Establishing criteria for protecting or improving wetland condition has often focused on physical and chemical factors, which can paint an incomplete picture of wetland quality. To protect the biological integrity of aquatic environments, identifying criteria based on biological responses to pollution is essential. We hypothesize that assessment of multiple taxonomic groups and response thresholds will provide statistically defensible effects-based methods to define reference condition and establish biological criteria. We used regression tree analysis to identify non-linear biological responses of three taxonomic groups (macrophytes, epiphytic diatoms, and plant-associated zooplankton) collected from 36 depressional wetlands in the Muskegon River watershed (Michigan, USA). Multi-metric biotic indices were developed for all three taxonomic groups and these indices were combined to reveal biologically relevant thresholds along a gradient of human disturbance. We found these three taxonomic groups responded at similar levels of impairment and could be used to classify wetlands into three groups: reference sites representing the highest quality wetlands in the landscape; slightly altered sites where the most sensitive organisms responded (sensitive plants, diatoms); and degraded sites where extensive changes in community structure occurred, which may reflect a shift to an alternate state. For the Muskegon River watershed, in particular, this analysis allowed us to identify sites in need of restoration, including approximately one-third of the depressional wetlands in the watershed. This study outlines a method for identifying criteria that could be used for regulatory purposes. In particular, we recommend the use of community-level metrics in identifying broad-based changes in community composition that may represent shifts to alternate states, as well as the use of sensitive indicators, such as the occurrence of sensitive plant and diatom taxa.