This paper addresses questions of whether positively- and negatively-worded items measure the same construct and whether the rating scale categories "strongly agree" to "strongly disagree" are used in the same way for both types of items. Item response theory (IRT), specifically the Andrich Rating Scale Model (B. Wright and G. Masters, 1982) is used to analyze a survey on job satisfaction. The methodology presented provides a strategy for exploring the effect of including positively- and negatively-worded items on a Likert-type survey. The Job Satisfaction Survey (JSS) (P. Spector, 1983; 1992) is a 36-item instrument designed to measure job satisfaction, which purports to measure nine aspects of job satisfaction and overall job satisfaction. Seventeen items are positively worded, and 19 are negatively worded. The survey was administered to 706 respondents from diverse work settings. Data were analyzed with the Andrich Rating Scale model. Analysis indicated that the positively and negatively worded items appeared to be measuring the same construct. These results confirm that, for these data, positively and negatively worded items can be scaled together. The IRT analysis provided a method for determining the impact of mixing positively and negatively worded statements on the same scale. (Contains two tables, two figures, and seven references.) (SLD)