Although thermal methods have been popular and successfully applied in heavy oil recovery, they are often found to be uneconomic or impractical. Therefore, alternative production protocols are being actively pursued and interesting options include water injection and polymer flooding. Indeed, such techniques have been successfully tested in recent laboratory investigations, where X-ray scans performed on homogeneous rock slabs during water flooding experiments have shown evidence of an interesting new phenomenon-post-breakthrough, highly dendritic water fingers have been observed to thicken and coalesce, forming braided water channels that improve sweep efficiency. However, these experimental studies involve displacement mechanisms that are still poorly understood, and so the optimization of this process for eventual field application is still somewhat problematic. Ideally, a combination of two-phase flow experiments and simulations should be put in place to help understand this process more fully. To this end, a fully dynamic network model is described and used to investigate finger thickening during water flooding of extra-heavy oils. The displacement physics has been implemented at the pore scale and this is followed by a successful benchmarking exercise of the numerical simulations against the groundbreaking micromodel experiments reported by Lenormand and co-workers in the 1980s. A range of slab-scale simulations has also been carried out and compared with the corresponding experimental observations. We show that the model is able to replicate finger architectures similar to those observed in the experiments and go on to reproduce and interpret, for the first time to our knowledge, finger thickening following water breakthrough. We note that this phenomenon has been observed here in homogeneous (i.e. un-fractured) media: the presence of fractures could be expected to exacerbate such fingering still further. Finally, we examine the impact of several system parameters, including core length, wettability and injection rate, on the extent and efficiency of the finger swelling phenomenon.