To investigate the effects of wording in attitude test items, a five-point Likert-type rating scale was administered to 173 undergraduate education majors. The test measured attitudes toward college and self, and contained 38 positively-worded items. Thirty-eight negatively-worded items were also written to parallel the positive statements. Correlations between scores on the parallel items were significant for 27 of the 38 item pairs, and ranged from .43 to .01, with a median correlation of .22. The reliabilities of total and subscale scores were above .80, with higher reliabilities noted for the negatively oriented scales. Factor analysis of the 76 items yielded an eight-factor solution; eight of the items did not load significantly on any factor and seven items loaded on two factors. The positive and negative items loaded positively on different factors on four out of eight factors, indicating that the item format influenced student responses. There was little evidence of bipolarity. Negative loadings were present, but for the most part, small. It was concluded that scoring negatively-oriented attitude test items by reversing the scoring procedures used for positive items may not yield a valid measure of the attitudes tested. (GDC)