The purpose of this study was to determine if correlations among student ratings items designed to be diagnostic could be lowered through use of special instructions to raters. The authors argue that the lowering of inter-item correlations is indicative of a reduction of the halo effect which leads to greater item diagnosticity. The experimental group first ranked items in terms of importance, then rated the course with the diagnostic items, then rated the course with the general items. This order was reversed for the control group. The correlations among items were significantly lower for the experimental group. (Author)