Abstract
The purpose of this study is to reveal the nature and the dimensionality of the social desirability factors active in the Y G Personaliity Inventory which is one of the most popular ones in Japan. Twelve scales of the inventory were administered to 200 male and 100 femal college students both in the ordinary and the valuation situations. A correlation matrix of 24 × 24 order was factor-analysed after iterative estimtaions of the communalities by the principal factor method. Ten factors were rotated by Varimax, Promax and then by Rotoplot method. Three of them were considered to be the SD factors which were called the Social Desirabitliy for Emotionality, the Social Desirability for Extraversion-In-troversion and that for Reflectiveness as shown in TABLE 1 and 2. After extracting these SD factors, 7 temperamental factors: Emtional Stability, Dominance, Impulsiveness, Reflectiveness, Fantasticalness, Frustrativeness and Aggressiveness, identified in several previous studies,were found. This fact reveals that in the domain of self diagnosis by personality inventory there is more than one SD factors and there still remains enough variances of the temperamental factors in spite of Edwards' finding that the SD factor was very highly correlated with the test scores for measuring the tempermental traits.
The second-order factor analysis was done upon the correlation matrix of 10 primary factors. Five second-order factors were extracted. One of which was the general factor of the second-order of the Social Desirability which heavily loaded on the three primary SD factors and also upon the primary Emotional Stability factor, and it did not correlate with the other temperamental second-order factors: Emotionlly Stable, Primary Function, Restraint and Manic-Depressive, the last of which was a new factor which divided the emotionality scales of the inventory into two groups of C,I,N and D,O, Co.
As the present analysis was an ordinary common factor analysis by oblique rotation, the factorial structure revealed could not be developed in terms of the structure of the temperamental factors plus the SD factors independent to the formers. The next step of our approach would be to construct the new scales independent to these SD factors by partialing the variances of the SD factors out of the natural responses by testees to the questionnaire items which are considered as a sum of three vectors of “Perception of Self”,“Personal” and “Social Frame for Desirability” as shown in Figure 3. This can be done statistically by the partial correlation method using a modified group principal factor method and also by applying the principle of factortrueness proposed by Cattell and Tsujioka.
