Monotonic and cyclic triaxial testing provides key information for a wide range of sensitive
geotechnical problems. This paper assesses the potential impact on stress-strain measure-
ments of several error sources and discusses how test quality may be improved. External vol-
ume gauges are shown to be subject to significant errors that depend on the pressure level.
While high-resolution local radial strain measurement presents considerable challenges,
especially in long-duration cyclic tests, problems with “floating” radial-belt and alternative
“L-configuration” systems were overcome by steps that allow strains as low as 10−4 %to
be resolved reliably. Sample end conditions are shown more important than is commonly ap-
preciated. Employing smooth, enlarged, and lubricated end platens can avoid the recording of
misleadingly high shear resistances, which are most significant with relatively loose specimens
tested under low effective stresses. Stiffnesses and dilation trends were also recorded more
reliably in tests employing smooth, enlarged, and lubricated end platens. The arrangements
overcome significant strain errors even in tests employing local instruments and specimens
with initial height-to-diameter ratios of 2.