Although considerable experience has been gained in model updating, the critical issues that remain are the choice of parameters and how to deal with ill-conditioning. Although a number of theoretical tools exist to help with both of these tasks, the techniques are advancing by gaining experience with a diverse range of structures. This paper adds to this debate by updating an experimental bridge model with a geometric scale of 1:50 that represents a typical multi-span continuous deck motorway bridge. The bridge has four identical straight spans and an irregular distribution of piers, and the central pier is shorter than the others. Four configurations corresponding to different pier stiffnesses and the inclusion of an isolation-dissipation device were considered. An initial test without the piers present was also performed. The measurement of data in these different configurations allows the model updating to be performed sequentially, where parameters identified in earlier configurations maintain their estimated values in subsequent configurations. This approach means that each configuration has a small number of uncertain parameters to be identified, leading to a set of well-conditioned estimation problems based on predicting four natural frequencies of the structure. The procedure was successful, and all of the measured natural frequencies were estimated accurately with a maximum error of under 2.5%.