Natural microbial communities are species-rich assemblages that display ecosystem properties in the laboratory. Chemical perturbations are expected to alter species richness, standing crop biomass, nutrient pools, and community primary production and respiration. Experiments using microcosms developed from natural microbial communities on artificial substrata and continuously dosed toxicants (including heavy metals, simple organics, and pesticides) and complex mixtures (effluents) showed that microcosm variability was sufficiently low to detect adverse effects on species richness, standing crop biomass, and measures of ecological function. Experiments assessing heavy-metal toxicity, including effluent mixtures with heavy metals, showed adverse effects at toxicant levels at or below current water quality criteria. For example, copper reduced species richness and biomass at 10 to 20 μg/L in microcosms; zinc reduced chlorophyll biomass at <10 μg/L. Microcosm responses to toxicants differed both quantitatively and qualitatively from responses of standard test species and showed that compensatory mechanisms in complex assemblages sometimes prevented the expression of toxic effects, especially when degradable toxicants affected communities. Zinc toxicity elevated the rate of microbial phosphate recovery. Low levels of atrazine (<100 μg/L) stimulated species richness and standing crop biomass, but oxygen production did not increase. Chlorpyrifos did not affect microcosms even when concentrations exceeded water quality criteria by three orders of magnitude. Trinitrotoluene (TNT) only affected microcosms at >400 μg/L although provisional water quality criteria predicted chronic effects at 40 μg/L. Microcosms were sensitive to chlorine in the laboratory (6 μg/L) but were insensitive to a nutrient rich, chlorinated effluent (>400 μg/L), which was acutely toxic to daphnids. Reference toxicant experiments showed that effects on naturally derived microcosms are repeatable within and between ecosystems. Differences between microcosm responses and expected toxicity based on standard bioassays are a result of complex interactions including toxicant degradation, the interaction of toxicants and nutrients, and the lack of species sensitive to some toxicants, and these interactions allow microcosms to display a range of ecological responses not predictable from single species tests.