The present study argued that the meaning of verbal labels of a Likert-type response scale was affected by the presentation order of the scale labels. It was proposed that subjects tended to choose the first alternative acceptable to them from among the ordered response categories so that a primacy effect was predicted. Findings supported the hypothesis. In addition, this response-order effect interfered with the threshold values, with factor structures estimated by factor analysis based on polychoric correlations, and with the item and person parameters estimated by the graded response model. Practical implications of the response-order effects were discussed.