Examining validity in a performance test: the listening summary translation exam (LSTE) - Spanish version

This article reports on a project undertaken by the Center for Applied Linguistics (CAL) to develop and validate a criterion-referenced perform ance test of listening summary translation ability for the Federal Bureau of Investigation (FBI).1 The listening summary translation exam (LSTE- Spanish version) is designed to assess ability to comprehend and summarize in written English recorded conversations spoken in Spanish. The language and topics of these conversations are intended to be representative of the conversations which the FBI routinely monitors. Bachman (1991) mentions two fundamental requirements for ensuring the validity of effective language testing procedures. First, the language abilities measured by the test must correspond to those abilities needed to carry out tasks in the target-language use situation. Secondly, features of the test tasks, or test method characteristics, must correspond to critical features of target- language use tasks. Bachman (1990; 1991) has suggested that the degree to which a language test meets these requirements can be assessed by evaluating the situational and interactional authenticity of the test tasks. Although the Bachman framework was not available to us at the time we developed this test (1988-89), we employ the framework here to present the test and to provide evidence of its validity. Evidence of the situational and interactional authenticity of the LSTE-Spanish is offered, as well as for convergent/divergent validity through comparisons of correlations with cri terion measures.