Because virtually all tectonophysical processes are marked by the overburden, or occur to slowly for adequate observation in anthropocentric time, or both, they must be studied in carefully controlled laboratory experiments that simulate the natural environment as realistically as is practicable. Extrapolations of laboratory data in space and time are invalid unless the experimental and natural phenomenologies are essentially the same. The size of conventional specimens is of the order of 10 cm, whereas the discontinuities (defects in a continuum) in real rock-masses are often much larger, of the order of 1 m or more. Furthermore, such discontinuities as macrofractures (joints) may well dominate the mechanical and fluid-transport properties in nature. Adequate sampling of rock-mass properties will probably always require in-situ testing, but testing machines much larger than any now available could provide useful data at least at intermediate scale.