Personality of Mice and Men
Heymans Bulletin HB-81-532-EX, State University of Groningen (RUG), 1981. by Popko P. van der Molen and Agnes A. de Graaf.
PERSONALITY OF MICE AND MEN
(Re-arranging Personality Dimensions in a Six-dimensional Adjective Space)
* The preparation of this paper has been made possible by the kind and patient support of the department of Personality Psychology of the Rijks Universiteit Groningen. For computational assistance the writers are indebted to Frank Brokken, Haring Land, Jos ten Berge and Hans Hommes.
Suggestions for improving a first and second draft were given by Jos ten Berge, Haring Land, Annemarie van der Molen, Hans van de Velde and especially by Boele de Raad. Margi Rothengatter corrected the worst mistakes in our use of the English language.
- 1 Abstract
- 2 Introduction
- 3 Dutch Personality Descriptive Adjectives
- 4 Model of the Dynamics of Personality in Mice and other Socially Living Mammals
- 5 Application of the model of personality dynamics to other species
- 6 Method and Results
- 7 Investigation of the residual scores on Brokken's 6 dimensions
- 8 Constructing adjective scales for four additional dimensions of personality
- 9 Correlations between personality dimensions
- 10 Discussion
- 11 Conclusion
- 12 References
✰✰ <level 2> As Brokken (1980), following Goldberg's (1977) approach, has shown, the greater part of the correlations between personality descriptive adjectives, as used in daily language, may be represented by means of a six-dimensional vector space. In this study the suggestions of Guilford & Hoepfner (1969), Cattell & Sullivan (1962) and Gorsuch (1974) amongst others, have been followed, namely that if generally agreed upon external criteria are available, a criterium rotation may render factors ( -and labels- ) to span the resulting 6-dimensional personality-adjective space in a way more convenient for experimentation and manipulation than a simple (- rotated or unrotated -) principal component solution could possibly be.
Criteria from model of personality trait dimensions and social-role dimensions and their interrelations, as derived from experimental research on animal behaviour, are used in this study to replace 6 orthogonal factors of personality descriptive adjectives by some 9 dimensions, each of which represents either a temperament trait, a social-role dimension, or some sort of skill. The resulting dimensions are correlated in many instances and can easily be fitted in a systematic theory of personality dynamics. It is argued that this finding supports the hypothesis that the major part of the most conspicuous personality dimensions in men is due to the mere fact that humans are socially living mammals.
Finally, some implications are briefly discussed which follow from applying this dynamic model of personality dimensions to human behaviour.
✰✰✰ <level 3> This paper deals with dimensions of interindividual differences in personality or 'traits'. In traditional personality psychological research it has long been prevailent strategy to obtain basic data from individual subject scorings on tests or other variables that are a priori considered relevant for measuring 'traits'. Such a strategy is therefore ultimately bases on 'a priori's' concerning the relevance of certain sets of tests. As a consequence classical personality-psychological research has been strongly criticized, especially concerning these 'a priori's' (e.g. Mischel, 1976; Hogan e.a., 1977). In recent years, some psychologists have tried to avoid the shaky ground of these 'a priori's' by using exhaustive lists of personality-descriptive adjectives and ratings of individuals on these adjectives as basic material (Brokken, 1978; Hofstee, 1977; Goldberg, 1978). The factors resulting from factor analysis on such data can directly be interpreted by their strongest loading adjectives. In fact, such dimensions are vectors in a personality-descriptive adjective-space.
(Somewhat similar approaches have been applied before. Cattell's 16PF (personality factors) were ultimately based on reduction of an exhaustive set of trait names too (Allport and Odbert, 1936; Cattell e.a., 1970; Buss & Poley, 1976, pp. 71-77). In the present study it is prefered to deal directly with a primary, exhaustive and unreduced adjective space, rather than with a third- or fourth- degree derivate of it, especially since computational facilities for handling such enormous sets of data have recently become available.)
Having decided that a particular n-dimensional euclidian space is a proper representation of the considered set of data, another problem arises. The problem of determining which position of personality factors is most convenient. Guilford & Hoepfner (1969) remark: "It should be agreed that the aim of those who apply factor analysis for the purpose of discovering scientific constructs in psychology should be to achieve psychological significant factors, which can be replicated, which fit into systematic psychological theory, and which can be investigated meaningfully by other methods. Only in this way can there be general unambiguous communicability that science requires."
When dealing with a general and rather exhaustive trait space, generally agreed upon external criteria to determine where to put our labels of reference are mostly not available. Therefore internal criteria are used such as. e.g., varimax rotation of principal components. But, as Guilford (1975) states: "It is probably commonly known that the most disturbing deficiency of factor analysis is its indeterminancy - the lack of any completely dependable criterion as to where to place the reference axes. And when mathematical specifications for simple structure are written in the form of analytical rotation models (e.g., varimax or promax), the model may not fit psychological reality."
Were generally agreed upon external criteria available, a criterium rotation might render factors ( -and labels- ) to span the resulting n-dimensional personality space in a way, more convenient for experimentation and manipulation (Guildford & Hoepfner, 1969; Elshout e.a., 1975).
(For a clear demonstration of these principles on relatively simple data, i.e. measures of books or coffee cups, refer to Overall (1964), Cattell and Dickman (1962) and Cattell and Sullivan (1962).)
However, when we reach for such external criteria, some difficulties arise.
Personality dimensions refer to differences in personality. Differences in personality stem from a variety of variance-sources (Cattell, 1950; Hettema, 1967; Smid, 1975). To begin with, differences in personality refer to differences in judgment by other people. Important sources of variance can furthermore be found in the domain of innate properties, the domain of environmental influences on development, the domain of social roles, etc. and at the same time all these sources of personality-variance may be expected to interact with each other.
If a personal model is to be of maximum use from an experimental and pragmatic viewpoint, it would be most convenient if it contained orthogonal factors ( -and labels to describe them- ) each of which relates exclusively to just one of these sources of personality. However, this is impossible for the following reasons.
Suppose for instance that certain innate properties account for an important part of personality-variance (e.g., genetical factors influencing physical strength). Then almost by definition these genetic factors for strength will also influence:
a) the probability that an individual adopts certain social roles;
b) the probability of profiting by learning from or suffering irrepairably from environmental influences;
c) the probability of evoking certain judgments from other people;
When a strong individual has to compete for a dominant position with weak individuals, the chances of success are unequal. The acquired dominant position will determine in its turn part of the individual's behaviour. In the same way, strong individuals will be better able to learn from harsh environmental influences and avoid the suffering of irrepairable damage, thus increasing their chances of becoming more skillful. This in turn determines part of the individual's behaviour. In this way correlations between genetical factors and other types of personality-sources will prevail and the same probably is true mutatis mutandis for other combinations of personality sources.
This implies that it will be impossible to find pure genetical personality factors, pure developmental personality factors, pure social role factors, etc., that vary independently from one another. In other words, having found an n-dimensional trait-space that describes the correlations between personality descriptive adjectives, and having found an efficient description of the n-dimensional space by an orthogonal simple structure treatment of its principal components, we can almost be certain that each of these resulting orthogonal factors contains elements of many classes of personality-sources, therewith rendering the model sub-optimal from an experimental and manipulative point of view.
Perhaps a more attractive approach is one in which external rotation criteria are obtained from experimental data (Armstrong, 1967; Gorsuch, 1974; Guilford & Hoepfner, 1969). If such criteria are available, orthogonal factors from a simple structure procedure may be replaced by (correlated) canonical factors each of which relates to some major (set of) personality sources and which can be experimentally manipulated and verified.
In this paper we will summarily present a model of personality dimensions and personality dynamics, as may be derived from animal and human behavioural studies ('Model of the dynamics of personality dimensions in mice and other socially living mammals'). External criteria, as derived from that model, are used to replace 6 (orthogonal) principal dimensions of the dutch personality descriptive adjective space, as explored and described by Brokken (1978), by correlated dimensions. It will be argued that the resulting dimensions fit more satisfactorily into the above mentioned systematic theory of personality differences and can therefore more readily be employed in meaningful experimentation and manipulation.
Dutch Personality Descriptive Adjectives
✰✰✰ <level 3> Brokken (1978) developed and investigated an exhaustive list of (1204) dutch personality descriptive adjectives. Ratings on these adjectives were obtained from 200 self and 200 partner raters each of whom used the complete list of 1204 adjectives to rate him/her self or a partner.
The ratings on these adjectives were factor analysed and varimax-rotated for self- and for partner-raters separately. Subsequently the two sets of factors were subjected to a two- sided orthogonal Procrustes rotation. After this procedure Brokken found a one-to-one correspondence between six pairs of self- and partner-factors with Phi(φ)-coefficients of about .80 or higher. Cattell's (1952) scree test for determining the number of factors to be maintained also suggested six dimensions to be a useful number.
These six dimensions were interpreted as is shown in table 1.
Brokken (1978, pp. 51-52) comments on the position of his factors: "It should be noted that the labels for the factors are only very general ones, capturing the tendency of meaning of each factor at most. Frequently, rather unexpected adjectives appear which suggests that other, aesthetically more appealing positions of the factors may be found. In the present study, in which the prediction of variables is a central issue, aesthetically appealing positions of the factors are relatively unimportant because such positions do not increase the predictability of other variables by the factors. Furthermore, as three of the five Norman factors are presented in the factors in their current position while another factor (Conscientiousness) is represented by the factors Dominance and Orderliness, it was decided to postpone new rotations to future research." (The Norman factors Brokken refers to in this quotation are 'Surgency (or Extraversion)', 'Agreeableness', 'Conscientiousness (or Dependability)', 'Emotional Stability' and 'Culture'. With these five dimensions of trait ratings Norman (1963) summarized the structure he found in the use of traits for describing oneself and others.)
table 1: Some high and low loading adjectives on Brokken's six orthogonal dimensions
Model of the Dynamics of Personality in Mice and other Socially Living Mammals
✰✰✰ <level 3> Van der Molen (1972, 1981a) collected ethological data on the social behaviour of male house mice, which were living group-wise together for some months in large (150 x 75 cm) observation cages, supplied with food and water ad libitum. The data were analysed by means of an R-type factor analysis in order to be reduced to a limited set of behavioural trait dimensions. In order to rotate the resulting principal components in a useful and meaningful way, rotation criteria were obtained from additional experiments with social roles. These experiments indicated the importance of:
- Three basic types of social roles: α (dominant), β(compliant subordinate) and ω (outcast), i.e. at least two orthogonal dimensions, and
- Interindividual differences in: 'Level of relevant social skills'.
Rotation of the principal components was subsequently performed in such a way that the factors were either interpreted as social- role factors or as within-role variation between individuals. The analysis resulted in two social role factors and two factors of within-role differences between individuals. The two social role factors accounted for the differences between α, β and ω -males and were labeled as 'Dominant versus subordinate' and 'Socially accepted versus outcast'.
The two personality factors of within-role variance were labeled as 'Active versus inactive' and 'Explorative and Self-willed versus Social and Compliant'.
Although a good distinction was thus obtained between role-differences and other individual differences, orthogonality of factors had to be abandoned in favour of operational and functional clarity.
No R-type factor was extracted that could account for the differences in social skill, a dimension which was experimentally shown to be of some importance. Not finding this factor was probably due to the fact that the experimental design in question was poorly suited for measuring differences in skills in so far as they are related to social roles.
In mice, such differences in skill are expressed through differential probabilities of obtaining the most desired social roles, whereas our ethological raw data were obtained within one social role for each individual.
A dynamic model of personality structures was designed to account for the derived behavioural trait dimensions and their intercorrelations, for differences in skill, and for the experimental results of manipulation of social roles (fig. 1). Explaining in detail how this dynamic model was arrived at, is beyond the scope of this paper. A full account of its construction is given elsewhere (Van der Molen, 1981a, ch. VI). Here we shall just indicate the model's most elementary features.
The model allowed for dependency (obliqueness) between dimensions of social behaviour and traits of temperament, thus allowing for functional (cor)relations between these two categories of personality dimensions.
In addition, the axis' of reference in the domain of temperament traits were chosen in such a way, that their significance for the distribution of social roles was clear a-priori.
The model was in accord with the assumption that the inter-individual variance in three independent and genetically determined qualities is of particular importance for the development of the observed behavioural differences between individuals. These congenital quality-dimensions were called 'Basic energy level' (1), 'Basic need for social contact and interaction' (2) and 'Basic need for having one's way' (3). These three dimensions of basic qualities were suggested to represent together the same three-dimensional trait space as the set of three basic qualities called 'Basic activity level' (8), 'Stability' (7) and 'Self-will versus compliance' (6) (figs. 1 & 2).
This latter trait dimension was of particular importance for understanding the relations between social-role dimensions and dimensions of temperament. The 'balance between self-willed and compliant tendencies' (6) was defined as the predisposition to assume an outcast ω -role, rather than a subordinate β -role. This basic trait was considered to be expressed in overt behaviour as the personality dimension 'Explorative versus social' (6). According to our model the basic trait 'Basic activity level' (8) is expressed in overt behaviour as the trait dimension 'Active versus inactive'(8). The basic congenital trait 'Stability'(7) is expressed at the level of overt behaviour as the likelihood of acquiring 'Useful skills'(11) given certain circumstances. But it was also indicated that accidental circumstances (12) do play a crucial role in the acquirement of 'Relevant skills'.
This model also stresses that the trait dimensions at the level of overt social role behaviour, viz. 'Dominance' (9) and 'Being socially accepted versus outcast' (10), mainly depend on accidental social circumstances (12) and that the congenital basic traits determine only to a limited extent an individuals chances of adopting one role or another. Many dynamic interactions between the trait dimensions are operating by way of the individual level of 'Relevant skills' (11). Therefore this 'Relevant skills' dimension is considered to play a central and crucial role in the dynamics of mouse personality .
A further characteristic of the model is its functional description of the interdependence between the various trait dimensions. These interdependencies predict i.a. correlations between the various trait dimensions (table 2).
Application of the model of personality dynamics to other species
Van der Molen (1981a) compared the above described set of behavioural trait dimensions with data from behavioural research on a score of other species. He concludes that these behavioural trait dimensions are also well suited to describe behavioural differences between individuals of other social mammal species. Also in the case of Man, many well established personality-psychological trait dimensions (as found by e.g., Cattell, 1946, 1950; Eysenck, 1953, 1967; Strelau, 1974; Buss e.a., 1973; Buss & Plomin, 1975) and social-psychological trait dimensions (as found by e.g., Leary, 1957; Foa, 1961; Schaefer, 1971; Benjamin, 1974) appear to fit the above described model.
However, some of the trait dimensions which regularly emerge in psychological literature are not represented by that model of mouse behaviour.
One of these dimensions is generally labeled as 'Positive versus Negative', 'Appreciation', 'Social Desirability' or 'Evaluation'. It is often the most conspicuous dimension in trait ratings of social interactive behaviour. In ethologically based research on human behaviour such an evaluation dimension is however not found. Therefore this dimension is only considered to be relevant for research on human behaviour provided it is based on trait ratings and questionnaires.(In the following this dimension shall be indicated as 'Appreciatedness'. It is meant to express the degree of being appreciated by the rater(s) in question.)
Other dimensions which were not represented in Van der Molen's (1981 ) model of mouse behaviour are the interpersonal variation in 'sensitivity' and various dimensions of acquired 'Skills', some of which may be indicated as 'Stability' in one or another (set of) situation(s).
In order to apply the dynamic model of mouse personality to other socially living mammals (including man), these dimensions were added following investigation of their relation with the dimensions already included in the model. This resulted in a hypothetical general model of temperaments and social behaviour and their dynamic interactions. It is beyond the scope of the present paper to review this model in detail, but fig.1 may provide an impression of the dimensions (boxes) and their interrelations (arrows). Fig.2 and fig.3 summarize the causal relationships between traits of temperament and the probability of assuming the various social-role types (the cosines between vectors being proportional to their intercorrelations).
Method and Results
Comparing the dynamic personality model with the correlation structure in the set of 1204 personality descriptive adjectives
✰✰✰ <level 3> Five (out of more than seven) dimensions from Van der Molen's model of personality dynamics are labeled with 20 items each from Brokken's set of 1204 dutch personality descriptive adjectives. No more than five dimensions from the model of personality dynamics were used, in order to be able to check whether substraction of the influence of these five dimensions from Brokken's factors would render a residue containing the remaining dimensions from the model of personality dynamics.
The correlations between the thus constructed five scales are subsequently computed from Brokken's raw material, viz. 400 x 1204 (self and partner) rating. These correlations appear to be congruent with the predictions from the experimental model of personality dynamics.
Brokken (1978) factoranalysed the correlations between his 1204 items and obtained six orthogonal factors of personality adjectives. The influence of the above mentioned five scales is substracted from the scores of Ss. on Brokken's six orthogonal dimensions, and it is shown that the residual scores on Brokken's six dimensions can also be represented by such factors as are predicted by the model of personality dynamics.
Constructing adjective scales for 5 dimensions from Van der Molen's model
✰✰✰ <level 3> In order to check similarity between the dimensions of the dynamic model of personality as presented in paragraph 3 and Brokken's six dimensions of dutch personality descriptive adjectives as described in the last paragraph, five trait dimensions from Van der Molen's model of personality dimensions (fig. 1) were each labeled with 20 adjectives from Brokken's (1978) list (table 4) considered representative of the trait in question. This resulted in five scales of 20 adjectives each. (For a justification of this number of 20, see the additional remarks in the note on page 20 ). No more than five (out of more than seven) personality dimensions from the model of personality dynamics were used - as explained in paragraph 2 - in order to subtract the 'influence' of these five scales from Brokken's six dimensions and subsequently to check whether the residue of his data did indeed contain the remaining dimensions from the dynamic model.
(That the most conspicuous dimensions of personality ratings in the domain of Dutch personality descriptive adjectives are comparable to those in the American language may be gathered also from Hofstee's comparison of Brokken's and Norman's data. Hofstee (1977) reduced the number of adjectives (1204) to some 400 and combined them into small clusters (scales) of about three adjectives. The scores of the subjects on these clusters were calculated from Brokken's materials and subsequently used as basic data for a factoranalytic procedure. Thus he was able to produce in the ratings which were used in Brokken's study, a remarkable replication of the five Norman factors.)
table 4: Adjective labels for five dimensisons of Van der Molen's model of personality dynamics (of each scale of 20 Dutch adjectives, 10 representative items have been translated)
|+ zelfverzekerd||(self confident)|
|+ zelfbewust||(self assured)|
|+ makkelijk||(easy going)|
|+ overbeleefd||(too polite)|
The scores of Brokken's 400 subjects on the thus constructed five scales and the correlation between the scales were calculated from his raw data. These correlations were compared with the predictions from the dynamic model (see table 1). As table 3 shows, the correlations between the adjective scales were reasonably well in accordance with the predictions from the dynamic model of personality traits. This indicates that the process of labeling the dimensions of our dynamic personality model with dutch personality descriptive adjectives did not seriously violate the correlation-structure between these adjectives as established in the dutch language.
|Dominant versus Subordinate||(9)|
|Accepted versus Outcast||(10)||(0) -0.19|
|Activity, Energy||(1,8)||(+) +0.60||(0) +0.10|
|Explorative and Self-willed versus Social and Compliant||(6)||(+) +0.22||(--) -0.53||(0) +0.12|
|Relevant Skills||(11)||(+) +0.46||(+) +0.05||(+) +0.62||(0) +0.35|
(One of the most conspicuous differences between the predictions in fig. 2. and the calculated correlations between the adjective-scales is the correlation between 'Dominance' (9) and 'Explorative versus Social' (6), which appeared to be somewhat low ( ̂ϱ((9),(6)) = + 0.22). This is probably due to the fact that in mice the exploration- and patrolling-activity is a typical α-job, rendering a high species-specific correlation between those two traits. This might not however be the case in man. Besides, the adjective labels for the scale 'Explorative and Self-willed versus Social and Compliant' (6) are aimed especially at that basic trait ( = predilections on the level of temperaments), thus reducing correlations, if any, between this trait-scale and actually assumed social roles. The lower than expected correlation between 'Skills'(11) and 'Acceptedness1(10) is due to the specific aspects of skills chosen here. Other skill dimensions (see table 7) appear to be correlated more strongly with 'Acceptedness'(10).)
A multiple regression analysis was carried out between each of Brokken's six adjective-factors and the above mentioned scales, viz. :
— 'Activity, Energy',
— 'Explorative, (Thing-oriented,) self-willed versus Social, Compliant',
— 'Relevant skills (in general)',
— 'Dominant versus Subordinate', and
— 'Accepted versus Outcast'; (see table 8).
The five scales of 20 adjectives each, appeared to account for 44% (= mean squared multiple correlation or (non-symmetric) index of redundancy; Steward & Love, 1968) of the variance of the scores on Brokken's six dimensions.
Investigation of the residual scores on Brokken's 6 dimensions
✰✰✰ <level 3> In order to investigate the part of the score variance on Brokken's six factors which had not been accounted for by these five scales, the 'influence' of the five scales on Brokken's six dimensions was partialed out and the resulting residual scores on these six dimensions of Brokken were factoranalysed.
If the hypothesis of similarity between Brokken's adjective space and Van der Molen's personality dimensions holds, it should be possible to predict from Van der Molen's model, which conspicuous personality-dimensions were left in the residue of Brokken's dimensions after substractions of the influence of the five scales.
According to the model of personality dynamics (fig. 1) the former five scales did not account for the trait dimension 'Sensitivity versus basic stability' (5,7) and for the evaluation-dimension 'Appreciatedness' (13). It furthermore predicts that a variety of 'Skill'-dimensions could probably be discerned in detailed analyses. The following dimensions of personality would therefore be expected in the residual scores:
-- 'Sensitivity versus Basic Stability';
-- 'Appreciatedness'-, 'Positive versus Negative'-, or 'Evaluation'-dimension;
-- Some dimensions of 'Skills'.
After carrying out factor analysis of the residual scores as indicated above, the highest and lowest loading adjectives for each of six principal components were investigated, which rendered interpretations as shown in table 5.
table 5: Percentages of explained variance and interpretation of the six unrotated principal components of the residual scores
|% explained variance;||and interpretation|
|1.||31.3||sensitivity vs stability|
|4.||11.8||composure, self-possession (introvert stability and introvert skills)|
|6.||3.7||strength of will (extravert stability and extravert skill)|
The third, fourth and sixth principal component of the residue ('Orderliness', 'Self-possession' and Strength of Will') may be considered as specific dimensions of skills. The fifth principal component of the residue ('Male-female Stereotype') is the only one which was not predicted from fig. 1. But its interpretation is self-evident. The fifth and sixth principal components may hardly be considered relevant. The percentages of variance they accounted for, were 5.2 % and 3.7 % respectively. These two dimensions accounted together for about 9 % of the residual variance and for roughly an extra 3 % of the total variance in Brokken's six orthogonal dimensions. We might therefore choose to restrict ourselves to four principal components of the residual scores, viz. :
-- 'Composure, Self-possession' (= Introvert Stability and Skill)
These interpretations of the principal components of the residue accord with the predictions and consequently support the hypothesis that Van der Molen's (1981 ) model of personality dynamics is to a great extent congruent with the correlation structure in the exhaustive set of Dutch personality descriptive adjectives as described by Brokken (1978).
(Another argument in favour of omitting the last two principal components is the fact that their adjective-scales correlate with the other dimensions in a similar way as the scales for 'Self- Will' (6) and 'General Skills' (11) respectively (refer to table 7 ). This suggests that the last two principal components of the residue ('Male-Female Stereotype' and Strength of Will') merely represent modifications of 'Self-will versus Compliance' (6) and 'General Skills' (11), and may therefore - within the six-dimensional adjective space - adequately be represented by the latter two dimensions. Should one prefer to retain all six principal components of the residual scores instead of only four, the fourth and sixth principal component ('Introvert Skills and Introvert Stability' and 'Extravert Skills and Extravert Stability') could be rotated over 45°, thus rendering their sum and difference viz., the dimension 'Stability and Skills' and the dimension 'Extraverted versus Introverted'. The dimension of 'Stability and Skills' is correlated +.64 with the scale 'Skilled' (11) from table 4 and fig.l. The dimension 'Extraverted versus Introverted' is correlated +.62 with the scale 'Active, Energetic' (1,8) and +.48 with the scale for 'Dominant' (9) from table 4 and fig.l. These two dimensions have been included in table 7 and in table 8 in their unrotated (scale 9 and 11) as well as in their rotated (scales 12 and 13) form.)
Constructing adjective scales for four additional dimensions of personality
✰✰✰ <level 3> For each of the six principal components of the residual scores on Brokken's six dimensions a scale was developed, consisting of 20 representative adjectives, chosen from the adjectives with the strongest loadings on the principal component in question.
In order to obtain the highest possible explanatory power of these scales, they were composed of those items from the set of 1204 adjectives, which loaded strongest on the principal components in question. A multiple regression analysis showed that these six scales, together with the original five adjective scales, accounted for 78.5 % of the score variance on Brokken's six dimensions.
Apparent synonyms were subsequently dropped in order to construct scales which could give within 20 adjectives the broadest possible representation of all types of strong loading adjectives for each factor in question. Table 6 contains this second version of the adjective scales for each principal component of the residual scores on Brokken's six factors.
In order to check the extent to which the explanatory power of the scales had been reduced by replacing synonyms (but higher loading) adjectives, another multiple regression analysis was carried out with this second version. This time the six scales, together with the original five adjective scales, accounted for 73 % of the score variance on Brokken's six dimensions, that is (73/78.5) x 100 % = 93 % of the amount of variance that could be accounted for by the version with the highest explanatory power.
table 6: Adjective labels for six principal components of the residual scores on Brokken's six factors (after subtraction of the influence of five scales from Van der Molen's model (of each scale of 20 Dutch adjectives, 10 representative items have been translated)
|- humorloos||(devoid of humour)|
|- slecht||(bad, evil)|
|+ doodnuchter||(quite sober)|
Overlap between Van der Molen's and Brokken's dimensions
✰✰✰ <level 3> As has been observed, the five mutually correlated adjective-scales which were directly derived from the model of personality dynamics accounted for about 44% of the score-variance on Brokken's (1978) first six principal dimensions of personality descriptive adjectives (this is 56% of the score-variance that could be explained by the set of eleven 20-items adjective scales with the highest explanatory power).
Multiple regression analyses showed that by adding adjective scales, derived from the first four principal components of the residue, the amount of explained variance could be raised to 90% of that maximum. And by adding the two scales which were derived from the last two principal components of the residual scores, a total amount of 70% + 3% = 73% of the variance could be explained (this is 93% of the score-variance that can maximally be accounted for by the set of eleven 20-item adjective scales with the highest explanatory power).
Speaking in terms of (abstract) 'personality dimensions' rather than in terms of factors or scales, we might summarize these findings by saying that the five mutually correlated dimensions from Van der Molen's model of personality dynamics accounted for about 56% of Brokken's six orthogonal dimensions of personality descriptive adjectives, and that all personality dimensions from the postulated model of personality dynamics (see figs. 1, 2 and 3), in which 'Relevant Skills' (11) is specified by the three Skill-dimensions 'Skill in General', 'Orderliness', and 'Self-possession', account together for about 90% of Brokken's six dimensions of personality descriptive adjectives; and thus for about 90% of the six most conspicuous orthogonal dimensions of Dutch personality-descriptive adjectives.
(One may wonder why no small adjective scales could be developed with more explanatory power than 78.5 %. To start with, none of the 1204 adjectives loaded stronger than between .40 and .70 on any of Brokken's six dimensions. (His six dimensions accounted roughly for only one third of the score variance of all 1204 personality descriptive adjectives). Fixing the number of adjectives on 20 per scale therefore puts a certain upper limit to the amount of score variance on Brokken's dimensions that those scales can possibly account for. The amount of explained variance can therefore only be raised higher than 78.5 % by increasing the number of adjectives per scale. But for concision's sake we limited ourselves to 20 items per scale.)
Correlations between personality dimensions
✰✰✰ <level 3> Of the six 20-item adjective scales representing the principal components of the residue, and of the former five scales which were derived directly from Van der Molen's (1981) dynamic model, the mutual correlations were calculated and also the correlations of all 1204 adjectives with each of the scales. (In table 8, the 30 highest - and 30 lowest correlated adjectives are listed for each scale). Because of the relative unimportance of the fifth and sixth principal components (scale 10 and 11), the correlations between the first 9 scales are most interesting. (See table 7)
Table 7: Estimated correlation coëfficients between the adjective scales
|Adjective scale number||Dimension number from fig.3-5||Interpretation||(9)||(10)||(1,8)||(6)||(11)||(13)||(5,7)||(11)||(11,7)||(11,7)||(11,7)|
|1||(9)||Ascendancy; Dominance versus Subordinance|
|2||(10)||Acceptedness vs being Outcast||-0.19|
|4||(6)||Explorative and Self-willed; Individualism vs Social Compliance||+0.22||-0.53||+0.12|
|5||Relevant Skills in general||+0.46||+0.05||+0.62||+0.35|
|6||(13)||Positive vs Negative Appreciatedness||+0.21||+0.33||+0.52||+0.06||+0.48|
|7||(5,7)||Oversensitivity vs. Stability||-0.26||+0.05||+0.01||-0.12||-0.13||+0.03|
|9||(11,7)||Composure; Introverted Skills & Stability||+0.01||+0.22||-0.06||+0.30||+0.34||+0.23||-0.37||+0.19|
|10||Masculinity vs Femininity||+0.14||-0.54||-0.15||+0.51||+0.07||-0.16||-0.11||-0.00||+0.10|
|11||(11,7)||Strength of Will; Extraverted Skill & Stability||+0.65||+0.14||+0.76||+0.08||+0.60||+0.52||-0.11||-0.06||+0.11||-0.17|
|13||Extraversion vs Introversion||+0.48||-0.06||+0.62||-0.16||+0.20||+0.22||+0.19||-0.19||-0.66||-0.21||+0.68||+0.02|
The social role dimensions 'Dominance versus Subordinance' (9)and 'Accepted versus Outcast' (10) span a two-dimensional space ( ̂ϱ((9),(10)) = - 0.19) and the Evaluation dimension 'Positive - versus Negative Appreciatedness' (13) is only weakly correlated with 'Dominance' (9) and 'Acceptedness' (10) ( ̂ϱ((9),(13)) = + 0.21; ̂ϱ((10),(13)) = + 0.33; the numbers between brackets and the subscripts refer to the numbers of the dimensions in figs. 1 to 3). Furthermore, the three basic-temperamental-trait dimensions 'Activity, Energy' (1,8), 'Sensitivity' (5) and 'Self-will versus Compliance' (6) span a three-dimensional space: their inter-correlations are negligible ( ̂ϱ((1,8),(5,7)) = - 0.01; ̂ϱ((1,8),(6)) = 0.12; ̂ϱ((5,7),(6)) = - 0.12).
Conspicuous correlations are present however between dimensions from the basic-trait domain on the one hand and dimensions from the social-role domain on the other. The level of 'Activity (Energy)' (1,8)and the level of 'Skills' (11) are positively correlated with 'Dominance' (9), and the level of 'Self-will versus Compliance' (6) is negatively correlated with 'Accepted versus Outcast' (10) ( ̂ϱ((1,8),(9)) = 0.60; ( ̂ϱ((9),(11)) = 0.64; ̂ϱ((6),(10)) = - 0.53) This is in accordance with the causal relations obtained from the dynamic model of personality (fig. 1 to 3).
The Appreciatedness (Positive versus Negative) (13) dimension - like the Dominance (9) dimension - is positively correlated with Activity (Energy) (1,8) and with Skill (11), ( ̂ϱ((1,8),(13)) = 0.52; ̂ϱ((11),(13)) = 0.48). This agrees with Van der Molen's (1981a) proposition that a high level of energy and skills tends to be converted in a dominant position and/or in attaining positive appreciation.
Fig. 1 suggests a strong dependence between the 'Appreciatedness (Positive versus Negative)' (13) dimension and the social role dimension 'Dominance' (9) and 'Accepted versus Outcast' (10). That model does not however predict that these dependencies render strong correlations. The fact is that it also depends on the social role of the rater himself whether the ratee's social-role behaviour is positively or negatively appreciated (Van der Molen, 1981a , chapter eight: α's appreciate β's above ω's, but ω's appreciate ω's above β's, and β's appreciate α's more than ω's do). The correlation between 'Appreciatedness (positive versus Negative)' (13) and 'Dominance' (9) and 'Accepted versus Outcast' (10) are indeed but weak ( ̂ϱ((9),(13)) = 0.21; ̂ϱ((19),(13)) = 0.06)
Finally, the level of 'Skills' (11) is positively correlated with the basic level of 'Activity, Energy' (1,8) ( ̂ϱ((1),(11)) = 0.62) This is also in accord with figs. 1 to 3.
The congruence between these estimated scale-correlations and the predicted relations between the dimensions of figs. 1 to 3, also supports the idea that Brokken's six-dimensional adjective-space may be adequately represented by the correlated personality dimensions from Van der Molen's general model of personality dynamics.
Unambiguity of the axis' of reference
✰✰✰ <level 3> It should furthermore be noted that the interpretation of the residual factors and of the final adjective scales in terms of correlated adjectives appeared to be rather unambiguous in comparison with Brokken's six orthogonal adjective dimensions (see table 7 and 8). This could be an argument - an argument of clarity - for replacing the six orthogonal adjective-dimensions by this set of some nine correlated adjective-dimensions. This can be added to the argument that the latter set allows for functional relationships between these dimensions.
Number and 'raisons d'être' of personality dimensions
✰✰✰ <level 3> Personality dimensions as found in mice were attributed to the fact that mice are a socially living species. The model of personality dynamics as developed for mouse behaviour was, after some refinements and extensions, suited for describing human behaviour of other socially living mammals and human behaviour as well. This suggests that the most conspicuous human personality dimensions are merely a necessary result of Man being a social species. (Van der Molen, 1981 , ch.nine)
This is not to imply that the dynamic model of personality dimensions as depicted in figs. 1 to 3 contains every imaginable trait dimensions, social-role dimensions and skill dimension that could possibly be the necessary result of existing as a social species.
The three genotypic trait dimensions of fig. 1 were derived i.a. from the notion of three basic functional conflicts, viz.:
the conflict between 'Need for Social Contact and Interaction' (2) on the one hand, and 'Self-will' (3) on the other hand;
the varying equilibrium between available 'Energy'(1), and the 'General Level of Needs or Sensitivity' (5);
the varying equilibrium between harsh environmental circumstances and selection for a high 'Basic Level of Activity' and 'General Capacities' on the one hand, and genetic degeneration on the other hand.
Whereas these functional conflicts may be the most conspicuous, a much greater number of such basic conflicting mechanisms may of course be considered. Such considerations hold even a fortiori for the number of 'Skill'-dimensions. Although our original model (fig. 1) contained merely one general dimension of 'Relevant Skills', a minimum number of three Skill-dimensions emerged in comparing the dynamic personality model with Brokken's six-dimensional space of personality descriptive adjectives. In more detailed analyses however, there may be a considerable increase in the number of useful orthogonal dimensions of Specific Skills (Cattell, 1971).
Obliqueness of factors and predictions
✰✰✰ <level 3> Our model contains non-orthogonal personality dimensions. This is a disadvantage if one considers the mathematical elegance and handiness of the model or its concision. We did however maintain oblique factors, in order to enable us to integrate descriptions of functional (causal) relationships between the various personality dimensions. The obliqueness of dimensions suggests in fact many testable hypotheses. The model predicts for instance which personality dimensions can be manipulated and how this could be achieved. For example:
a. The dimension of 'Social Desirability' is considered as a social tool on the - typically human - cognitive level. Since the score on this dimension indicates the relative social position of the rater towards the ratee (attraction or repulsion) the score may be manipulated by simply choosing raters who are in a different social role, relative to the ratee. Whereas low correlations were found between 'Positive versus Negative Appreciatedness' (13) and the social role dimensions 'Dominance' (9) and ' Acceptedness' (10) ( ̂ϱ((9),(13)) = 0.21; ( ̂ϱ((13),(10)) = 0.33)) the model of personality dynamics predicts strong correlations between the 'Appreciatedness' (13) dimension and interactions between the social-role scores of rater and ratee. In other words: the correlation between 'Appreciatedness' and social roles may be low, but that will probably change as soon as the social roles of the raters, relative to the ratee, are also taken into account.
b. The two dimensions 'Dominant versus Subordinate' and 'Accepted versus Outcast' indicate social role types. Social roles are only defined in relation to specific other individuals in specific circumstances. Therefore a completely different score on these dimensions may be obtained simply by having the ratee's rated by raters who know them from completely different areas of social life.
c. Apart from these three dimensions, the dimension of 'Relevant Skills' may also be manipulated. This can be achieved by either enhancing personal growth, or by blocking it through too many or too few difficult experiences (Van der Molen, 1981b).
d. The model also indicates which dimensions are considered as basic personality traits (or 'temperaments') and which dimensions are not. The former traits can by definition not be manipulated. They are however correlated with the traits from the level of social roles and skills because they determine the likelihood that an individual assumes certain roles and acquires certain skills in certain circumstances. Therefore such 'temperament'-traits can only be measured from overt behaviour when corrections are included for actually assumed roles and actually acquired skills.
e. An individual's social-role behaviour depends on the way in which the available social niches (roles, which can possibly be adopted) have been divided and developed in his particular group(s). And his social role influences the expression of his temperament on the level of overt behaviour. When on the other hand a group of individuals is considered, the latter biasing effects are averaged out and therefore it is predicted that differences in average temperamental condition (congenital basic traits) are easier to trace as between-populations and as between-groups-variance than as between-individuals-variance.
'Good/bad' and social-role-reflexes
✰✰✰ <level 3> Most of the dimensions discussed so far relate either to congenital basic traits or to some kind of social role or skill. The only one dimension which does not relate to one or another observable behaviour of the rated person in question is the Evaluation-dimension 'Positive - versus Negative Appreciatedness'. This dimension relates to the attitude of the rater rather than to the attitude of the ratee. It depends on, but is certainly not identical with, the social roles and the (social) skills prevalent between the rater and the ratee.
The 'Appreciatedness'-dimension is a typical human attribute, supported by widespread evidence (Horst, 1968; Benjamin, 1974). When forming an opinion concerning other people, humans apparently focus strongly on this dimension, which serves as a tool for the individual in his struggle to assert his position in the socio-dynamic processes at the level of social roles. Benjamin (1974) points out that classifying the social-role behaviour of Ss. is only possible after correcting the data for the influence of this dimension of 'Appreciation of Social Desirability'. Van der Molen (1981a) postulates that for claiming and sustaining social roles, individuals tend to focus on 'Appreciation' of the ratee instead of on unbiased 'estimation' of a ratee's qualities. Individuals thus activate a mechanism which actually serves to obscure opportunities for pure assessment of other peoples qualities. The tendency to be blinded by 'Appreciation' causes involuntary social reflex mechanism of repulsion and attraction to operate fully, without being disturbed by sober and intelligent use of the human faculty to assess other people's qualities. He concludes: "In this way primitive mechanisms of social selection can operate without being disturbed by our intellectual faculties. They may thus motor migrations and population cycles in man, just as in other socially living mammals."
The ability to avoid being trapped by the tendency to indulge in Positive versus Negative Evaluation therefore is likely to be a prerequisite for manipulating population- and group-cycles at will. And since human population- and group-dynamics tend to be worked out nowadays on the level of (nuclear) war, genocide and economic strangling techniques, the skill of controlling such population-dynamic forces might be a prerequisite for man's survival.
✰✰ <level 2> Personality dimensions can roughly be divided into three categories: traits of temperament; dimensions of skills; and dimensions of social-role behaviour. A positive/negative-evaluation dimension can eventually be added to the latter category.
Which of these categories will primarily be represented in any set of empirical or experimental data, depends on the point of departure and the focus of interest of the research in question. As a consequence, psychological personality dimensions as emerging from different (sub)disciplines often seem rather different or bear no relation to one another at all.
Data on personality dimensions from the whole area of psychological research may however be integrated into a functionally meaningful model in the following way.
1. In the first place one has to allow for dependency (obliqueness) between dimensions of social behaviour and traits of temperament, thus allowing for functional (cor)relations between these categories proper.
2. In the second place one has to select axis' of reference in the domain of temperament traits in such a way, that their significance for the distribution of social roles is clear a-priori.
To that end one may select the dimensions of social-role behaviour in such a way that the axis' of reference primarily describe:
- the differences between dominant and subordinate behaviour,
- the differences between 'incrowd'- and 'outcast'-behaviour,
- and in the case of ordering data from trait ratings one may add a dimension which describes differences in the degree of being appreciated positively/negatively.
A functional relation between the domain of social-role behaviour and the domain of temperament traits may subsequently be expected if one of the trait dimensions is defined as the predisposition to assume an outcast ω-role rather than a subordinate β-role, or as the 'Balance between self-willed and compliant tendencies'. This trait dimension, and the two dimensions 'Basic energy level' and 'Basic sensitivity or reactivity level' together span a three-dimensional trait space. This set of three orthogonal axis' of reference may be considered equivalent to the set of three orthogonal temperament traits 'Basic activity level', 'Basic stability level' and 'Self-willed versus compliant tendencies', the latter set being a rotation of the former.
Given some tendency to strive for control and social dominance within each individual, a positive correlation may be expected between the 'likelihood of acquiring dominance' and the temperamental conditions 'high basic activity' and 'high basic stability', and similarly between the 'likelihood of acquiring dominance' and a 'high level of acquired skills'.
Whereas the thus established three-dimensional trait space may be represented by a set of three orthogonal personality-trait dimensions - as may the space of social role differences - personality dimensions of the trait category can in general not be considered orthogonal to personality dimensions of the social-role category.
Interpersonal differences in skill levels can be viewed as intermediates between trait differences and social role differences. Skill-dimensions may subsequently be expected to correlate with personality dimensions of both the trait- and the social-role-category.
It was shown in this paper that if one integrates personality dimensions from the realm of temperament traits, from the realm of social-role behaviour, and from the realm of skill-dimensions, in the above described way, the resulting model of personality dynamics is consistent with common sense knowledge of personality, as sedimented in the day to day use of personality descriptive adjectives.
table 8: Strongest correlated adjectives for each of the 11 final 20-item adjective scales (Scale numbers (between brackets and bold) refer to the dimension-numbers in fig.1 and fig.2) (Scales 12 and 13 are a rotated version of scale 9 and 11. The scores on the first 9 scales account for about 90% of the score-variance on Brokken's six dimensions of personality descriptive adjectives that can maximally be accounted for by 11 adjective scales of 20 items each)
|blue||timid, bashful, shy||-0.59|
|schuw||withdrawn, retiring, shy||-0.58|
|dwars||obstinate, contrary, cross-grained||-0.54|
|slecht||bad, evil, wicked||-0.49|
|achterbaks||double-faced, secretive, sneaky||-0.46|
|doodnuchter||down to earth||-0.40|
|nalatig||negligent, neglectful, careless||-0.38|
|doodnuchter||down to earth||+0.44|
|bedaard||calm, collected, composed||+0.49|
|gelovig||religious, believing, pious||-0.39|
|vrolijk||gay, cheerful, merry||+0.58|
|neerslachtig||down-hearted, dejected, depressed||-0.66|
|onstandvastig||indetermined, unsteadfast, unstable||-0.56|
|gesloten||reserved, uncommunicative, unapproachable||-0.61|
|teruggetrokken||retiring, withdrawn, aloof||-0.47|
|bedaard||composed, calm, collected||-0.40|
Allport, G.W., Odbert, H.W. (1936): 'Trait names, a psychological study', Psychological Monographs, 47(1).
Armstrong, J. Scott- (1967): 'Derivation of Theory by Means of Factor Analysis' or 'Tom Swift and His Electric Factor Analysis Machine', American Statistician, 21 (5), pp. 17-21.
Benjamin, L. Smith- (1974): 'Structural analysis of social behaviour', Psychological Review, 81, pp. 392-425.
Brokken, F.B. (1978): 'The Language of personality', Thesis, Rijks Universiteit Groningen.
Buss, A.H. Plomin, R., Willerman, L. (1973): 'The inheritance of temperaments', Journal of Personality, 41, pp. 513-524.
Buss, A.H., Plomin, R.A. (1975): 'A temperament theory of personality development', New York: Wiley.
Buss, A.R., Poley, W. (1976): 'Individual Differences Traits and Factors', New York: Gardner Press.
Cattell, R.B. (1946): 'The description and measurement of personality', New York: Yonkers, World Book Comp.
Cattell, R.B. (1950): 'Personality: A systematic Theoretical and Factual Study', New York: Mc Graw-Hill.
Cattell, R.B. (1952): 'Factor analysis', New York, Harper & Bros.
Cattell, R.B., Dickman, K.A. (1962): 'A dynamic model of physical influence demonstrating the necessity of oblique simple structure', Psychological Bulletin, 59, pp. 389-400.
Cattell, R.B., Sullivan, W. (1962): 'The scientific nature of factors: A demonstration through coffee cups', Behavioural Science, 7, pp.184-193.
Cattell, R.B., Eber, H.W., Tatsuoka, M. (1970): 'Handbook for the sixteen personality factor questionnaire', Champaign, 111.: Institute for Personality and Ability Testing.
Cattell, R.B.(1971): 'Abilities: their Structure, Growth and Action', Boston: Houghton Mifflin Cy.
Elshout, J.J., van Hemert, N.A., van Hemert, M. (1975): 'Comment on Horn & Knapp: On the subjective Character of the Empirical Base of Guilford's Structure-of-Intellect Model', Tijdschrift voor Onderwijsresearch, 1(1).
Eysenck, H.J. (1953): 'The structure of human personality', London: Methuen.
Eysenck, H.J. (1967): 'The biological basis of personality' Springfield: Thomas.
Epstein, S. (1979): 'The stability of behaviour: On predicting most of the people much of the time', Journal of Personality and Social Psychology, 37(7), pp. 1097-1126.
Foa, U.G. (1961): 'Convergences in the analysis of the structure of interpersonal behaviour', Psychological Review, 68, pp. 341-353.
Goldberg, L.R. (1978): 'Attribution of Trait-Descriptive Terms to Oneself as Compared to Well liked, Neutral, and Disliked Others', Journal of Personality and Social Psychology, 36 (9), pp. 1012-1028
Gorsuch, R.L. (1974): 'Factor Analysis', W.B. Saunders Cy.
Guilford, J.P., Hoepfner, R. (1969): 'Comparisons of varimax rotations with rotations to theoretical targets', Educational and Psychological Measurement, 29, pp. 3-23.
Guilford, J.P. (1975): Factors and Factors of Personality', Psychological Bulletin, 82(5), pp.802-814.
Hettema, P.J. (1967): 'Trekken, processen en persoonlijkheidstests', Nederlands Tijdschift voor de Psychologie, 22, pp. 618-641.
Hofstee, W.K.B., Tomas, A. (1976): 'Cognates, Cultures and Crap: Attemps to find English Translations for Dutch Personality Descriptive Adjectives', internal report, Groningen: Dept. of General Psychology, Rijks Universiteit Groningen.
Hofstee, W.K.B. (1977): 'Preliminary steps in structuring the set of 1204 Dutch personality descriptive adjectives' Heymans Bulletin HB-77-302-IN, Dept. of Psychology, University of Groningen, Groningen
Hogan, R., De Soto, C.B., Solano, C. (1977): 'Traits, tests and personality research', American Psychologist, 32, pp. 255-264.
Horst, P. (1968): 'Personality: Measurement of dimensions' San Francisco: Jossey-Bass.
Leary, T. (1957): 'Interpersonal diagnosis of personality', New York: Ronald Press.
Mischel, W. (1976):'Personality and assessment', New York, Wiley, 1968, 1976.
Molen, P.P. van der (1972): 'Faktoranalytisch onderzoek naar gedragsvariaties tussen huismuismannetjes', internal report, Dept. of Behavioural Genetics, Rijks Universiteit Groningen, Netherlands.
Molen, P.P. van der (1977): 'Uitspraken over muizen: Een differentieel-ethologische analyse van het sociale gedrag met bijzondere aandacht voor agressie' , Heymans Bulletin HB-77-308-IN, Dept. of Psychology, Rijks Universiteit Groningen.
Molen, P.P. van der (1981a): 'Dimensions and Dynamics of Personality from Mice to Men; A Biologists view on personality', Heymans Bulletin HB-78-401-EX, Psychol. Dept., Rijks Universiteit Groningen.
Molen, P.P. van der (1981b): 'Striving, Playing and Learning' or 'Aggression and the Dynamics of Learning in the Play and Struggle called Life', Heymans Bulletin, Dept. of Psychology, University of Groningen, Netherlands.
Norman, W.T. (1963): 'Toward an adequate taxonomy of personality attributes: Replicated factor structure in peer nomination personality ratings', Journal of Abnormal and Social Psychology, 66, pp. 574-583.
Overall, J.E. (1964): 'Note on the scientific status of factors', Psychological Bulletin, 61, pp. 270-276.
Osgood, C.E., Suci, G.J. and Tannenbaum, P.H. (1957): 'The measurement of meaning', University of Illinois Press, Urbana.
Schaefer, E.S. (1971): 'From circular to spherical conceptual models for parent behaviour and child behaviour', In J.P. Hill (ed.): 'Minnesota Symposia on Child Psychology', 4, Minneapolis: Union of Minnesota Press.
Smid, N. (1975): 'Schets van een persoonlijkheidsleer als psychologie van het oordeel,' Heymans Bulletin HB-75-174-EX,Dept. of Psychology, Rijks Universiteit Groningen.
Steward, D., Love, W. (1968): 'A general canonical correlation index', Psychological Bulletin, 70, pp. 160-163.
Strelau, J. (1974): 'Temperament as an expression of energy level and temporal features of behaviour', Polish Psychological Bulletin, 5, pp. 119-127.