What's Wrong with Teaching to the Test?

Abstract

Mr. Posner speculates on the evolution of a species that has been educated only to succeed on standardized tests. IN CALIFORNIA, where I live, and in many other states, the quality of public education -- and by extension the competence of its teachers -- is being measured by students' scores on standardized achievement tests. The pressure on teachers and administrators to improve these scores is enormous. Up until the recent budgetary crises in California, teachers in districts whose scores improved sufficiently (relative to national percentile rankings) were eligible for cash bonuses and extra money for programs. Though these positive incentives are gone, most of the negative incentives continue. Teachers in schools and districts whose scores fail to improve adequately are branded and subjected to various indignities. Schools that continue to fail to improve may be closed, and districts that continue to fail may be subject to state takeover. Opponents of this so-called high-stakes testing complain that such intense pressure causes teachers to devote virtually all classroom time and resources to preparing students for the standardized This phenomenon is called to the test. Proponents of high-stakes testing respond that that is exactly as it be. They argue that the tests measure success in teaching the curriculum and so to the test is to the curriculum. And after all, isn't that what we want teachers to do? I was led to consider this notion while thinking about the accomplishment of a former colleague who recently made a major breakthrough in a famous unsolved problem in mathematics (though he did not arrive at a complete solution). He has been working on this and related problems for more than 25 years, and some of these problems have been under attack for more than a century. I wondered whether the skills and mental processes necessary to attack problems of this magnitude were qualitatively different from those required to solve more routine problems or whether the intellectual requirements were essentially the same but applied over a much longer period. The kinds of problems that can appear on a standardized are, of course, quite limited in form and complexity, as the student is allocated only a minute or two to complete each one. If the intellectual processes required to solve a really complicated problem are not essentially the same as those required to solve these simpler problems, then a student prepared only to solve standardized problems could well lack the mental preparation required to attack really hard problems. Part of my concern about this matter is that routine problems are the most amenable to solution by computer. Thus individuals equipped only with the ability to solve routine problems would be those most vulnerable to displacement by automation. Of course, solving famous unsolved problems in mathematics is a special calling and probably not a reasonable model for what we expect from most of our students. As a model for evaluating whether teachers teach to the test, we use something more typical of the kind of everyday problems that concern us as workers or parents or citizens. But we needn't look far. The very question we are considering -- Should teachers teach to the test? -- strikes me as a typical example. Would the capabilities required to solve problems on standardized tests enable a student to attack this problem? As stated, the problem might seem too vague. The student might well respond, What do you mean by should? But that is the way real problems usually confront us. Should the U.S. invade Iraq? Should I have sex? Should I smoke pot? Should I add this service or feature to my product line? Reducing these more or less vague problems to more concrete questions is a major part of the problem-solving process in the real world. We typically attack such should problems by analyzing the possible consequences of the proposed actions, along with the probabilities of those consequences and their relative costs and benefits. …