The purpose of this study is to survery the factorial structure of preference to universities and colleges and its developmental change. Students from the seventh grade to the eleventh grade were asked to rate in five-point rating scale how intensively they desired to enter each of the universities and colleges. The total number of students was 1694. The correlation matrix of each grade was calculated separately. For the eleventh grade, the residual correlation matrix was computed in order to partial out the effect of individual differences on rating tendencies. The residual correlation matrix was factor-analyzed by principal axis method and varimax rotation. In consequence, eight significant factors were obtained, each of which could be tentatively named 'Teachers College', 'Old Sytem Imperial Universities', 'Technical Colleges', 'Colleges of Foreign Languages', 'Medical Colleges', 'New System National Universities', 'Private Universities A' and 'Private Universities B', respectively. Nwxt, differences in factorial structures among the five grades were examined. For that purpose, after each original correlation matrix of each of the four grades other than the eleventh was factoranalyzed by principal axis method, each solution was roated by procrustean transformation in terms of maximizing the similarity of roated solution to 'hypothetical matrix' (8 factor-varimax solution of the eleventh grade). In this roatation, 10 factorprincipal factor solution, instead of 8 factors, was employed as intial matrix, and 10 factor-decoy matrix, instead of the hypothetical matrix, was employed as target matrix, in order to exclude, out of 8 factor-rotaed solution, a factor which reflected the effect of rating tendencies and an unnecessary factor. This decoy matrix was composed of the hypothetical matrix and two column vectors whose elements were all unspecified as hypothetical factor loadings. The resulting factorial structures of each of the four grades were all surprisingly coincident with that of hypothetical matrix, each of which had the same factors with the ones of the eleventh grade. The measures of similarity (Harman, 1967) of the four grades were 0.127, 0.112, 0.107, 0.098 in order of grade. These values also demonstrate that the factorial structures of each of the four grades are all greatly similar to that of the eleventh grade, but, at the same time, that factorial structure of the upper grade is somewhat more similar to the hypothetical matrix, than the lower. Furthermore, methodological investigation as to several procrustean transformations was carried out. First, the procrustean solution based on the residual correlation matrix and the one based on the original correlation matrix were compared with each other, and it turned out that there exist little difference between them so far as the decoy matrix was employed as target matrix. Second, the complete procrustean solution, which employed the hypothetical matrix as target matrix, and the incomplete procrustean solution, which employed the decoy matrix as target matrix, were compared with each other, and the effectiveness of using decoy matrix was demonstrated. And then some theoretical problems such as reason for that effectiveness were discussed.