Scales constructed under procedures and criteria outlined by the various traditional and latent trait methods were examined as to whether they varied in characteristics related to scale quality. Scales were constructed from a common pool of items analyzed in full form according to Likert and a one-parameter Rasch model for non-dichotomous data. Scales were analyzed in dichotomous form for the Guttman, two-parameter Birnbaum, and one-parameter Rasch models. Findings indicate that, for this data set, a traditional method based on item to total score correlation produced a slightly more valid scale. All five method-defined scales, however, were remarkably similar in other characteristics related to scale quality. (Author/PN)