This is a study of a well-known controversy in computer science, between artificial intelligence and neural networks researchers. It examines claims made by participants in the field that the controversy was triggered, shaped and finally resolved in connection to activities of state research agencies, particularly the US Department of Defense Advanced Research Projects Agency (ARPA). In contrast to these claims, this account concludes that, although ARPA's relations to the controversy among researchers were important, they were indirect. This study thus beings the analytical resources of studies of scientific controversy to bear on a domain which has so far gone largely unexamined in studies of 20th-century controversies - namely, state research policy and management. The account is based on contemporary documents, historical literatuare, and interviews with researchers and ARPA officials.