Many large-scale connectivity initiatives have been proposed around the world with the aim of maintaining or restoring connectivity to offset the impacts on biodiversity of habitat loss and fragmentation. Frequently, these are based on the requirements of a single focal species of concern, but there is growing attention to identifying connectivity requirements for multi-species assemblages. A number of methods for modelling connectivity have been developed; likewise, different approaches have been used to construct resistance surfaces, the basic input data for connectivity analyses. In this study we modelled connectivity for a multi-species group of vertebrates representative of heavily fragmented forests in north-central Victoria, Australia. For each species, we used least-cost modelling and compared two alternate resistance surfaces, based on species distribution models and on expert opinion, respectively. We integrated the connectivity results across individual species to obtain a multi-species connectivity map for the region. A resistance surface based on expert assessment of the relative use of land-cover classes by the target species was more informative than one based on species distribution models. The former resulted in pathways more strongly aligned with existing patches and strips of native vegetation. In this region, pathways aligned with streams and their associated riparian vegetation have relatively high ecological potential and feasibility to contribute to regional connectivity for the assemblage of forest vertebrates.