The behaviour of cohesionless soils is known to be influenced by the method of reconstitution. It is generally accepted in the literature that different reconstitution methods produce samples of varying fabric and thus samples of varying behaviour. Very little evidence has been presented to validate this statement. The main aim of this is thesis is firstly to observe the fabric of in situ and reconstituted gold tailings samples and secondly to investigate the difference in behaviour between these samples at the same state. The investigation focused on testing in situ and reconstituted gold tailings samples obtained from 3 positions on a tailings dam; pond, middle beach and upper beach. Laboratory reconstitution methods included moist tamping and slurry deposition. Fabric analysis involved the use of SEM images to classify the observed differences in the fabric of the undisturbed and reconstituted gold tailings samples. A particle interaction model based on the observed fabric was postulated to explain the differences or similarities in behaviour. The scope of behaviour investigated included sedimentation, collapse and swell, consolidation and compressibility, creep, stiffness and shear behaviour. The fabric analysis indicates that differences in the fabric of undisturbed and reconstituted gold tailings samples are visible. Moist tamping produces an aggregated fabric while slurry deposition yields a homogeneous fabric similar to that of the undisturbed samples. Comparison of behaviour indicates that neither moist tamping nor slurry deposition can replicate the behaviour of the undisturbed sample fully. Consolidation and compression is a function of the fabric while friction angle is independent of the fabric. Available shear strength and liquifaction potential is also affected by the preparation method and the resulting fabric. / Thesis (PhD)--University of Pretoria, 2009. / Civil Engineering / PhD / Unrestricted