We analyse the effects of agents' decisions on the creation of and reaction to, congestion on a centralised network with a ring-and-hub topology. We take a fixed network model and numerically determine the global transport costs across the network as a function of capacity. These results show that as the capacity of the hub is reduced the system dynamics are driven by an interplay between stable states and critical points. The stable states are studied in detail allowing us to derive an analytic expression for the probability of crowding within the central hub. The analytic solution is in excellent agreement with the numeric results.