An empirical comparison of the validity of three Likert-type attitude scales measuring fourth to sixth grade students' attitudes toward school integration was conducted. The content-parallel forms differed only in terms of item phrasing and were composed of either all positive, all negative or a mix of items. Different factor solutions underlying the scales were hypothesized for the All-Positive, All-Negative and Mixed Forms. The different structures were compared statistically using the LISREL program to determine whether elementary children respond differently to positive and negative phrasing and to examine the utility of alternating item phrasing to enhance validity. Results indicated that the subjects responded differently to the same phrase stated positively and negatively, implying that the technique of balancing item phrasing to control response sets is not effective for younger children. (Author/CM)