A FACTOR ANALYTIC STUDY OF AN ANXIETY SCALE

In the previous study (12), a multidimensional structure of Taylor's Anxiety Scale was suggested by the scale analytic method and at the same time it was discussed that her uni-dimensional scale alone would not be appropriate for the analysis, but that a lot of anxiety items should be newly added to. With these suggestions, the present study was carried out to search for the structure of anxiety measured psychometrically.<br>The investigation was proceeded by the following steps.<br>1) Collection of items: those presumed to identify anxiety were collected referring mainly to MAS, MMPI and Eysenck's MMQ (4), out of which 131 were used as test items.<br>2) Administration of items: the questionnaire of those 131 items was administered with Yatabe-Guilford Personality Inventory to 292 university students under the condition to respond by dichotomy.<br>3) Item analysis and construction of subscales: only those data of 262 male students were analyzed in parallel by the following two operations;<br>(i) collecting items of apparent similarity, and<br>(ii) selecting items of functional similarity such as those comparatively high in reproducibility (Rep).<br>When a subscale of items of apparent and functional similarity was extracted, the same operations were repeated to the remaining items. Thus, six subscales of ten items each were extracted with maximum Rep from .81 to .85 and minimum Rep form .62 to. 71 (Table 1, 2).<br>4) Factor analysis: Pearson's product moment correlations were computed among those six subscales and eight from Yatabe-Guilford Personality Inventory (Table 4). The correlation matrix was factored by Thurstone's complete centroid method and four factors were extracted (Table 5). The centroid factor pattern revealed, among other things, the comparatively low communality of the so-called somatic items (items concerning physical complaints supposedly accompanying anxiety), which was probably related to the comparatively low validity of these items.<br>5) Rotation of axes: both oblique and orthogonal solutions were obtained (Table 6, 7, 8). Interpreting them psychologically, the orthogonal solution was preferred to the oblique solution. That is, according to the former, the present anxiety scale would consist of a “general factor”, with high loadings on all subscales, and three “special factors”, if named, those of “introverted anxiety”, “extraverted anxiety” and “inferiority feelings”. On the other hand, the oblique solution appears difficult to clarify factors C and D. Factor A and B, however, would be interpreted as “introverted anxiety” and “extraverted anxiety” respectively.<br>Through these analyses, it may be concluded that a multidimensional scale is needed for the dynamic study of anxiety.