Ecological thresholds have been applied conventionally to characterize relationships between stressors and biological responses. We demonstrated how ecological thresholds can be used to assess recovery in a system where stressors have been removed. We applied a relatively new statistical approach, the significant zero crossings (SiZer) model, which approximates a response function and its derivatives and then shows how these functions vary across the range of an explanatory variable. We applied the model to a long-term macroinvertebrate data set collected from the Arkansas River, a Colorado stream undergoing restoration from historical mining pollution. SiZer analysis identified distinct threshold responses of benthic communities through time and related these shifts to improvements in water quality. With respect to restoration success, a recovery threshold was observed ∼10 y after clean up was initiated. Recovery trajectories in the Arkansas River were relatively complex, with multiple thresholds, and these responses were strongly influenced by temporal variation in heavy metal concentrations and stream discharge. A unique aspect of SiZer is that it easily handles multiple thresholds and displays significant change points along a predictor variable graphically. In addition, SiZer allows investigators to examine potential thresholds at different levels of resolution and provides a transparent representation of the threshold and how it responds to variability of the data.