For the past three years Harvard has been experimenting with self-paced instruction (SPI) in several sections of its introductory economics course, Economics 10. This paper is a report of the what, how and for of that experiment: what the benefits (and costs) of SPI have been, how SPI changes the learning process, and for whom among our students it has been particularly helpful. The reader will note the lack of the evangelicalism often found in descriptions of educational innovations. This is not intended to suggest that self-paced instruction is without merit; indeed our evidence suggests quite strongly that under the circumstances, the marginal product of SPI is a 10-20 percent increase in scores and that it inspires students to take more courses in economics. However, it is also true that producing the right circumstances is initially rather costly in time and energy; that the distribution of the benefits is not uniform; and that after an initial period of enthusiasm, students are not happier in self-paced courses than in conventional courses. In short situations in which an SPI system would not dominate a conventional course are plausible. I. Some Background on the Harvard Experiment