Methods proposed by L. A. Goodman (1987) for analyzing intrinsic properties of cross-classified nominal or ordered categorical variables were used to examine the performance of items measuring psychological adjustment (PA). These methods were applied to 35 3x4 tables cross-classifying 1,158 persons according to their mental health status (MHS) and responses to 35 Likert-type items measuring PA. The subjects included: 562 normal individuals; 336 persons who had recently attempted suicide; and 260 psychiatric inpatients. Items in this study were drawn from the Suicide Probability Scale. Two sets of orthogonal metric scores were derived for categories of MHS and PA variables. Scores in the first set--representing the direction of MHS and PA--were used to assign scale values to categories of the variables, making it possible to examine the consistency between intrinsic and a priori orderings. The second set of scores, representing the strength or intensity of the variables, formed J- or U-shaped curves when mapped onto the first set. These curves were used to distinguish positive and negative PA and MHS categories. For all tables, the intrinsic ordering of MHS categories was consistent with their a priori ordering; this finding was true for only 23 of 35 tables for PA. The location of the zero point indicated that categories representing positive PA varied depending on the content of the item, with some patterns evident across items of similar content. Two tables and 14 graphs are included. (SLD)