The extent to which the emergent structure of an organization remains the same, regardless of the turnover of the members, is one of the most interesting questions raised by the analyses of its evolution. This paper uses longitudinal network analysis to provide an answer to this question. Its shows the usefulness of combining both dynamics and comparative statics (here, blockmodelling) in the study of this evolution. An empirical study examines the evolution of the structure of the advice network among judges of the Commercial Court of Paris. The combination of dynamics and comparative statics answers the initial question: Radical structuralism turns out to be wrong. An answer depends on the dimensions of the structure on which observers focus. The pecking order in the advice network remains relatively stable, regardless of members' turnover. However, social differentiation measured in terms of role relationships and division of work shows that the relational structure does not remain the same regardless of members' turnover. Specifically, relational processes within the organization, such as collective learning, impose varying constraints on different kinds of members over time and the overall relational structure reflects these members' changing reactions and responses to these constraints. Since the former vary, so do the latter, and so does the resulting emergent overall structure.