Testing our patience rather than student performance

See all posts

by Julie Greenberg, Senior Policy Director

First off, let's be clear that the staff at NCTQ, most of whom have considerable teaching experience, is probably more supportive than the average teacher for the use of standardized tests in evaluating both student and teacher performance.

However, our patience with testing is being sorely tried, as the states' separate testing fiefdoms demonstrate how strained they are to test with integrity, much less innovate to improve testing.

The latest news about problems caused by poor tests is from New York. According to a newspaper expose of recent score increases on the state's math tests, teachers in the state long ago caught wind of the fact that only a fraction of the topics that are supposed to be taught shows up on the test each year. For example, in 7th grade, just over half of the topics have been tested repeatedly in each of the last four years, with the remainder never having been assessed.

Even the best teachers may question why they would take on the task of teaching the full curriculum when there will be no reward, no matter how well the untested concepts were learned. And what teacher is going to refuse to drill students on those questions that come up year after year, especially when colleagues across the hall are doing so?

A better test design would have a fairly large but random selection of all topics tested each year and each question pulled from a deep test bank, allowing for unpredictable variations in the way the relevant concept is presented.

But buying into those common sense principles isn't in the interest of states, any more than it is in the interest of schools or teachers. Even if states did care as much about test integrity as showing increasing achievement, making better tests is more expensive.

If we are to create better tests that foster and reward good teaching rather than undermining it, one critical component is taking the tests out of the hands of individual states. Oklahoma tried just such a move this year, only to have the bill vetoed by its governor. However, prospects for common state tests are improving with Arne Duncan's announcement that the U.S. Department of Education will be supporting development of assessments tied to common state standards.

Another way to free up more resources for good test design is to control the costs of test administration. To that end, we should improve machine-scoreable testing, a.k.a. multiple choice tests, to keep it a viable alternative to more sophisticated formats such as short answer or essays. To demonstrate just what such an improvement might look like in mathematics or science, we offer a test we've developed that introduces to most Americans a format imported from Japanese assessments. The format requires that test-takers construct numeric responses, rather than allowing the process of elimination or trial and error guesswork possible in the standard multiple-choice test format.

Standardized testing is part and parcel of a good education system. Imperfect as it may be, it has demonstrated its power by making broad ethnic achievement gaps evident and allowing us to compare the performance of our students to those around the world on selected measures. We've got a long way to go, however, before our states produce tests that live up to the real potential of testing.