Abstract Mid-rise to tall mass timber buildings are becoming internationally popular and are required, for instance in the Eurocode or the Australian building code, to be designed against progressive collapse. Designing against progressive collapse is especially important for mass timber buildings as, when compared to reinforced concrete and steel, timber is a more brittle construction material and mass timber buildings are deemed to be more elastic and have limited rotational capacities at the beam-to-column connections. However, whilst the ability of reinforced concrete and steel buildings to resist such an extreme event has been widely researched, limited studies were carried out on mass timber buildings. Their load transfer mechanisms and structural response after the loss of a load-bearing element are currently unclear. Consequently, this paper presents the outcomes of three experimental tests performed on three scaled-down, 2 × 2-bay, post-and-beam mass timber substructures under an edge column removal scenario. The capacity of the 3D substructures to resist progressive collapse was investigated for two types of beam-to-column connectors, namely two tests performed with a connector type commonly used in Australia in mass timber buildings and one test with a proposed novel connector. In the tests, Uniformly Distributed Pressures (UDP) were applied to the floors in two stages: (i) a constant UDP of 4.8 kPa was first applied to the bays not adjacent to the removed column and (ii) an idealised UDP was then increasingly applied to the remaining two bays through a hydraulic jack connected to a six-point loading tree. The load redistribution mechanisms or alternative load paths, structural response and failure modes were recorded and are presented in this paper. Results showed that the applied load was principally transferred to the three columns closest to the removed column and that the Cross Laminated Timber (CLT) panels spanning over two bays were efficient in resisting and transferring the load. The substructure with the proposed novel connector showed an 8.5% increase in capacity and higher ductility than the substructures assembled with the commonly used connector. A simplified theoretical model consistent with the methodology currently used by industry to predict the collapse resistance capacity of post-and-beam mass timber buildings was compared to the test results. The model underpredicted the test capacity by 53%.