Although most scientific and engineering disciplines view empiricism as a basic aspect of their discipline, that view has not been the tradition in software engineering. There should be a symbiotic relationship between the development of theories and empirical studies that test and evolve that theory. This talk discusses of the role of empirical study in the understanding and improvement of the software product and process. It offers a personal, historical perspective of the use of empiricism through a series of example applications that demonstrate the various roles that empiricism can play. The examples are taken from the author?s own experience and include the use of empirical study to improve an organization's product quality and productivity in the Software Engineering Laboratory at NASA/Goddard and to evaluate and evolve the effectiveness of various software technologies for use in improving the dependability of software. It also suggests research directions for empirical study in software engineering.