Towards authenticity of task in test development

This article describes a working model used to determine the Target Language Use (TLU) (Bachman and Palmer, 1996) in a Language for Specific Purpose (LSP) test project: The Listening Summary Translation Exam in Taiwanese (LSTE/T). The purpose of the LSTE/T was to evaluate the summary translation ability of applicants who want to work as linguists in Law Enforcement Agencies (LEAs) in the USA. The discussion focuses on the authenticity of task (Douglas, 2000) and how it can be achieved by the development of a taxonomy and a structured verification procedure in the process. Authenticity is treated in this article as the cornerstone for the content validity and reliability in this typical LSP test, which claims to serve as an assessment tool in the real world. Explained in detail are the steps followed in the project to create authentic language materials that mirror the linguistic reality facing LEAs. Major steps in the process include: • collecting data from task performers in the TLU domain; • analysing data; • identifying salient features in both form and content; • creating a taxonomy and sample scenarios; • producing simulated conversations based on scenarios; and • creating test tasks based on simulated conversations. A structured verification procedure that was used in the project is charted and discussed to reveal the pros and cons of such an approach. It is argued that, as one type of LSP testing, job-relevant tests are becoming increasingly important and deserve more attention from practitioners in the testing field. This discussion of a working model to determine the TLU in one project will contribute to the understanding of the role of authenticity in test development in general.