Content area
Abstract
The present study describes development of a new measure of prejudice against accented English. Scarce research in this area has been due to methodological preference on experimental designs. However, the self-reported questionnaire development can provide an alternative method to maximize our confidence in past results. Moreover, when employing item response theory (IRT), the major weakness of classical test theory (CTT) can be overcome by applying an appropriate model in order to assess an examinee's true latent trait (i.e., prejudice). Preliminary, participants completed generated items with 7-point Likert-scale online. Items were further analyzed under the most general IRT model, the nominal response model, utilizing its unique ability to examine within-item category functioning. Upon examining individual item as well as its response function, each item yielded its unique reliability that allowed us to construct more reliable and stronger scale. This empirical examination leads establishment of the scale that contains fewer items as and response categories; yet, maintaining the high reliability. Construct validation results supported that the finalized scale correlated positively with xenophobia and anti-immigrant scales, but did not correlate with social desirability scale. Moreover, the scale was found to have modest predictive validity.
