Bromilow's log-log time-cost (BTC) model is tested and refitted with a new set of data for Australian construction projects completed between 1991 and 1998. It is shown that, as anticipated by earlier research, different parameter estimates are needed for different project types, with smaller industrial projects taking less time to complete than the smaller educational and residential projects. This results in the development of two separate models, one for industrial projects and one for non-industrial projects. No changes in parameter estimates are needed for projects with different client sectors, contractor selection methods and contractual arrangements. Alternatives to the log-log model failed to produce any improved fit. Finally, the results are compared with previous work to indicate the extent of changes in time-cost relationships in Australian construction projects over the last 40 years. This indicates a clear improvement in construction speed over the period. Furthermore, the 'public' sector group in particular has exhibited a greater variation (up to 132%) over the years.