The “sociotype” construct: Gauging the structure and dynamics of human sociality

Exploring the pertinence of a "sociotype" construct, established along the conceptual chain genotype-phenotype-sociotype, is the essential purpose of the present paper. Further, by following the sociotype’s conceptual guidelines, a new psychometric indicator has been developed in order to gauge the level of social interaction around each individual—the sociotype questionnaire (SOCQ). A first version of this questionnaire has been elaborated by gathering data about the different classes of social bonds (family, friends, acquaintances, and work/study colleagues) in general population and about the dynamic update of these bonds via face-to-face conversation and other modes of interaction. A specific fieldwork was undertaken, involving 1,075 participants, all of them Spanish adults (with diverse social and regional backgrounds). The data obtained were analyzed by means of the correlational method with an analytical cross-sectional design: the number of factors and the consistency and reliability of the resulting scales were evaluated and correlated. The new sociotype indicator resulting from that fieldwork, in spite of its limitations, seems to be valid and reliable, as well as closely associated with widely used metrics of loneliness and psychological distress. It is interesting that the construct noticeably varies throughout the life course and circumstances of individuals, based on their gender and age, and adjusting to the different situations of social networking. This is the first study, to the best of our knowledge, which has tried to reach both a theoretical and an operational formulation of the sociotype construct, by establishing an ad hoc psychometric questionnaire. We think that the information provided by this operational definition opens a new direction of work that could be useful to guide the development and evaluation of programs aimed at improving and strengthening social networking in people at risk, especially for the elderly.


INTRODUCTION
The publication of Psychoticism as a Dimension of Personality (Eysenck and Eysenck, 1976) and the Eysenck Personality Questionnaire (EPQ; Eysenck and Eysenck, 1975) was accompanied by numerous criticisms of the psychometric properties of the P scale (Block, 1977a, b;Bishop, 1977) as well as other aspects of the concept itself; some of these were answered by Claridge and Birchall (1978), Eysenck (1977) and Eysenck and Eysenck (1977). A summary of the available evidence to date on all the points has been presented by Claridge (1981), and there is no intention here to discuss these issues any further. Our intention is to try and improve what are undoubtedly psychometric weaknesses in the P scale of the EPQ, by designing new items, based on the development of the original concept, testing the relevance of these items by new factor-analytic studies and construct improved questionnaires in the hope of improving upon the original version of the P scale.
There are three major faults in the original P scale which, while not apparently interfering too much with its validity in group comparisons, made individual application rather hazardous. The first of these faults is the low reliability of the scale, 0.74 for males and 0.68 for females. (These are  internal reliabilities; test-retest reliabilities are rather higher, viz. 0.83 and 0.71.) The second fault is the low range ofscoring, with means of 3.78 for males and 2.63 for females. The fact that standard deviations were almost identical with means (3.09 and 2.36) indicates the third fault, namely the grossly skewed distribution of scores, which almost resembled a Poissonian distribution. The fact that in spite of these faults the scale was found to behave very consistently and predictably (Claridge, 1981) suggests that the validity was not depressed too much by these psychometric faults, but clearly improvements should be made if possible.

SAMPLES AND METHODS
Two studies were, in fact, carried out to this end. The first (A) used a 90-item questionnaire which contained the EPQ P scale plus possible new P items, together with 12 E and 13 N items selected from the EPQ. There were no L-scale items. The sample tested consisted of 384 males and 290 females, whose ages ranged from 17 to 70 yr, their respective age means being 40.01 _+ 14.39 and 37.99 f 14.47 yr. Sample A was collected by random approach in the street and by house-tohouse circulation of questionnaires. The second sample (B) used a 117-item questionnaire which contained all the 90 items of the EPQ plus almost all the possible new P items used for the questionnaire of Study A. This sample was collected in a more orthodox manner, with groups of students, teachers and other willing and varied Ss being approached to complete the questionnaire and returned by post. There were 408 males and 494 females aged 38.44 + 17.67 and 31.80 + 15.84 yr, respectively, the range being the same as in Study A.
In both studies the items were intercorrelated separately for males and females and the resulting matrices factor-analysed by principal components. In Study A, three factors were extracted and rotated through Varimax and then Promax, these being identified as P, E and N. In Study B, however, four factors were extracted and were similarly rotated using Direct Oblimin rotation. These were clearly identified as P, E, N and the L scale.
Both sets of loadings were scrutinized for suitable P items to improve the original scale. Table 1 gives the 32 items which were finally chosen to make up the P scale of the EPQ-Revised (EPQ-R). It will be seen that only the P loadings are given for Sample A, but all P, E, N and L loadings are given for Sample B. This is because of the few E and N items in Study A, and the complete absence of L items.

Although
Sample A were somewhat older than Sample B (especially the females), and the collection of the data was somewhat different, the P-scale means and standard deviations were pooled. These are given, for different age groups, in Table 2 and contain Ss from Samples A and B combined for P, but only from Sample B for E, N and L. However, just as a matter of interest, Table 3 gives the actual means and standard deviations on P obtained by each of the age groups in Samples A and B, respectively.
However, 13 new P items were incorporated into the P scale which, along with 19 of the original EPQ P scale items now totals 32 items. In addition, two new E items warranted inclusion in the new E scale and one extra N item was added to the N scale, leaving L at the original 21 items.
Reliabilities (cr-coefficients) and intercorrelations of the scales are given in Table 5. It will be noted that separate reliabilities for P are given for Samples A and B. For interest, the correlations between the new P scale (EPQ-R) and that of the EPQ were calculated and the values were 0.88 for males and 0.81 for females (Sample B only).
Finally, it seemed desirable to try to devise a short scale of the EPQ-R for use when time is very limited. Twelve items were chosen from each of the scales and the short scale questionnaire is given in Appendix 2. The means and standard deviations of all the age groups are given in Table 6. Again, as with the longer scale, Table 6 gives the P means and standard deviations of Samples A and B pooled, while Table 7 shows the individual age group means for the separate samples on P. Reliabilities for both samples are given in Table 8 and also intercorrelations of the scales for Sample B. The scoring key on Table 9 gives item numbers to correspond with the 48-item short scale given in Appendix 2.   : 3,7,11,15,19.23. 32,36,44,48 NO: 27,41 0 @I YES: I. 5.9. 13. 17,21,25,30,34. 38.42,46 0 0 YES: 4,16.45 NO: 8,12,20,24.29,33,37,40,47 0 DISCUSSION Since, admittedly, the P scale of the EPQ suffered from several psychometric shortcomings, an attempt has now been made to rectify some of these. The three main areas of criticism have been listed in the Introduction as: (1) low reliability, (2) low range of scoring and (3) grossly skewed distribution.
The reliabilities (see Table 5) are 0.78 for males and 0.76 for females which contrasts with 0.74 for males and 0.68 for females as quoted in the EPQ manual (Eysenck and Eysenck, 1975). Although this is still not as high as the reliabilities achieved for E, N and L, it must be remembered that the P scale taps several different facets (hostility, cruelty, lack of empathy, non-conformism etc.) which may hold reliabilities lower than would be true of a scale like E which comprises largely sociability and activity items only (Cattell and Tsujioka, 1964). At any rate, the new reliabilities are now acceptable and are certainly an improvement on those of the EPQ P scale. The low range of scoring of the 25-item P scale was a great disadvantage in that means of 3.78 + 3.09 for males and 2.63 f 2.36 for females, quoted in the manual of the EPQ, gave very little room for differentiation.
The nature of the scale, unfortunately, is such that most discriminating items are somewhat 'way out' for most people. The original PEN P scale contained even more paranoid and schizoid type items than the EPQ P scale, and some criticism of this latter scale from researchers like Claridge (1981) dwells on the better discrimination of the old PEN P scale as far as schizophrenics are concerned. Hopefully the present changes of the P scale will not further reduce the usefulness of the EPQ-R for work with schizophrenics, but this remains to be seen when the scale is applied to clinical groups in due course. Meanwhile, Table 2, shows means of 7.19 + 4.60 for malesand 5.73 + 3.85 for females; this contrasts with 4.15 + 3.42 for males and 2.97 f 2.59 for females when the same groups are scored for P on the original 25 items of the EPQ. This difference seems a considerable improvement giving greater leeway in scoring. Finally, the question of skewness and kurtosis of distribution was investigated. Since all 25 original EPQ P scale items are contained in the questionnaires used in both Studies A and B, it was possible to compute the moments for this scale as well as for the 32-item new P scale form. Both distributions, of combined Studies A and B Ss, were plotted as histograms and are given in Figs 1 and 2. It can be seen that there is a somewhat improved skewness in the new distribution, especially so for the females, and this is reflected in the drop in both skewness and kurtosis values listed in the figures. It would be unreasonable to expect a dramatic normalizing of the distribution for the P scale, since the scale itself must inevitably, by its very nature, constitute some departure from normality. This is so, not only by reason of the type of items which differentiate high and low P scorers, but also highlights a proven hallmark of high P scorers to be uncooperative as far as helping researchers with questionnaire replies. "Did you mind filling in this form?" gives consistent loadings on P and suggests that high P scorers would be much less likely to complete a personality questionnaire than low P scorers. Thus, however improved the P scale, it seems likely to be doomed to a slightly J-shaped distribution as long as high P scorers are able to decline to cooperate.  Nevertheless, we feel that all these areas of improvement which we set out to tackle, have yielded some success and, hopefully, researchers may try the new questionnaire out in future studies.