The big-five model of personality is an advance on preceding models insofar as it was founded on reappraisals of
Cattell's sixteen factor model of personality, as measured by the
16 Personality Factor Questionnaire, or the "
16PF" (Cattell
et al.,
1970). However there has been debate over whether the
big-five model is an adequate model of personality. If it is not, then it is not that much of an advance on Cattell's original work.
It is true that there is a trend in personality psychology to regard the five-factor approach as an absolutely fundamental model, but its theoretical basis may leave a lot to be desired (Cooper, 1998). However, the five factors that have now been generally accepted as encapsulating the five-factor model are those defined by Costa and McCrae (1992) as "Openness to Experience", "Conscientiousness", "Extraversion", "Agreeableness", and "Neuroticism" - easily remembered by the acronym "OCEAN". The question that has been asked of this model, and the question that should be asked of any model of personality, is whether or not it is theoretically sound, and also, if is capable of describing personality.
An extensive criticism of the big-five approach has be put forward by Block (1995) which seriously questions the legitimacy of the approach of advocates of the big-five model. Whether or not the big-five model is theoretically sound can be determined by looking at how it came about - this is part of Block's critical approach.
The big-five model was derived from work by Cattell. Cattell was a proponent of the "lexical hypothesis" of personality theory, which states that terms describing all aspects of human personality that are important would have found their way into the language. On this rationale, Cattell reduced a list of 4,504 "non-judgemental trait-names" by Allport and Odbert (1936) and a series of his own terms used to describe conditions discovered by psychologists to 35 bipolar rating dimensions which he called the "standard reduced personality sphere". Cattell's model consists of 12 primary factors which emerged from his factor analyses. However, Block proposes that in making use of his own judgement and prior theoretical notions to decide upon his set of dimensions, Cattell may have entailed that his primary factors emerged from his analysis. This is an example of what is known as "prestructuring", which seems to be a running theme in Block's criticisms of the big-five model's development.
The big-five model really came about because of work by Tupes and Christal (1962) who factor analysed the results of Cattell's 16PF applied to 8 samples of young Air Force officers. In these analyses, 6 had 8 factors extracted, another had 5 factors, and the last, 12 factors. Tupes and Christal's claim was that "five fairly strong and recurrent factors emerged". Block, however, does not think that this recurrence is so striking because Tupes and Christal used the labourious (by hand) centroid method of factor analysis for their first analysis, which extracted 8 factors, and this was rotated subjectively to make three of these factors "unimportant". Furthermore, their subsequent analyses used the less labourious multiple-group method, where the variables were grouped into five subsets prestructured so as to correspond to the five rotated factors decided upon in the first analysis. In other words, they were designed to conform to the structure of the first analysis and factors other than these five were residualised. More importantly, according to Block, was that Tupes and Christal may have prestructured their analysis in that it was based on the Cattellian variables, which contains many semantically related variables. Rating scales that contain many synonyms will almost always generate a mathematical factor. Block suggests that the factorial equivalence of many of the variables found in the Tupes and Christal study may only represent reliable and coherent scorings of the rater, not underlying factors. However, the reliability of the raters are also questioned on the basis that they received only very little training (about two hours at the most) and did not know their peers they were rating for very long (as little as three hours) (Block, 1995; Cooper, 1998). The final criticism of Tupes and Christal given by Block is that they named their five factors on the basis of a study by French (1953) without any elaboration on what they mean. They called them "Surgency or Extraversion", "Agreeableness", "Dependability", "Emotional Stability, or the Opposite of Emotionality", and "Culture", which Block understandably calls "allusive". Block also makes the point that French's study was based on many studies that were conceptually or methodologically insufficient.
The next advance in the big-five theory was made by Norman (1963). On the basis of Tupes and Christal's factors, Norman selected 20 of their Cattellian variables - four for each factor. He then used these variables for undergraduate peer ratings, which were then factor analysed. Block raises the very valid point that five factors obtained were a "forgone conclusion" due to a prestructured set of variables. However, Norman's study is seen as empirical support for the big-five structure. In addition to this study, Norman (1967) wanted to find any previously omitted or new personality terms, so he added 175 single-word descriptors to Allport and Odbert's list and used his own decision rules to exclude terms he judged quantitative, evaluative, ambiguous, metaphorical, vague, difficult, obscure, little-known, anatomical, physical, grooming characteristics, temporary states, or social rules. Norman only included those terms he deemed "stable traits", and after having these rated by undergraduates on their knowledge of these words, he was left with 1,431 terms. These terms were assigned to the positive or negative poles of his five dimensions and Norman further formulated what he judged to be semantic clusters within each pole.
Norman obtained 75 clusters, which were used by Goldberg in addition to some adjectives of his own to try to go beyond Cattell's work. Goldberg also found five factors in his studies, but this is hardly surprising given he use of Norman's clusters which were built around five factors in the first place - more prestructuring. It seems that the big-five models is based largely on the results of prestructured analyses, which does nothing to suggest that the model is a good advancement of the understanding of personality.
The big-five model has been derived largely from the lexical hypothesis, which Block does not think is totally sound. Block suggests that it is a "psychologically insufficient" hypothesis, drawing on the observation of McCrae and Costa (1985) that psychologists have uncovered important aspects of personality that were not encoded in the language. They make the good point that there is no reason to suspect that simple analysis of common English terms for parts of the body will provide an adequate basis for the science of anatomy, so why should this be true of personality, which is after all a much more intangible science? Block also adds that there are many aspects of personality that cannot be captured with a single-word term, such as "nasty to individuals of lower rank".
A further problem with the background of the big-five model is that Goldberg used undergraduates to determine if his stable trait descriptors were understandable by laypersons. Block suggests that undergraduates are not the best judge of this and it would be better to use a more expert judgement to avoid losing fairly unheard of, but possibly vitally important descriptors. He says that novices do not make the same discriminations that experts do.
Yet another problem for the idea of the big-five factors is that they have repeatedly been found to be non-orthogonal and correlate with each other. For instance, Goldberg's (1992) 100-item five-factor marker set is problematic. Although it produces five nearly orthogonal factors when data is taken from self-ratings and ratings of liked others, it does not produce orthogonal factors at all if the sample also includes ratings of disliked others. There are intercorrelations between Agreeableness, Conscientiousness, Emotional Stability, and Culture/Intellect/Openness to Experience in the 30s to 50s (Block, 1995). This just goes to show how the nature of the sample can affect the results of a factor analysis.
Finally, Costa and McCrae's (1992) questionnaire for testing their interpretation of the big-five factors has not escaped the criticism of Block. Costa and McCrae (1976) had initially found three clusters in Cattell's 16PF - Extraversion, Neuroticism (both in agreement with Eysenck at that time), and also, a third and not well represented factor they termed "Openness to Experience". They then created a test using the 16PF variables and 6 additional ones for measuring Openness to Experience and found (unsurprisingly) that their third factor was now better represented. From this they constructed a new questionnaire based on their "NEO trait model" (Costa and McCrae, 1980). Block criticised the fact that this was not based on theory, just their personal thinking about how they should articulate their three domains.
However, after becoming influenced by the five-factor approach, Costa and McCrae wished to construct a questionnaire for the big-five which would integrate their NEO trait model. To do this they selected adjectives for their previously unconceptualised factors of Agreeableness and Conscientiousness from Goldberg (1980). For the fifth factor, they tried to include aspects of Openness to Experience that they believed were underrepresented by Goldberg. Their factor analysis found five factors, the fifth of which now looked more like Openness to Experience than Goldberg's Intellect dimension. This seems to be a case of more prestructuring and indeed, even Costa and McCrae (1985) admit that their results may have been due to bias in the items that picked. Block rejects Costa and McCrae's (1985) claims that they showed that the five-factor model could encompass their NEO with Agreeableness and Conscientiousness being orthogonal to the NEO dimensions, instead suggesting that all they did was link two "nominally equivalent and predictably related lexical factors". Costa and McCrae's (1985) NEO Personality Inventory (NEO-PI) "grafted" the extra big-five factors onto the NEO Inventory and was found by Costa and McCrae (1987) to yield five-factors in analysis. However, Block suggests that this should not be seen as yet more validation of the big-five model, once again due to prestructuring. Even the latest version of Costa and McCrae's questionnaire, the Revised NEO Personality Inventory fails to validate the big-five model. In fact, it seems worse than it was before the revision as the Neuroticism and Conscientiousness scales have a -0.53 correlation, with a 0.40 correlation between the Extraversion and Openness to Experience scales. Furthermore, Parker et al. (1993) found that the five factors did not emerge as expected. Block compounds the defeat of Costa and McCrae's approach as he sees that they allow their model, rather than eigenvalues or scree tests, to determine the number of factors they extract.
Advocates of the five-factor approach seem to be allowing "fiveness" to blind them somewhat to other possibilities. In addition to this, the big-five also fails to provide an adequate description of personality. Firstly, there is not enough clarity over what the factors actually mean. For instance, Costa and McCrae see "warmth" as a facet of Extraversion, while Goldberg sees it part of Agreeableness. Also, "Impulsivity" is classed as being part of Neuroticism by Costa and McCrae, and as being part of Extraversion by Goldberg. This shows that these factors are not clear, and perhaps this lack of clarity comes from the five factors being rather broad with fuzzy overlapping boundaries. This is something else that a good account of personality should not have. Summarising all the aspects of personality under five broad categories creates a great loss of information and no way of looking at more subtle traits. According to Block, a good model of personality also should not rely on factor analysis to decide the concepts of the model, and should also come from a conceptual language decided upon by experts rather than novices. The big-five model seems to fail on all these points and does not seem to have a solid theoretical background. The major problem seems to be that its apparent supporting evidence sufferers chronically from prestructuring. Therefore, the big-five does not make any advances in getting towards an understanding of what makes up personality.
References:
Cooper, C. (1998) Individual Differences. London: Arnold
Block, J. (1995) A contrarian view of the five-factor approach to personality description. Psychological Bulletin, 117: 187-215