in the testing game
is that "language and language ability are abstract theoretical entities" (Oller, Jr. 1987:42).
Because of this abstractness and of the high complexity of language, linguistic theories have flourished and withered often to resurface in modified attire not only as our knowledge of linguistic processes has increased but as our philosophical, social, and political trends have dictated. To go back to the mid 1970s only, tests in education and instruction have emphasized from discrete item tests to integrative tests (such as the cloze), from the viewpoints of structural linguists through those of psycholinguists to those of sociolinguists and communicative buffs.
Today, we still wonder about who should design, develop, and administer tests: the teacher, the school, the school board, or the commercial publisher? What kinds of tests are most profitable not only to the political, social, and administrative structures but to the learners? What purposes do tests serve? How does one evaluate tests (and who should do the evaluating?) in terms of their reliability, validity, effectiveness, efficiency, cost/effectiveness, acceptability, and in their meeting the instructional and/or educational goals? Should tests of foreign-language competency be different in methodological approaches to those of native-language competency? What should we test for in TESOL, anyway? Competency? Proficiency? Fluency? Accuracy? Knowledge about the target-language or knowledge of the language? Or both? Should we really test at all: is it worth the effort and the possibly aleatory results?
Davies (1978) noted a tension between the analytical approach to testing and the integrative one. He cut the Gordian knot wisely by recommending that language testing should be based on a combination of both views. In any case, he argued, no test could reasonably be wholly analytical or integrative.
Oller (1979:212) flails the analytical method:
D...