A large body of work argues that scientific research increases the rate of technological advance, and with it economic growth. The precise mechanism through which science accelerates the rate of invention, however, remains an open question. Conceptualizing invention as a combinatorial search process, this paper argues that science alters inventors' search processes, by leading them more directly to useful combinations, eliminating fruitless paths of research, and motivating them to continue even in the face of negative feedback. These mechanisms prove most useful when inventors attempt to combine highly coupled components; therefore, the value of scientific research to invention varies systematically across applications. Empirical analyses of patent data support this thesis. Copyright © 2004 John Wiley & Sons, Ltd.