The classification of arid soils is shown to be unsatisfactory using existing classification systems. The argument is made for a comprehensive development of the current systems to provide a single, universal system, which is thus equally applicable to, for example, arid soils, tropical soils, and residual and relic soils. This paper concentrates, however, on the engineering classification requirements for arid soils. Arid soils are defined and their distinctive properties are described. It is established that the most widely used and most generally applicable classification is the Unified Soil Classification system. In the context of this, other commonly used systems are briefly reviewed and the philosophy behind their development is briefly discussed.

The properties that must be taken into account (i.e. described) in a classification that includes arid soils are categorised into three groups and suggestions are made for their measurement. For example mineralogy is of prime importance, but its direct measurement is not generally feasible. Thus simple field or laboratory tests that indicate the effects of different mineralogies are required, with the aim of obviating the need for precise definition but providing the experienced geotechnical engineer with detailed clues to type(s) of mineral present. The primary additional considerations for arid soils concern the structure of the soil and its response to load and water. This is discussed and suggestions for simple tests are made. A brief reference to engineering applications is made, although the point of the classification system is that interpretation of the data for any specific application should be possible and thus the system should be independent of final application.

The paper concludes with brief comments on the expansion of immediate engineering classification to a more general classification, for example via mapping, and thereafter to provide unification of scientific and engineering 100classifications, although no firm conclusions are drawn on these aspects. The aim of the paper is to provide a discussion document for a working group to adopt as a basis for the development of definitive tests and classification categories, and as such is open for comment.