The Upper Permian Castile Formation of the Delaware Basin in northwest Texas and New Mexico consists of up to 600 m of evaporites and is subdivided into units of anhydrite overlain by halite. The Castile Formation has commonly been interpreted as a deep-water, deep-basin deposit in which sediments were laid down in several hundred metres of water or brine. Recent textural observations within anhydrite units, in which the thick-bedded anhydrite horizons have been interpreted as being of shallow-water origin, have challenged this assumption. This geochemical study of the oldest anhydrite unit in the Castile Formation (the Anhydrite 1 Member) attempts to resolve some of the problems regarding brine depth and evolution in the basin. The Anhydrite 1 Member has been subdivided into five major cycles on the basis of the distribution of stratigraphic units of thick-bedded anhydrite. Stable isotopic analyses of sulphur from anhydrite, and oxygen and carbon from calcite show that the basin waters were chemically homogeneous during precipitation of anhydrite, and do not indicate any significant input of meteoric, continental-derived waters. Throughout the section studied progressive enrichment of O-18 upwards within cored intervals indicates continuous evaporation of the water body. Carbon isotopes appear to indicate fluctuations in organic activity within the cycles. Trace elemental analyses of Fe, Mg, Sr, Mn, Al, Ba, Zn, Pb and Cu from the sulphate fraction of the samples show a very high variability. There is a distinct increase in trace elemental abundances at the tops of cycles which may indicate variations in precipitation kinetics. Analyses of texturally defined cycles show that up-core trends for many of the trace elements correlate with changes in delta(18)O, indicating a progressive increase in the influence of evaporation. In addition, cyclical variations in trace elemental composition indicate changes in basin conditions with around a 350-year cyclicity, These changes are independent of delta(18)O values. The geochemical data do not provide conclusive proof of water depth during deposition of the Castile Formation. The data are interpreted as reflecting small-scale changes in conditions of deposition, despite the fact that water input remained essentially constant in terms of chemical composition.