Loess has a metastable structure and undergoes structural hydrocollapse when loaded and wetted, leading to subsidence and damage of overlying structures, which presents enormous engineering problems in many countries in the world. This problem is a focus of much research on both the nature of the phenomenon and the means. of overcoming its negative consequences. However, there remains a need for fundamental experimental and theoretical studies aimed at understanding the many uncertainties involved in the hydrocollapse phenomenon. An experimental programme has been conducted to study the fundamental nature of the hydrocollapse phenomenon in loess, and the primary features of the metastable structure controlling it. High quality disc-shaped specimens were prepared from natural undisturbed, very soft and slightly cemented loess soils using two simple devices which were designed and manufactured in the laboratory. Artificial loess materials were prepared by mixing pure silt particles of different geometrical characteristics (crushed sand, ballotini glass balls) with various clay types (kaolinite, bentonite) to produce mixes with different clay/silt ratios. Soil specimens of metastable structure were created from natural and model loess materials using the air-fall technique, which allowed full control over the critical variables. The index properties of the materials were determined experimentally, together with their geometrical characteristics from scanning electron microscope (SEM) studies. The compressibility characteristics and hydrocollapse behaviour were measured via one-dimensional (oedometer) compression tests. Two testing methods were used: the single oedometer test and the double oedometer test. The results indicated that the experimental approach used in this research can be used successfully to investigate the hydrocollapse problem of loess deposits. The small-clay loess model was constructed and examined. The results proved that this model is valid. The collapse behaviour of reconstituted and undisturbed loess specimens prepared from the same material was qualitatively similar, although the reconstituted materials exhibited greater collapsibility, particularly under low normal effective stresses. Specimens prepared from pure silt revealed very little collapse in spite of relatively high initial voids ratios (e ≤ 0.9). The existence of bonding materials, such as clay minerals at the points of particle contact, is thus apparently essential for hydrocollapse to occur. Specimens prepared from pure silt-bentonite mixtures exhibited low hydrocollapse values in comparison with those for pure silt-kaolinite mixtures. It was observed that there is an optimum clay mineral content for maximum hydrocollapse and it depends on the type of clay mineral and the level of applied stress. Higher values of hydrocollapse were obtained with smooth, spherical glass balls as the silt fraction, thus confirming that the geometrical properties of the silt particles also have a significant effect on the hydrocollapse behaviour of loess deposits. The position and distribution (mixing method) of clay particles inside the soil structure were equally found to have a significant effect on the hydrocollapse behaviour. The index properties of Libyan loess were found to be very similar to those of other loess deposits world-wide. The Tripoli loess can be classified as silty loess with high susceptibility to structural hydrocollapse in the same manner as other better known loess deposits in the world.