Sensitive soils derived from weathered pyroclastic materials have contributed to major landslides in the Bay of Plenty. Sensitive soils have a high ratio of peak to remoulded undrained strength. While it is known that (a) sensitive soils flow once failed, causing long runout distances, and (b) these failures often occur following heavy rainfall, the mechanisms that lead to failure are less understood. The aim of this thesis is to determine static and cyclic failure mechanisms of sensitive soils sampled from the failure scarps of two recent landslides in the Tauranga Region. Revelations about how these soils fail will allow slope stability models to more accurately capture geomechanical behaviour. Recent publications on sensitive soils derived from glacial till materials have indicated that these soils are brittle materials displaying undrained strain softening behaviour, where deviator stress drops significantly following peak stress. Failure is governed by rate dependant, excess pore pressure gradients accumulating during undrained, consolidated triaxial compression (Gylland et al. 2013c; 2014; Thakur et al. 2014). These publications provided a methodological backbone for this thesis. Field methods included geomorphological and stratigraphic site characterisation, and sampling of extra sensitive soil suspected of contributing to failure. Laboratory methods included geotechnical tests (Atterberg Limits, moisture content, bulk and particle density, particle size distribution, and static and cyclic undrained, consolidated triaxial tests). Static triaxial testing was undertaken at a high compression rate of 0.5mm/min to model rapid undrained during slope failure. Different combinations of average and cyclic shear stresses allowed replication of Anderson’s (2015) cyclic contour plot. Shear zone microstructure of failed triaxial samples was analysed using thin section and micro-CT techniques. Two coastal cliff landslide sites were characterised and sampled: (1) a significant landslide at Bramley Drive, Omokoroa, which initially occurred in 1979, with reactivations in 2011 and 2012, and (2) a landslide on the south side of Matua Peninsula, which occurred in 2012. The bowl-shaped landslide crater at Bramley Drive and long runout component of sensitive material are likely due to failure within an over-thickened sequence pyroclastic material (Pahoia Tephras), which initially accumulated in a paleovalley. At Matua, the failure surface was long, slightly rotational, and comprised a sequence of variable sandy lenses and silty clays. Landslide debris comprised remoulded sensitive material underlying intact overlying blocks, indicating failure of a sensitive soil layer at depth. Material sampled at Omokoroa (OM1) was an extra sensitive (St = 15 ± 3) silty CLAY, 19 m from top of profile within Pahoia Tephras. Material at Matua (M1) was an extra sensitive (St = 10 ± 1) silty CLAY, 16 m from the top, within the Matua Subgroup. Clay mineralogy of these soils is known to be various morphologies of hydrated halloysite. Samples from both sites have dominant clay fractions (OM1: clay: 62.6%, silt: 37.3%, sand: 0.1%, M1: clay: 40.1%, silt: 22.3 %, sand: 37.6%). High porosity (OM1: 70% M1:66%), void ratio (OM1: 2.3 M1: 1.8), and moisture content (OM1: 72%, M1: 64%), together with low wet and dry bulk densities (wet b.d: OM1:1320 kgm-3, M1: 1690 kgm-3, dry b.d: OM1: 760 kgm-3, M1: 980 kgm-3), are in keeping with previously published values of halloysite-rich clays derived from pyroclastic material. Atterberg Limits are high for both materials (Liquid limit: OM1: 66 M1: 52, Plastic limit: OM1: 41, M1: 37, Plasticity index: OM1: 25 M1: 15, Liquidity index: OM1: 2.9 M1: 1.8). M1 and OM1 both plot below the A-line in the range of high compressibility silts (MH). M1 and OM1 both have low activity, reflecting the hydrated halloysite composition (OM1: 0.4, M1: 0.4). Static undrained, consolidated triaxial tests show that failure occurs at less than or near to 5% strain for all tests, indicating brittle failure. Two main types of failure mechanisms were recognised from triaxial results. Post failure, type A was characterised by significant strain softening, contractive, left trending p’q’ stress paths, and a rise in global pore pressure after failure. Type B response post-peak deviator stress showed minor to no strain softening, dilative, right trending p’q’ paths, and a drop in global pore water pressure. In general, test rate, confining pressure and material affect the type of failure: higher compression rates and confining pressures correlate with type A failure, whereas the opposite is true for type B failure. Failure modes observed in failed triaxial samples were either wedge or shear, with the exception of M1a (tested at 75 kPa confining pressure) which failed by barrel deformation. Strain softening increased with effective confining pressure (R2= 0.58). Average effective cohesion and friction remain essentially consistent between peak and residual states (OM1: c’f = 26, c’r = 24, φ’f = 31, φ’r=26, M1: c’f = 17, c’r = 17 φ’f = 32, φ’r = 29). Thin sections captured shear zones tested at 240 kPa and 340 kPa (OM1), and 150 kPa and 255 kPa (M1) confining stress. Riedel shears (R and R’) and P shears were observed in all thin sections. Evidence for progressive failure, most notably changes in the abundance and spacing of shears along the same shear zone, was found in both materials. Clay mineral realignment was observed in shear zones. Micro-CT results showed clay matrix material to be denser in shear zones, implying localised contraction of microstructure. I infer that type A failure mechanism is comparable to sensitive soils that Gylland et al. (2013c; 2014) studied. During compression, pore pressure does not have time to dissipate, leading to excess pore pressure gradients, which initiate brittle failure where a release of potential energy results in R shear fractures and R’ fractures which become linked by P shears. Microstructural collapse within these fractures induces further excess pore pressure, liquefying material in shear zones, and resulting in a loss of material resistance as evidenced by the strain softening behaviour observed. Realigned material in shear zones provides a pathway for excess pore pressure to dissipate, finally registering as a rise in pore pressure in the post-peak region. Integrity of cohesive bonds and asperity interaction is preserved during shearing, resulting in little to reduction of c’ and φ’. For type B failure, lower confining pressures and/or test rates mean that pore pressure has ample time to dissipate during compression, so that when the critical state line in p’q’ diagrams is reached, grains interlock, causing pore pressures to drop (dilation). Boulanger & Idriss (2007) conclude that for sensitive materials, it is difficult to assess the strain or ground displacement that will reduce the clay from peak to residual strength during cyclic loading. In this study, I utilised a new geotechnical tool, a cyclic contour plot (Anderson, 2015), that predicts the cycles to failure, and the average shear strain and cyclic shear strain at failure, for combinations of applied average and cyclic shear stresses. Seven samples were tested at different combinations of average and cyclic shear stresses. Tests with high average and low cyclic shear stress applications resulted in progressive, positive strain accumulation. Tests with no average but high cyclic shear stresses resulted in progressive accumulation of strain in both positive and negative directions. In comparison to Drammen Clay (Anderson, 2015), in general, for the same application of average and cyclic shear stress, failure occurs after a lesser number of cycles, but both average and cyclic strain accumulation is lower. Although limited microstructural evidence was analysed, observations tests show similar mechanisms as described above are responsible for failure under cyclic stresses; post-failure strain softening occurs, and excess pore pressure increases. One micro-CT sample of an entire failed sample tested at high (60 kPa) cyclic shear stress and zero average shear stress shows intense contraction in the shear zone. It is likely that following heavy rainfall events, excess pore pressure gradients develop in sensitive material at Bramley Drive and Matua, resulting in localised fracture development. Collapse of the disturbed sensitive soil in developing shear zones releases additional pore water, enhancing pore water pressure gradients and leading to progressive fracture. Ultimately, breakdown of the sensitive material results in liquefaction along a macroscopic failure surface and rafting away overlying material.