The present study addresses some of the problems of traditional item analysis during the construction of attitude instruments by using a convergent analysis of three metric and nonmetric multivariate techniques for item analysis. Employing data from a recent administration of a science anxiety questionnaire, the use of maximum likelihood factor analysis followed by oblique rotations, multidimensional scaling, and hierarchical cluster analysis is shown. A high reliable instrument with equally reliable subscales is developed in the process. The reliability of the instrument and its subscales lie between .88 and .92.