A large, and purportedly increasing, number of research fields in modern science require scholars from more than one discipline to understand their puzzling phenomena. In response, many scholars argue that scientific work needs to become more interdisciplinary, and is indeed becoming so. This paper contributes to our understanding of the evolution of interdisciplinary research in new fields. We explore interdisciplinary co-authorship, co-citation and publication patterns in the recently emergent research field of open source innovation during the first ten years of its existence. Utilizing a database containing 306 core publications and over 10,000 associated reference documents, we find that inquiry shifts from interdisciplinary to multidisciplinary research, and from joint puzzle solving to parallel problem solving, within a very few years after the inception of the field. "High-involvement" forms of interdisciplinary exchange decline faster than "low- involvement" forms. The patterns we find in open source research, we argue, may be quite general. We propose that they are driven by changes in task uncertainty and the ability to modularize research, among other factors. Our findings have important implications for individual scholars, research organizations, and research policy.