Previously developed regression formulae for estimating body mass in dogs and wolves based on cranial and mandibular dimensions are evaluated using modern canid specimens of known weight at death. Some of these equations proved reliable, but others have large standard errors of estimate and likely produce unreliable mass estimates. New sets of equations for estimating body mass in dogs and wolves are produced using our datasets, including a set of equations developed from combining the dog and wolf biometric data into a single population. The resulting regression equations allow body mass to be estimated from a series of cranial and mandibular dimensions with relatively low errors. Further, our datasets include larger numbers of specimens of larger ranges of body mass than in these previous studies. When the equations are applied to a suite of dogs and one wolf from Eastern Siberia, several patterns emerge. First, hunter–gatherers' dogs in this region vary widely in terms of body size, even within a limited geographic area and time period. Some were quite large, similar in size to modern Siberian huskies. Second, pastoralists' dogs are less variable in terms of body mass, but this may reflect the nature of our samples. In particular, pastoralists' dogs nearly all were sacrificed juvenile dogs, some of which appear to have been eaten. These dogs seem to have been approached adult body size when they were selected for sacrifice. Finally, our findings help to highlight the need for further refinement in methods used to study ancient canid remains. Copyright © 2014 John Wiley & Sons, Ltd.