Abstract Metastable soils may collapse because of the nature of their fabric. Generally speaking, these soils have porous textures, high void ratios and low densities. They have high apparent strengths at their natural moisture content, but large reductions of void ratio take place upon wetting and, particularly, when they are loaded because bonds between grains break down upon saturation. Worldwide, there is a range of natural soils that are metastable and can collapse, including loess, residual soils derived from the weathering of acid igneous rocks and from volcanic ashes and lavas, rapidly deposited and then desiccated debris flow materials such as some alluvial fans; for example, in semi-arid basins, colluvium from some semi-arid areas and cemented, high salt content soils such as some sabkhas. In addition, some artificial non-engineered fills can also collapse. In the UK, the main type of collapsible soil is loess, though collapsible non-engineered fills also exist. Loess in the UK can be identified from geological maps, but care is needed because it is usually mapped as ‘brickearth’. This is an inappropriate term and it is suggested here that it should be replaced, where the soils consist of loess, by the term ‘loessic brickearth’. Loessic brickearth in the UK is found mainly in the south east, south and south west of England, where thicknesses greater than 1 m are found. Elsewhere, thicknesses are usually less than 1 m and, consequently, of limited engineering significance. There are four steps in dealing with the potential risks to engineering posed by collapsible soils: (1) identification of the presence of a potentially collapsible soil using geological and geomorphological information; (2) classification of the degree of collapsibility, including the use of indirect correlations; (3) quantification of the degree of collapsibility using laboratory and/or in situ testing; (4) improvement of the collapsible soil using a number of engineering options.