The success of ecosystem restoration in semi-arid grasslands is variable and unpredictable. Water retention dikes (7.5-30 cm tall, 50-150 m long) were installed at the Jornada Experimental Range (3-6 dikes/site) in the Chihuahuan Desert from 1975-1981 in an effort to restore ecosystem structure and function. There was little detectable change for several years after installation of the dikes. Compared to nearby reference areas, soil moisture was 20-45% higher behind the dikes (except TW), even though ponded saturated infiltration rates were 20- 80% slower behind the dikes. Soil aggregate stability was consistently greater behind the dikes. Clay and silt soil fractions were greater behind the dikes generally, but one site (AT) had the opposite pattern. Differences in soil texture were likely influenced by transport processes and source areas. Percent cover of vegetation behind the dikes was 1.5-3 times the cover at nearby reference areas, but 20 times greater at one site (TW). Plant species richness behind the dikes was about double that in the reference areas, but 6 times greater at one site (TW). Mesquite (Prosopis spp.) and unpalatable shrubs were less dominant behind the dikes, while cover of small shrubs, grasses and forbs was greater than in reference areas. Fertilization with nitrogen and micronutrients increased vegetation cover behind dikes, but not in reference areas. The addition of native grass seed (black grama, Bouteloua eriopoda) allowed establishment of plants in some sandy areas at both dike and reference sites (mainly at DA), but only when other above ground vegetation was removed. Installation of water retention structures was sufficient to initiate the formation of islands of increased resource availability and increased biological productivity, but the development of soil and vegetation interactions was influenced by the local site characteristics, the landscape context, and the rainfall pattern.