Denver's Pay for Performance pilot was the most ambitious experiment in teacher compensation ever attempted. The verdict: test-based for performance doesn't work. Mr. Gratz, who led the research team for the first half of the pilot, reports on the improved plan that grew out of it. TEACHERS IN Denver recently approved a historic revision to the teachers' contract, in which compensation is linked, in part, to student success in the classroom. If voters fund the proposed contract in fall 2005, it could mark a new era in which teachers and communities formally accept a range of indicators that can be used to determine the compensation of individual teachers. Denver's Pay for Performance pilot, from which the new contract plan emerged, was jointly run by the district and the teacher association, supported by the business and philanthropic communities, seriously implemented (though not without problems), and thoughtfully reviewed. It forced significant improvements in the way the district does business -- improvements demonstrated both in aggregate student achievement and in the opinions of participants -- and led to the formation of a joint task force to develop a new plan for teacher compensation. The recently accepted plan is not a test-driven for performance plan, based on the experience of the pilot. Instead, it takes into account both successful teaching and service to the school community, an approach that entails a much broader assessment of performance than was piloted. It addresses problems identified through the pilot as well as long- standing issues and concerns felt in Denver and many other communities. The Denver pilot also demonstrates why, even with thoughtful pilot leadership and broad support, a strict for performance system -- where performance is defined as student achievement -- is an inappropriate model for education. Indeed, Denver's pilot was a resounding success not because it proved the efficacy of for performance in education based on test scores, but because of how it was run, what was learned, and the highly significant vote of teachers to abandon the common steps and lanes approach to compensation in favor of a new four-point strategy. At the same time, the pilot demonstrates how attempts to gauge student achievement with precision or to tie teacher performance to student achievement can create reverse incentives and negative consequences for all parties. Thus the experience of the pilot provides powerful information both about what is useful and about what is not. It thereby opens up new possibilities even as it closes the door on others. I draw the conclusions I express here from research and from my own experience in Denver, leading the research team for the first half of the pilot. A Brief History of Merit Pay and Pay for Performance Variously called or pay for performance, financial incentives have been used in some forms of business for many years. These incentives have also been tried -- without much success -- in school districts dating back nearly 300 years. They revived as recently as the 1980s. Definitions are fluid, but is often seen as a system in which merit is determined by a supervisor, whereas performance is supposed to be a more objective measure. In many instances, however, and performance may both be determined by a test, by some demonstration of teacher skill or knowledge, or by student achievement. No uniform definition exists, so arguments for and against are often confused. Historically, and performance have often been used interchangeably, making it difficult to determine the content of a particular program or experiment by the name used to describe it. As early as 1710, teachers' salaries in parts of England were based on their children's scores on examinations in reading, writing, and arithmetic. This practice had grown by the mid-19th century and was incorporated into the English system of education in 1862 as part of the Revised Education Code, where it remained for more than 30 years. …