This situation forces teachers to guess regarding which curricular goals will be tested each year. And, of course, a good deal of inaccurate guessing unavoidably takes place. As a result, many teachers end up emphasizing what isn't tested, and failing to emphasize what actually is tested. In most states, teachers really have no clear idea about what's going to be measured on their state's upcoming accountability tests.
If teachers truly understand the nature of the skills and bodies of knowledge being assessed, then they can teach toward such skills and knowledge rather than toward a test's items. Teaching to a test's items is deplorable; teaching to the skills and knowledge measured by a test's items is admirable.
Next, let's look at the instructional sensitivity of the tests that most advocates of test-based teacher evaluation would have us use. An instructionally sensitive test will identify which students have been well taught and which students haven't. But, at the moment, there is no evidence whatsoever that the tests being touted for test-based teacher evaluation are up to that task.
State accountability tests, the annually administered standardized tests used as part of a state's accountability tests, are accompanied by no evidence -- none at all -- that they can tell the difference between students who have been taught well and those who haven't. That's right, there's no documentation that these annual accountability tests are instructionally sensitive. On the contrary, available evidence suggests that today's state accountability tests are instructionally insensitive.
These tests have been constructed using traditional procedures designed to produce comparative score-interpretations, for example, to allow us to say, "Kelly scored at the 78th percentile, that is, outperformed essentially 78 percent of other test-takers." For such tests to provide these sorts of comparative interpretations, however, it is necessary for the tests to produce a considerable amount of spread in students' total test scores.
But to attain such score-spread, many of the items on state accountability tests end up being linked to students' inherited academic aptitudes, such as a child's innate quantitative potential, or to the socioeconomic status of a student's family. Because inherited aptitudes and family status are nicely distributed variables, test items influenced by these factors tend to create the needed spread in students' test scores. Yet, inherited academic aptitudes and family status reflect what students bring to school, not what they are taught once they get there. Many of today's accountability tests are laden with items tending to make them instructionally insensitive.
Can these two problems be addressed so we can carry out defensible test-based teacher evaluation? Absolutely! Serious efforts can be made to communicate upcoming testing targets to teachers. Solid evidence can be collected to indicate whether a test is, in fact, instructionally sensitive.
Test-based teacher evaluation can be made sensible -- but only if we first let teachers know what's going to be tested, and then make sure the tests we use are suitable for this purpose. Otherwise, with or without federal dollars, test-based teacher evaluation will surely be specious.