The importance of the presentation order of items on Likert-type scales was studied. It was proposed that subjects tend to choose the first alternative acceptable to them from among the response categories, so that a primacy effect can be predicted. The effects of reversing the order of the response scale on the latent factor structure underlying the items and the item and person parameters estimated by item response theory were also studied. Five Likert-type items, theoretically constituting a Personal Distress Scale, were answered by 102 senior high school students (49 males and 53 females) in Taipei City (Taiwan). Five weeks later, the same students responded to the scale with the ordering of the scale labels reversed. Primacy effects were supported, and the items' underlying factor structures were changed regardless of whether the factor analysis was based on Pearson or polychoric correlations. Estimated item parameters indicated that items with a traditional response scale order (from positive to negative) were "easier" than were those with reversed order. Persons whose parameters were estimated from items in traditional response scale order had higher trait levels than those estimated from items with the reversed order. Thus, primacy effects interfered with the item response theory theoretical property of parameter invariance. Practical implications of the findings are discussed. Five tables present study data, and a 44-item list of references is included. (SLD)