Test decisions over time: tracking validity

Abstract

This study took a grounded theory approach in investigating the relationship between the criteria indigenously drawn for an English for academic purposes (EAP) test and the outcomes of test decisions over time. Test outcomes were examined across three theoretical samples: (1) an EAP teacher-identified sample of misplaced students; (2) a database-generated sample that matched the key variables in the EAP teachers’ sample; and (3) a random sample of test-takers drawn from a single administration of the test. Although analysis revealed two sources of testing error, namely, under-specification of bands in the writing sub-test scale and an under-valuing of the listening sub-test in the overall weighted average of the test, it also revealed the relationship between key patterns of performance on the test and performance in university classrooms that signposted students at risk. As such, the study provides evidence of the usefulness of inquiry that examines the relationship between EAP test performance and the use of English as a mediating tool in academic performance.