ABSTRACT
Most engineered structures are designed with, the designers hope, a “remote” chance of failure. This may correspond to something in the order of 1 in 1,000 to 1 in 10,000 or more. In other cases, for example open pit mines and some forensic work, the engineer can operate at higher probabilities of failure in the order of 1 in 10 to 1 in 3 or so. In geotechnical work it is rare to have large numbers of relevant results from which to judge parameters. A few tens of results is usually as much as is available, and often less. Given this it is questionable whether estimates of reliability are very accurate or useful, especially when the limited number of results are coupled with uncertainties of the scale effect (autocorrelation distance). This paper addresses these issues and provides an assessment of these influences. It is considered that the results confirm what most mainstream practitioners have intuitively assessed about replacement of the conventional factor of safety approach with reliability appraisals in such areas as slope stability in soil slopes. Some areas for the more fruitful application of statistical methods are suggested
