Given current threats to biodiversity, understanding the effects of diversity changes on the functions and services associated with intact ecosystems is of paramount importance. However, limited realism in most biodiversity studies makes it difficult to link the large and growing body of evidence for important functional consequences of biodiversity change to real-world losses of biodiversity. Here, we explored two methods of incorporating realism into biodiversity research: (1) the use of two-, five-, and eight-species assemblages that mimicked those that we observed in surveys of seaweed biodiversity patterns on a northern California (USA) rocky shore and the explicit comparison of those assemblages to random assemblages compiled from the same local species pool; and (2) the measurement of two fundamental ecosystem functions, nitrate uptake and photosynthesis, both of which contribute to growth of primary producers. Specifically, we measured nitrate uptake rates of seaweed assemblages as a function of initial nitrate concentrations and photosynthetic rates as a function of irradiance levels for both realistic and random assemblages of seaweeds. We only observed changes in ecosystem functioning along a richness gradient for realistic assemblages, and both maximum nitrate uptake rates (V(max)) and photosynthetic light use efficiency values (alpha(p) = P(max)/I(K)) were higher in realistic assemblages than in random assemblages. Furthermore, the parameter affected by changes in richness depended on the function being measured. Both V(max) and alpha(p) declined with increasing richness in nonrandom assemblages due to a combination of species identity effects (for V(max) and overyielding effects (for both V(max) and alpha(p)). In contrast, neither nitrate uptake efficiency values (alpha(N) = V(max)/K(s)), nor maximum photosynthetic rates (Pmax) changed along the gradient in seaweed species richness. Furthermore, overyielding was only evident in realistic assemblages, and the parameters exhibiting overyielding, including V(max), alpha(N), P(max), and alpha(p), changed along a gradient in species richness. Our results suggest that in realistic assemblages of species (1) some functions may be maximized at low levels of species richness, and (2) it is not only diversity, per se, that is important for sustaining multiple ecosystem functions, but also the range of diversity values in an ecosystem.