A comparison of two different methods for setting performance standards for a test with constructed-response items

The trustworthiness of performance standards influences the credibility of criterion-referenced large-scale testing. In this paper, two standard-setting methods are evaluated and compared, when applied to a test with polytomously scored constructed-response items. A version of the Angoff method is chosen as representative of the class of test-centred standard-setting procedures and the borderline-group method represents the class of examinee-centred procedures. The evaluation is based on procedural, internal and external evidence. The results indicate that both methods provide reasonable and trustworthy approaches to standard setting, but also confirm some of the potential problems with these methods.