one's functioning.

Both the person/situation interaction, and the interaction of intra-individual characteristics may be addressed by applying a set of internalized if...then relationships, which are a sound candidate for the replacement of single-word adjectives. Such if...then descriptors allow for measurement of the interaction of complex dynamic personality organization with environmental contingencies. In contrast to the trait approach, Mischel's (1995) framework provides both a structured and more comprehensive approach to personality research, and is a much sounder candidate for idiographic assessment in clinical settings.

Because of the number and complexity of the possible individual variations of if...then relationships, using an interactional, structural approach cannot result in a situation-free, grand theory of personality. Instead, it requires a local application in more specific contexts and for more specific purposes. This may be considered a weakness by these theories who have reductionist tendencies. However, as evident from the present overview of personality theory, the alternative, exclusively nomothetic, static, trait approach to personality is a gross oversimplification of the complete picture. The trait approach does not inform us about personality. Though there may be pragmatic reasons for using such an approach, as for the purpose of prediction, this should not be confused with construct validity (Cronbach & Meehl, 1955).

Of Global Characteristics of Personality Structure?

My final point is that the NEO FFM is capable of reliably assessing only its five global dimensions. The reliability of the six facets of each dimension is inadequate, as they are redundant and their meaning has questionable theoretical relevance (Block, 1995). The NEO questionnaire was created for the purpose of obtaining five, as orthogonal as possible, global dimensions, and its facet scales were prestructured to serve that goal. Thus, even as a theory of static personality structure, the Five Factor model is too global to be a useful clinical tool.

Skimming the Surface of the Wetlands of Personality

In the present article, I discussed the limitations of the trait theory in general, and the NEO FFM (Costa & McCrae, 1985) in particular. The development of the NEO FFM was neither a discovery, nor a product of psychological theory. Instead, the NEO FFM was constructed using a mixture of factor-analysis and theoretically unelaborated arbitrary decisions. The exclusion of psychological language and insight in the development of the NEO FFM, and the lack of control for response bias in the NEO self-report questionnaires further limits the validity of the model. Most importantly, single-word, global adjectives are inadequate descriptors of the intra-individual and cross-situational variations in the meaning of personality characteristics. Personality is complex and the methodology used to study personality should reflect this complexity. Costa and McCrae (1995, p.216) pride themselves, stating that the "perhaps most impressive achievement of the FFM is its reduction of conceptual jangle". Given the presented evidence, should such a reduction be considered an achievement?

The recent suggestion that the NEO FFM be used in clinical settings (Costa, 1991) is not an adequately considered statement. While some may find the NEO questionnaires useful for studying individual differences on global and static dimensions of personality structure or for the purpose of prediction, the model attends to too narrow a range of personality phenomena to be adequate for the "comprehensive assessment of individuals" (McCrae, 1989, p.243).

Finally, in respect to long efforts of philosophers and psychologists, the term personality should not be used, without necessary explicit qualifications, to refer to a much narrower concept of factor-analytically developed, lay person, self-reported, single-word, adjectival descriptors of global characteristics of personality structure. Using the presently popular NEO FFM to study personality is just skimming the surface of the wetlands of personality.

References

Allport, G.W. & Odbert, H.S. (1936). Trait names: A psycho-lexical study. Psychological Monographs, 47, No.211.

Baldwin, M.W. (1992). Relational schemas and the processing of social information. Psychological Bulletin, 112, 461-484.

Block, J. (1995). A contrarian view of the five-factor approach to personality description. Psychological Bulletin, 117, 187-215.

Bowlby, J. (1969). Attachment and loss: Vol. 1. Attachment. New York: Basic Books.

Cattell, R.B. (1943a). The description of personality: Basic traits resolved into clusters. Journal of Abnormal and Social Psychology, 38, 476-506.

Cattell, R.B. (1943b). The description of personality: I. Foundations of trait measurement. Psychological Review, 50, 559-594.

Cattell, R.B. (1945). The description of personality: Principles [sic] findings in a factor analysis. American Journal of Psychology, 58, 69-90.

Cattell, R.B. (1947). Confirmation and clarification of primary personality factors. Psychometrica, 12, 197-220.

Cliff, N. (1987). Analyzing multivariate data. Orlando, FL: Harcourt, Brace, Jovanovich.

Cohen, R.J. (1992). Psychological testing and assessment. Mountain View, CA: Mayfield.

Costa, P.T. (1991). Clinical use of the Five-Factor Model: An introduction. Journal of Personality Assessment, 57, 393-398.

Costa, P.T., Jr. & McCrae, R.R. (1976). Age differences in personality structure: A cluster analytic approach. Journal of Gerontology, 31, 564-570.

Costa, P.T., Jr. & McCrae, R.R. (1978). Objective personality assessment. In M. Storandt, I.C. Siegler, & M.F. Elias (Eds.), The Clinical Psychology of Aging (pp. 119-143). New York: Plenum.

Costa, P.T., Jr. & McCrae, R.R. (1985). The NEO Personality Inventory manual. Odessa, FL: Psychological Assessment Resources.

Costa, P.T., Jr. & McCrae, R.R. (1992a). Four ways five factors are basic. Personality and Individual Differences, 13, 653-665.

Costa P.T., & McCrae, R.R. (1992b). Revised NEO Personality Inventory (NEO-PI-R) and the NEO Five-Factor Inventory (NEO-FFI): Professional manual. Odessa, FL: Psychological Assessment Resources, Inc.

Costa P.T., & McCrae, R.R. (1995). Solid grounds in the wetlands of personality: A reply to Block. Psychological Bulletin, 117, 216-220.

Costa, P.T., McCrae, R.R., & Dye, D.A. (1991). Facet scales for agreeableness and conscientiousness. Personality and Individual Differences, 12, 887-898.

Cronbach, L.J. & Meehl, P.E. (1955). Construct validity in psychological tests. Psychological Bulletin, 52, 281-302.

Erikson, E.H. (1980). Identity and life cycle. New York: Norton.

Eysenck, H.J. (1991). Dimensions of personality: 16, 5, or 3? - Criteria for taxonomic paradigm. Personality and Individual Differences, 12, 773-790.

Eysenck, H.J. & Eysenck, S.B.G. (1965). Manual for the Eysenck Personality Inventory. London: University of London Press.

Freud, S. (1920). The Meaning of Symptoms. In A general introduction to psychoanalysis. New York: Washington Square Place. (pp.291-311)

Fromm, E. (1941). Escape from Freedom. New York: Holt, Rinehart, & Winston.

Goldberg, L.R. (1990). An alternative "Description of Personality": The Big-Five factor structure. Journal of Personality and Social Psychology, 59, 1216-1229.

Goldberg, L.R. (1992). The development of marker variables for the Big-Five factor structure. Psychological Assessment, 4, 26-42.

Goldberg, L.R. (1993). The structure of phenotypic personality traits. American Psychologist, 48, 26-34.

Hamilton, N.G. (1989). A critical review of Object Relations theory. American Journal of Psychiatry, 146, 1552-1559.

Hathaway, S.R. & McKinley, J.C. (1989). MMPI-2. Minneapolis, MN: University of Minnesota Press.

Higgins, E.T. (1987). Self-Discrepancy: A theory relating self and affect. Psychological Review, 94, 319-340.

Horney, K. (1937). The neurotic personality of our time. New York: Norton.

John, O.P., Angleitner, A., & Ostendorf, F. (1988). The lexical approach to personality: A historical review of the trait taxonomic research. European Journal of Personality, 2, 171-203.

Jung, C.G. (1928). The development of personality. Collected Works, Volume 17. (324-345) (Marriage as a psychological relationship).

Markus, H. & Ruvolo, A. (1988). Possible selves: Personalized representation of goals. In L. A. Pervin (Ed.), Goal concepts in personality and social psychology (pp. 211-242). Hillsdale, NJ: Erlbaum.

May, R. (1969). Existential psychology. New York: Random House.

McCrae, R.R. & John, O.P. (1992). An introduction to the Five Factor Model and its application. Journal of Personality, 60, 175-215.

Mischel, W. (1968). Personality and assessment. New York: Wiley.

Mischel W. & Shoda, Y. (1995). A Cognitive-Affective System theory of personality: Reconceptualizing situations, dispositions, dynamics, and invariance in personality structure. Psychological Review, 102, 246-268.

Norman, W.T. (1963). Toward an adequate taxonomy of personality attributes: Replicated factor structure in peer nomination personality rating. Journal of Abnormal and Social Psychology, 66, 574-583.

Norman, W.T. (1967). 2800 personality trait descriptors: Normative operating characteristics for a university population. Ann Arbor: University of Michigan, Department of Psychological Sciences.

Pervin, L.A. (1980). Personality: Theory, assessment, and research. New York: Wiley & Sons.

Rogers, C.R. (1965). Client-centered therapy. Boston: Houghton Mifflin Co.

Skinner, B.F. (1987). Whatever happened to Psychology as the Science of Behavior? American Psychologist, 42, 780-786.

Sroufe, L.A. (1990). An organizational perspective on the self. In D. Cicchetti & M. Beeghly (Eds.), The self in transition: Infancy to childhood (pp.281-308).

Tupes, E.C. (1957). Relationship between behavior and trait ratings by peers and later officer performance of USAF Officer Candidate School graduates (USAF PTRC Tech. Note No. 57-125). Lackland Air Force Base, TX: U.S. Air Force.

Tupes, E.C. & Christal, R.E. (1992). Recurrent personality factors based on trait ratings. Journal of Personality, 60, 225-251. (Reprinted from USAF ASD Tech. Report No. 61-97, 1961, Lackland Air Force Base, TX: U.S. Air Force.)

Vaillant, G.E. (1977). Adaptation to life. Boston: Little, Brown & Co.