To determine whether placement of items on a teacher rating scale affects the factor structure underlying the scale and to determine whether changing the item format alters the ratings given, a 12-item, high-inference student rating scale was developed containing two global items pertaining to overall teacher effectiveness and 10 evaluative items pertaining to more specific teacher characteristics. Four different formats were constructed and randomly distributed to 363 college students within 27 university classes. The results of a principal components analysis suggest that the factor structure underlying the rating instrument is stable and not affected by changes in the placement of items. The nonsignificant results from a multivariate analysis of variance (MANOVA) provide additional support for the stability of the ratings. The MANOVA results in particular suggest that student ratings do not increase or decrease as a result of changes in item placement. The results of both analyses detract from the view that the halo effect may be influenced by the positioning of items. There is thus little reason to believe that preliminary items may set the stage for responses on following items. (RC)