In cancellous bone testing of cored samples, side artifacts are the underestimation of the true (i.e. in situ) mechanical properties due to the severing of the trabecular network during specimen preparation. Although other researchers have suggested correction factors derived from finite element method (FEM) models, it is proposed that side effects can be minimized by increasing the specimen diameter. Six different diameter specimens (3.1–10.6 mm), from two different anatomic sites (bovine femoral condyle and bovine lumbar vertebrae), were mechanically tested in elastic tension using an epoxy endcap protocol to eliminate end artifacts. Elastic modulus was found to be significantly affected by diameter in both sites. For example, the 5.1 mm samples underestimated the elastic modulus of the 10.6 mm samples by an average of roughly 20%. Yet no statistical difference was detected between the 8.3 and 10.6 mm samples in either anatomic site, suggesting that 8.3 mm diameter specimens were sufficiently large to avoid side artifacts. FEM models created from micro-CT images reveal that modulus approaches an asymptotic value with increasing diameter, and demonstrate an architecture-dependent drop in modulus at decreasing diameters. These results confirm, both experimentally and numerically, that side effects can be ignored given a suitably large specimen diameter and that this minimum diameter will be dependent on the cancellous architecture. An important implication of the latter result is that specimen diameters must be chosen appropriately when comparing test groups with different architectures (e.g. normal versus osteoporotic) to ensure that the magnitude of side artifacts does not confound the true differences between the groups.