In 1994 the Merriespruit gold tailings dam in South Africa failed, resulting in 17 deaths. The post-failure investigation provided no explanation as to why the catastrophic flow failure, which contradicted all previous experiences of failures of gold tailings dams in South Africa, occurred. The documented history of the dam describes insufficient freeboard provision and often poor pool control, which is argued to have resulted in some areas of the dam having high in situ void ratios. Some of the undrained triaxial tests carried out on specimens obtained from zones adjacent to the failure scar exhibited nondilative behaviour. Laboratory triaxial tests that were conducted on reconstituted specimens and are reported in a companion paper defined a series of steady state lines that were dependent on the particle-size distribution of the tailings. Void ratios obtained from undisturbed samples taken during the post-failure investigation are compared with these steady state lines and it is shown that an appreciable percentage of the specimens were likely to have been contractant. The inference drawn is that a large volume of tailings was in a metastable state in situ and overtopping and erosion of the impoundment wall exposed this material, resulting in static liquefaction of the tailings and a consequent flow failure.