This study criticizes current approaches to language test development for their restrictive views of how language should be measured. Language tests are based either on an input component (what goes into a test) or an output component (how should the test be measured). An integration of both components on language tests is advocated. The descriptions provided by Stern and Cummins are recommended for addressing the input issues, and measurement theories are suggested as sources for output. However, it is noted that because theories of language proficiency are not complete, many of the available sources may have only preliminary information useful for test construction, and that measurement theories still do not provide sufficient and full direction regarding the psychometric procedures which should be used to deal with communicative and authentic language. As a result, it is suggested that special efforts be made to advance these two components toward their integration in language testing. (MSE)