In a salt marsh restoration effort that involved eight native halophytes, biomass increased with species richness, supporting biodiversity-ecosystem function (BEF) theory.  Randomly drawn sextets accumulated more biomass than solos, suggesting that diverse assemblages would persist indefinitely.  However, the site declined in richness in the ~10 years since planting (average richness was 2.8 species/0.25m2, below the reference datum of 4.5 species/0.25m2). BEF theory (based on random assemblages) predicted a decline in productivity with fewer complementary species. Diversity loss was nonrandom in the salt marsh, however: two short-lived plants died out, and two productive perennials became dominant. These shifts were consistent with competition and selection effects, not complementarity effects (positive interactions).  Further experimentation in the greenhouse indicated little evidence of complementary performance among the 8 halophytes.  All trios that included a productive species overyielded, but most of the biomass was contributed by a productive species, notably Salicornia viriginica.  Because S. virginica is a competitive dominant, we do not anticipate a loss in vascular plant productivity with further diversity loss in the field.  Whether or not other functions will decline is uncertain. In the southern California salt marsh, productivity can be sustained by an aggressive monotype, but other functions, such as resilience, arthropod support, and bird use might follow the classic BEF curve.  We recommend that BEF theory expand to address more of the critical functions of ecosystems in order to be more relevant to restoration ecology.