An inter-battery factor analysis of the comrey personality scales and the 16 personality factor questionnaire

The scores of 700 Afrikaans-speaking university students on the Comrey Personality Scales and the 16 Personality Factor Questionnaire were subjected to an inter-battery factor analysis. This technique uses only the correlations between two sets of variables and reveals only the factors that they have in common. Three of the Big Five personality factors were revealed, namely Extroversion, Neuroticism and Conscientiousness. However, the Conscientiousness factor contained a relatively strong unsocialised component and in this regard it is similar to Eysencks Psychoticism factor. The results support the construct validity of the Comrey Personality Scales and the 16 Personality Factor Questionnaire. Implications for personality questionnaire design and validation are discussed. Opsomming Die tellings van 700 Afrikaanssprekende universiteitstudente vir die Comrey Persoonlikheidskale en die 16 Per- soonlikheidsfaktorvraelys is aan 'n interbattery-faktorontleding onderwerp. Hierdie tegniek gebruik slegs die korrelasies tussen twee stelle veranderlikes en ontbloot slegs faktore wat die twee stelle veranderlikes gemeen het. Drie van die faktore van die vyfFaktormodel is blootgele, naamlik Ekstroversie, Neurotisisme en Konsensieusheid. Die Konsensieusheidsfaktor het ook n relatief sterk ongesosialiseerdheidskomponent ingesluit en in hierdie sin is dit soortgelyk aan Eysenck se Psigotisisme faktor. Die resultate ondersteun die konstrukgeldigheid van die Comrey Persoonlikheidskale en die 16 Persoonlikheidsfaktorvraelys. Implikasies vir die ontwerp en validering van persoonlikheidstoetse word bespreek.

This study examined the factors common to the Comrey Personality Scales (CPS; Comrey, 1970) and the 16 Personality Factor Questionnaire Form A (16PF; Cattell, Eber &Tatsuoka, 1970) for a sample of Afrikaans-speaking university students. Both Comrey and Cattell proposed exhaustive taxonomies of personality traits after extensive programmes of factor analytic research. However, the two systems show marked di¡erences. Cattell claims to have identi¢ed about twenty fundamental personality factors (Smith, 1988), while Comrey (1995) recognises only eight basic personality factors. Boyle (1989) suggested that these di¡erences may be a re£ection of di¡erent levels of explanation in the two systems, with Cattell's factors lying on the ¢rst order level and Comrey's factors on the second order level.
Two joint factor analyses of the CPS and 16PF in Australia (Noller, Law & Comrey, 1987 1 ) and Israel (Montag & Comrey, 1990), respectively, produced solutions in which between ¢ve and seven factors provided the best theoretical and statistical ¢t to the observed data. At least four of the factors identi¢ed in these studies appeared similar to the factors of the ¢ve factor model of personality.The ¢ve factor model has strongly emerged as a comprehensive and integrative higher order model of the structure of personality in the last decade (e.g. Digman, 1990;John, 1990;McCrae,1989). Although di¡erent names have been attached to the Big Five factors, they are commonly referred to as Extroversion, Neuroticism, Conscientiousness, Agreeableness and Openness to Experience (or Culture or Intellect).
Another prominent model of the structure of personality is that proposed by Eysenck (Eysenck & Eysenck, 1985). This model postulates three orthogonal basic trait factors, namely Extroversion, Neuroticism and Psychoticism. The ¢rst two factors correspond to the Extroversion and Neuroticism factors of the ¢ve factor model and were also found by Noller et al. (1987) and Montag and Comrey (1990). Eysenck (1991) believes that Agreeableness and Conscientiousness of the ¢ve factor model are components of his broader Psychoticism factor and that they should not be regarded as basic trait factors in their own right (however, also see Costa & McCrae, 1992). Boyle (1989) re-analysed Noller et al. 's (1987) data and concluded that they extracted too many factors. He noted that his reanalysis provided strong support for the validity of ¢ve of Cattell's second order trait factors, namely Extroversion, Anxiety,Tough Poise, Independence and Control.
The goal of this study was to examine whether factors similar to those reported in the studies of Noller et al. (1987) and Montag and Comrey (1990) could be identi¢ed in an Afrikaansspeaking South African sample using a method speci¢cally designed to uncover only factors that two batteries of tests have in common, namely inter-battery factor analysis (Browne, 1979). A further goal was to relate the factors found in this study to the factors of Eysenck and those of the ¢ve factor model.

Participants and procedure
The CPS and 16PF were completed by 700 Afrikaansspeaking ¢rst-year university students. The participants consisted of approximately an equal number of men and women, with an average age of 18 years.

Instruments
The traits measured by the Comrey Personality Scales and the 16PF respectively, are listed and described in Table 1. This table also includes the internal consistency reliability coe⁄cients computed for the present sample.
The manual of the 16PF contains a wealth of information re-  (Cattell et al., 1970). Unfortunately there is no consensus on the factorial validity of the scales, with many authors suggesting that there are less than 16 factors underlying the total set of items (e.g. Barrett & Kline, 1982;Howarth, 1976;Matthews, 1989, McKenzie, 1988. In addition it can be seen from Table 1 that the reliabilities of the scales appear to be unsatisfactory. However, Cattell (e.g. 1973) has consistently argued (in contrast to most psychometricians) that internal consistency is not a necessary requirement for a good psychometric instrument.

Data analysis
The present study employed the maximum-likelihood interbattery factor analysis of Browne (1979). The aim of this technique is to identify those factors that are common to two sets of variables, by factor analysing only the correlations between the two sets of variables. Only those factors that are common to both batteries are revealed. Factors that are present in only one of the batteries remain uncovered (Panter,Tanaka & Hoyle,1994). Figure 1 shows that the intercorrelation matrix of two sets of variables can be divided into four quadrants. In an interbattery factor analysis only the upper right or lower left quadrants are analysed (Gorsuch, 1983).

RESULTS
Eight, seven, six and ¢ve factor solutions were obtained. The varimax rotated six factor solution provided the most parsimonious and theoretically meaningful solution. Although the chi-square for the residuals was signi¢cant ( 2 = 37,266, df = 18, p 5 0,05), theTucker-Lewis reliability coe⁄cient suggested that in practical terms the six factor solution provided an acceptable ¢t with the data (TLI = 0,98). Table 2 contains the varimax rotated factor matrix. Next, each of the factors are discussed.   (Krug & Johns, 1986). The high loading of the CPS E scale (0,86) supports the conclusions of Montag and Comrey (1990) and Noller et al. (1987) that it is equivalent to Cattell's second order Extroversion factor. This factor matches the Extroversion factors of Eysenck and the ¢ve factor model. The loading of the CPS A scale suggests that people scoring high in extroversion are more active than people whose scores lie in the direction of introversion. This pattern of 16PF scales is clearly recognisable as Cattell's well-known second order Anxiety factor (Krug & Johns, 1986). The high loading (0,71) of the CPS S scale suggests that it is a good indicator of Cattell's second order Anxiety factor. This factor can be regarded as similar to the Neuroticism factor of the ¢ve factor model and Eysenck's model. The loading of the CPS Tscale suggests that people with high scores for this factor are suspicious and hostile toward others, while the loading of the CPS A scale indicates that emotionally stable individuals have more energy than individuals with low scores for emotional stability. In the last place the loading of the CPS M scale suggests that tough-minded people are more emotionally stable than sensitive people. Montag and Comrey (1990) and Noller et al. (1987) reported similar factors to the one found in this study. This factor was de¢ned by scales related to morality and the adherence to rules (16PF G, CPS C), orderliness and meticulousness (CPS O), energy, stamina and the will to excel (CPS A), and to a lesser degree, empathy (CPS P). High scorers will probably be hard-working and reliable, strive to excel, be cooperative and conforming, and be concerned with the well-being of others. This factor appears similar to the Conscientiousness factor of the ¢ve factor model, but it also includes a strong conformity component and an empathy component. In this regard, this factor also appears similar to the opposite pole of Eysenck's (1992) broad Psychoticism factor (high scorers on the Psychoticism factor typically are unorderly, unreliable, non-conforming and egocentric). In his re-analysis of the Noller et al. (1987) data, Boyle (1989) found a similar factor to the one reported here and he equated it with Cattell's Control/Superego second order factor. Eysenck and Eysenck (1985) pointed out that Cattell's second order Superego factor 'looks very much like psychoticism? (p. 124).
The scales with signi¢cant loadings on this factor are related to interpersonal warmth (16PF A), assertiveness and dominance (16PF E), experimentation, a critical attitude and radicalism (16PF Q1, CPS C), trust (CPS T) and masculinity and toughmindedness (CPS M). High scorers will probably be cold, critical, rebellious, dominant, unemotional and suspicious. In addition high scorers may be open to new and alternative ideas that deviate from the accepted norm. It therefore appears that this factor contains elements of Agreeableness and Openness to Experience. Montag and Comrey (1990) and Noller et al. (1987) reported similar factors to the one reported here. Montag and Comrey (1990) commented that this factor matched Cattell's second order Independence factor.
This factor is not interpreted because it is only weakly de¢ned by two scales. It is also not clear what the 16PF N and CPS O scales have in common. However, it was necessary to extract this factor because factors four and six were forced together when only ¢ve factors were extracted.
The scales that de¢ned this factor are related to sensitivity (16PF I, CPS M), imaginativeness (16PF M), and empathy (CPS P). This factor is similar to Cattell's Tough-mindedness or Tough Poise second order factor (Krug & Johns, 1986). This factor also corresponds with elements of the Agreeableness and Openness to Experience factors of the ¢ve factor model. None of the factors reported by Montag and Comrey (1990) or Noller et al. (1987) corresponded directly with this factor.

Discussion
As stated in the introduction, the goals of this study was twofold. The ¢rst was to compare the factors of the present study to those reported by Montag and Comrey (1990) and Noller et al. (1987).The second goal was to compare the factors found in the present study with the factors of the ¢ve factor model and Eysenck's three factor model.
The results revealed a strong similarity between four of the factors found in the present study and the factors of the Montag and Comrey (1990) and Noller et al. (1987) studies. These results underline the robustness of the factors, because the analytic techniques and the participants in the studies di¡ered substantially. Montag and Comrey (1990) and Noller et al. (1987) made use of principal factor analysis, while the present study employed an inter-battery factor analysis. Regarding the participants, Montag and Comrey (1990) made use of Israeli driver's licence applicants, Noller et al. (1987) made use of volunteers representative of the Australian population and the present study made use of Afrikaans-speaking university students. The inter-battery factor analysis provided support for the construct validity of the CPS and the 16PF on the second order level. It is clear that the two instruments largely measure the same broad constructs. However, because of the superiour reliability of its subscales and the replicability of its ¢rst order factor structure, the CPS is probably the instrument of choice for research and applied purposes. Next, the factors found in the present study will be compared to the factors of the ¢ve factor model and Eysenck's three factor model.
The Extroversion and Emotional Stability factors found in this study are similar to the Extroversion and Neuroticism factors of the ¢ve factor model and Eysenck's model. In addition, the Conscientiousness/Psychoticism factor found in this study appears similar to the Conscientiousness factor of the ¢ve factor model and Eysenck's Psychoticism factor. However, both the Independence and Tough-mindedness factors found in the present study do not directly match any of the factors of the ¢ve factor model or Eysenck's model.The Independence factor is probably most similar to the Agreeableness factor of the ¢ve factor model (due to the loadings of 16PF A, 16PF E, CPS T, and CPS M), but it also contains an element of openness to change (16PF Q1 and CPS C) that is mostly associated with the Openness to Experience factor of the ¢ve factor model. Likewise theTough-mindedness factor contains elements related to openness to feelings and ideas (16PF I and 16PF M), but also elements related to sensitivity and empathy (16PF I, CPS M and CPS P) that is mostly associated with Agreeableness. Perhaps a hand rotated or Procrustes solution might show that the two factors can be rotated to positions more clearly aligned with Agreeableness and Openness to Experience.
In conclusion, the study provides support for the importance of the Extroversion, Neuroticism and Conscientiousness factors of the ¢ve factor model in an Afrikaans-speaking sample. These factors correspond largely with the traits of Eysenck's three factor model. These higher order traits have now been identi¢ed in several di¡erent countries and in several di¡erent languages (Costa & McCrae, 1992;Eysenck & Eysenck, 1985). The space de¢ned by these three higher order factors provides researchers with a conceptual ?map? of the domain of personality traits and can serve as a guideline for the development of personality measures. By administering new personality questionnaires with established questionnaires, such as the 16PF and CPS, the meaning of the new measure can be determined (at least in part) from the position that it occupies in the three-dimensional factor space de¢ned by Extroversion, Neuroticism and Conscientiousness/Psychoticism. This should lead to greater conceptual clarity over the meaning of the traits that a questionnaire measures. Overlap between trait measures with di¡erent names will be revealed if they occupy the same position in the factor space. Similarly, di¡erences between trait measure that have the same name, but really measure di¡erent constructs will also be revealed if they occupy di¡erent positions in the factor space.