In 1951, it was first suggested by E. M. Anderson that faults in nature will form at specific angles to the maximum principal stress. For most cases, this concept agrees with observations. However, in nature there are some notable exceptions to this rule, with some faults becoming activated in unfavourable orientations. This project was designed to examine conditions that might affect the angle of faulting and the friction coefficient for sliding.Faults that do not conform to generally accepted 'Andersonian' theory include low-angle detachment faults (e.g. Basin and Range Province (USA)), the San Andreas Fault (USA) the Nordfjord Sojn Detachment (Norway), high angle (steeply dipping) normal faults (Britain) and the Zuccale Fault, Italy. Various explanations have been given for sliding on these unfavourably-oriented faults, including high pressure CO2 or water infiltration (Zuccale Fault), low-friction minerals growing in fault gouge (Zuccale fault, San Andreas Fault), stress refraction (San Andreas Fault), reactivation of thrust faults (Nordfjord Sojn Detachment) and fracture under a combination of compression and tension (UK).Experiments were performed in compression and extension using a triaxial deformation rig. Darley Dale sandstone and Pennant sandstone blocks were cut into cores and tested under the following conditions: (a) intact rock failure (dry); (b) sawcut at 35°, 45° and 55° (dry) to the core axis using (i) constant confining pressure, (ii) constant normal stress (iii) constant mean stress; (c) with pore pressure in all three sawcut angles; and (d) with a gouge layer of 70% quartz to 30% kaolinite mixed with 0, 2, 5, 10, 20, 50 and 100% wt of either graphite or talc. Microstructural studies were carried out on deformed samples.These experiments showed that: The angle of failure in extension was 16-20° to the maximum principal stress for Darley Dale sandstone, and 18-21° for Pennant sandstone whereas in compression it was >30°. This angle in extension is lower than expected, but microstructural analysis indicated occurrence of stress refraction, which may help explain this result. The friction coefficient does not appear to change with pre-cut fault angle in dry samples. Pore pressure tests confirmed the general applicability of the law of effective stress, but anomalous apparent reduction of friction coefficient and production of an apparent cohesive strength in pore pressure tests suggested tests should be run slower to avoid disparity between applied pore pressure and true pore pressure in the sample. Attempts to induce hydraulic fractures showed that high overpressures may often be required to do this.Addition of a low-friction phase (talc or graphite) to fault gouge reduced friction by a disproportionately large amount for very planar faults. This was shown to be due to mechanical smearing of the weak phase over the fault plane, increasing its apparent area of coverage.It was concluded that commonly some combination of high fluid pressure in fault planes coupled with low-friction fault gouges may be required to explain slip on natural, unfavourably -oriented faults.