In this paper social network terms are applied to conceptualize the intermediary sector of heterogeneous research networks between the institutional contexts of university research and industrial research. White’s notion of ‘commit interface’, and Burt’s elaboration of this notion to account for change in social networks, are used to explain changes in research networks comprising research institutes, agencies and industrial firms. The theory is tested for the case of a subsystem of the Dutch innovation system: university research in biotechnology with applications in the sectors of environmental research, agriculture and health. The analysis reveals that structural autonomous organizational actors in such networks show different types of behavior than predicted in the hypotheses of Burt: some show less role conforming behavior, others however show role conforming behavior, but both tend to have higher survival rates. Moreover, processes of change in the network also result in the occurrence of structural autonomous actors in later stages of the development of the network. White’s notions of differentiation and dependence are used to account for the differences between the attributes of the network studied and the networks initially analyzed by Burt, i.e. economic markets.