Body mass is a pivotal quantity in palaeobiology but must be estimated from an imperfect fossil record. We analyse the precision of skeletal predictors of mammalian body mass as a mean to inform the Mesozoic mammal record, including a new eutriconodont from North America. We focus on the critical small end of the size spectrum – critical because the earliest mammals were small, because small size persisted onto the stems of the major extant radiations, and because small mammals compose a large proportion of crown diversity. Linear regressions based on extant small mammals indicate a universal correlation of body mass with observed measurements, but with clear differences in precision. Postcranial predictors outperform jaw and dental metrics, with certain femoral joint dimensions providing surprisingly precise estimations. Overall, our data indicate small-mammal evolution during the Mesozoic unfolded in patterns of underappreciated complexity. Studying these dynamics is only possible when estimating body mass within a strict, highly focused phylogenetic context. The heuristic value of the estimators we provide here are not limited to the Mesozoic but are phylogenetically justified for any small-bodied mammal regardless of age.

For extant samples, craniodental measurements were taken using Paleo-Tech Dental Caliper with a Mitutoyo 0-150 mm digital scale to the nearest 0.01mm. Limb measurements and measurements of the specimens from KUNHM were taken with a SPI Dial Caliper Model 31-415-3 to the nearest 0.05 mm. If specimens had both left and right sides, we used measurements from the more complete side and supplemented from the contralateral side if necessary.

For fossil specimens, we recorded measurements from literature. For specimens lacking published measurements of preserved predictors, we used published figures and ImageJ to add measurements where possible. This approach included calibrating the line tool with the published scale bar and then using this tool for measurement. Where possible, we also compared our figure-based measurements with published values to ensure accuracy. Because we could not verify the angle of the limb elements in specimens preserved in slabs, we measured the midshaft diameter of the femur and humerus and included these measures as both anteroposterior and mediolateral diameters for analysis. We acquired measurements of Gobiconodon ostromi from quality casts of MCZ 19965 and MCZ 19860 housed the Center for Functional Anatomy and Evolution at Johns Hopkins University. From these specimens, we added or amended a limited number of postcranial measurements.